A construction scheme generation method and device, electronic equipment and storage medium

Abstract

The present application relates to the field of data processing, in particular to a construction scheme generation method and device, electronic equipment and storage medium. The present application firstly automatically extracts the key nodes, node descriptions and corresponding key indicators of the construction content, ensuring the accuracy of information extraction and changing the current situation that manual analysis is difficult to extract key links and set accurate indicators. Secondly, these key information is generated into structured data, providing clear and standardized input for large language models, so that they can generate preliminary construction planning based on data, improving the effectiveness of information integration. Thirdly, by dynamically correcting the predicted construction planning and the construction standard, the parts that do not meet the standard are found and corrected in time, greatly reducing the probability that the construction planning does not meet the standard, and avoiding the delay of construction period and the increase of cost caused by unreasonable scheme.

Description

Technical Field

[0001] This invention relates to the field of data processing, and specifically to a method, apparatus, electronic device, and storage medium for generating construction plans. Background Technology

[0002] In the construction industry, the formulation of traditional construction plans often relies on manual experience, which not only consumes a lot of manpower and time, but is also limited by the individual knowledge of engineers, making it difficult to comprehensively and scientifically consider all aspects of the construction project. When undertaking a project, the construction company needs to collect information on the construction content and standards, but relying on manual analysis makes it difficult to accurately extract the key construction links and ensure the accurate setting of key indicators. During the construction plan generation phase, existing methods cannot effectively integrate key construction information, resulting in unreasonable construction plan planning and frequent discrepancies between the construction plan and standards during construction. This not only causes delays in the construction period but also significantly increases construction costs. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a method, apparatus, electronic device and storage medium for generating construction plans, in order to solve the problem that existing methods cannot effectively integrate key construction information, resulting in unreasonable construction plan planning and frequent discrepancies between construction plans and standards during construction, which not only cause delays in the construction period but also significantly increase construction costs.

[0004] In a first aspect, embodiments of the present invention provide a method for generating a construction plan, the method comprising: Obtain construction requirements uploaded by users, wherein the construction requirements include the construction content and construction standards of the construction project; The construction content is analyzed to obtain each key construction node and node description, and key indicators corresponding to the key construction nodes are generated based on the construction standards. Structured data is generated using the aforementioned key construction nodes and key indicators; The structured data is input into the trained large language model so that the large language model can make predictions based on the structured data to obtain a predicted construction plan, and then dynamically correct the predictions based on the predicted construction plan and construction standards to obtain a construction scheme.

[0005] Furthermore, the analysis of the construction content yields various key construction nodes and node descriptions, including: Obtain the project type corresponding to the construction project; The construction content is analyzed using the knowledge graph corresponding to the project type to obtain the key construction nodes; The node description is generated based on the node attributes of the key construction nodes and their associated paths in the knowledge graph.

[0006] Furthermore, the analysis of the construction content using the knowledge graph corresponding to the project type to obtain key construction nodes includes: The construction content is broken down into multiple sub-tasks and task descriptions corresponding to the sub-tasks, and keywords corresponding to the sub-tasks are extracted from the task descriptions. Using the keywords, match the corresponding candidate nodes from the knowledge graph corresponding to the project type; If there is only one candidate node, then the candidate node is designated as the key construction node; or, if there is more than one candidate node, then the key construction node is determined from the candidate nodes by using the relationship weights and upstream / downstream logic between nodes in the knowledge graph.

[0007] Furthermore, the step of inputting the structured data into the trained large language model, so that the large language model can make predictions based on the structured data to obtain a predicted construction plan, and dynamically correcting the predicted construction plan and construction standards to obtain a construction scheme, includes: The prediction network in the large language model predicts the initial construction plan for key construction nodes based on node descriptions in structured data. The initial construction plan is integrated by the fusion network in the large language model based on the logical relationship between the key construction nodes, and the sequence and connection process of each node are sorted out to obtain the intermediate construction plan. The intermediate construction plan is compared with the construction standard, and the contents that do not meet the standard are marked. The instructions are then reorganized and input into the large language model based on the marked contents in the intermediate construction plan, guiding the large language model to adjust the intermediate construction plan until the final construction plan matches the construction standard.

[0008] Furthermore, the step of predicting the initial construction plan for key construction nodes based on node descriptions in structured data using the prediction network in the large language model includes: The prediction network parses the node descriptions in the structured data and extracts key features for each critical construction node. Obtain similar construction content in scenarios similar to the described key construction nodes; The key features are matched with similar construction content to obtain an initial construction plan corresponding to each key construction node.

[0009] Furthermore, the step of comparing the intermediate construction plan with the construction standard, marking the content that does not meet the standard, and reorganizing the input instructions into the large language model based on the marked content in the intermediate construction plan, guiding the large language model to adjust the intermediate construction plan until the final construction plan matches the construction standard, includes: By comparing the intermediate construction plan with the construction standard, the abnormal content in the intermediate construction plan that does not match the construction standard is obtained; Analyze whether the aforementioned anomalies are feasible within the intermediate construction plan; If it is not feasible, the abnormal content is marked to obtain the marked content; According to the instruction rules of the large language model, the marked content and the construction standards are reorganized into instruction text; The instruction text is input into the large language model, which continuously optimizes and adjusts the marked content according to the construction standards until every item of the construction plan matches the construction standards.

[0010] Furthermore, the method also includes: Obtain the risk data associated with the construction project; The risk data is used to assess the potential safety risks currently existing in the construction plan; Obtain the corresponding response strategies for the potential security risks, and deploy them at the construction site corresponding to the construction project in accordance with the response strategies.

[0011] Secondly, embodiments of the present invention provide a construction plan generation device, the device comprising: The acquisition module is used to acquire construction requirements uploaded by users, wherein the construction requirements include the construction content and construction standards of the construction project; The analysis module is used to analyze the construction content to obtain each key construction node and node description, and to generate key indicators corresponding to the key construction nodes based on the construction standards. A generation module is used to generate structured data using the key construction nodes and key indicators; The prediction module is used to input the structured data into the trained large language model, so that the large language model can make predictions based on the structured data to obtain a predicted construction plan, and dynamically correct the predicted construction plan and construction standards to obtain a construction scheme.

[0012] Thirdly, embodiments of the present invention provide an electronic device, including: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the method described in the first aspect or any corresponding embodiment thereof.

[0013] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer instructions for causing a computer to perform the method described in the first aspect or any corresponding embodiment thereof.

[0014] This application first automatically extracts key nodes, node descriptions, and corresponding key indicators of the construction content, ensuring the accuracy of information extraction and changing the current situation where manual analysis struggles to extract key aspects and set accurate indicators. Second, it generates structured data from this key information, providing clear and standardized input for a large language model, enabling it to generate preliminary construction plans based on the data and improving the effectiveness of information integration. Third, by dynamically correcting the predicted construction plan against construction standards, it promptly identifies and corrects any non-compliance, greatly reducing the probability of discrepancies between the construction plan and standards, and avoiding delays and increased costs due to unreasonable solutions. Attached Figure Description

[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a method for generating a construction plan according to some embodiments of the present invention; Figure 2 This is a schematic diagram of the structure of a large language model according to some embodiments of the present invention; Figure 3 This is a structural block diagram of a construction scheme generation device according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] According to embodiments of the present invention, a method, apparatus, electronic device, and storage medium for generating a construction plan are provided. It should be noted that the steps shown in the flowcharts in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0019] This embodiment provides a method for generating a construction plan. Figure 1 This is a flowchart of a method for generating a construction plan according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain the construction requirements uploaded by the user, wherein the construction requirements include the construction content and construction standards of the construction project.

[0020] In this embodiment, a dedicated online platform, such as a web page or mobile application, is developed to provide an entry point for users to upload construction requirements. After a user uploads their construction requirements, the system parses the uploaded data. Natural language processing technology is used to classify and extract the text information, separating the construction content and construction standards. If the user uploads a document, the system uses a document parsing tool to extract the necessary information, ensuring that the obtained construction content and standard information are complete and accurate.

[0021] Step S102: Analyze the construction content to obtain each key construction node and node description, and generate key indicators corresponding to the key construction nodes based on construction standards.

[0022] In this embodiment of the application, the analysis of the construction content yields various key construction nodes and node descriptions, including the following steps A1-A3: Step A1: Obtain the project type corresponding to the construction project.

[0023] Specifically, a pre-established database covering various construction project types is accessed, such as civil buildings, industrial buildings, municipal engineering, and transportation engineering, and each type is categorized and defined in detail. Project type identification involves text analysis of the construction content to extract keywords, such as "residential construction," "road repair," and "bridge construction," and these keywords are matched with the project type database to determine the specific type of the construction project.

[0024] Step A2: Analyze the construction content using the knowledge graph corresponding to the project type to obtain the key construction nodes.

[0025] Specifically, based on the determined project type, the corresponding knowledge graph is invoked. A knowledge graph is a structured knowledge representation method that includes various entities and relationships related to the project type during the construction process, such as construction procedures, materials used, and equipment involved. Natural language processing technology is used to segment and semantically understand the construction content, matching key information within the content with nodes in the knowledge graph. For example, if the construction content mentions "foundation excavation," the corresponding "foundation excavation" node is found in the knowledge graph through matching, and related construction processes and technical requirements are extracted.

[0026] Based on the relationships between nodes in the knowledge graph, the key processes and steps of construction are identified, and critical construction nodes are determined. Simultaneously, nodes irrelevant to the core construction content are excluded to ensure that the extracted critical nodes accurately reflect the main steps of construction.

[0027] Step A3: Generate node descriptions based on the node attributes of key construction nodes and their associated paths in the knowledge graph.

[0028] Specifically, the construction content is analyzed using a knowledge graph corresponding to the project type to identify key construction nodes. This includes breaking down the construction content into multiple sub-tasks and their corresponding task descriptions, and extracting keywords from these descriptions. These keywords are then used to match relevant candidate nodes from the knowledge graph corresponding to the project type. If there is only one candidate node, it is designated as a key construction node. Alternatively, if there are more than one candidate node, the key construction node is determined from the candidate nodes based on the relationship weights between nodes in the knowledge graph and the upstream and downstream logic.

[0029] It should be noted that the construction content text uploaded by users undergoes cleaning to remove noise information, such as irrelevant punctuation marks, special characters, and whitespace, making the text more standardized and easier to process. Syntactic and semantic analysis techniques from natural language processing are used to analyze the construction content text. Based on the semantic structure and logical relationships of the text, the construction content is divided into multiple relatively independent sub-tasks. For example, a building construction project might be broken down into sub-tasks such as "foundation treatment," "main structure construction," and "decoration." Simultaneously, a corresponding task description is generated for each sub-task, detailing its specific content and requirements.

[0030] For each subtask's task description, keyword extraction algorithms, such as TF-IDF (Term Frequency-Inverse Document Frequency) and TextRank, are used. These algorithms select keywords that accurately represent the core content of the subtask based on the frequency of words in the task description and their importance in the entire corpus. For example, the description of the subtask "foundation treatment" as "using dynamic compaction to reinforce the foundation and ensure that the foundation bearing capacity meets design requirements" might extract keywords such as "dynamic compaction," "foundation reinforcement," and "foundation bearing capacity."

[0031] For the knowledge graph corresponding to the project type, an efficient index structure is pre-established to facilitate fast node querying and matching. Graph database indexing techniques can be used, such as node attribute-based indexing and edge relationship-based indexing. Extracted keywords are matched against node names, attributes, and other information in the knowledge graph. String matching algorithms, such as exact matching and fuzzy matching, are used to find nodes related to the keywords. For example, if the keyword is "forced compaction method," then the knowledge graph is searched for nodes whose names or attributes contain "forced compaction method." Matched nodes are then treated as candidate nodes.

[0032] After obtaining candidate nodes, check the number of candidate nodes. If the number is exactly one, it means that a unique corresponding node has been found through keyword matching. This unique candidate node is identified as the key construction node, because it can accurately represent the position and meaning of the subtask in the knowledge graph, which is of great importance for subsequent construction plan generation and analysis.

[0033] For multiple candidate nodes, analyze their relationship weights with other nodes in the knowledge graph. Relationship weights reflect the closeness and importance between nodes. For example, edges connecting certain nodes may have weight values, indicating their business relevance or technological dependence. The importance of each candidate node is assessed by calculating the sum of its relationship weights with other relevant nodes.

[0034] By analyzing the upstream and downstream logical relationships of nodes in the knowledge graph, the position and sequence of subtasks within the overall construction process are determined. The roles of candidate nodes in their upstream and downstream logics are analyzed to determine which nodes are prerequisites for subtasks and which are subsequent dependencies. For example, if the subtask is "main structure construction," the "foundation treatment" node in the knowledge graph might be a prerequisite, while the "decoration" node might be a subsequent node. Based on the upstream and downstream logic, candidate nodes that best match the logical sequence of the subtasks and the construction process are selected as key construction nodes.

[0035] Taking into account both relationship weights and upstream / downstream logic, multiple candidate nodes are compared and evaluated. Nodes with higher relationship weights and conforming to upstream / downstream logic are selected as key construction nodes. If multiple candidate nodes perform similarly in terms of relationship weights and upstream / downstream logic, other factors, such as node attribute information and their importance in actual construction, can be further considered to make a final decision.

[0036] Step S103: Generate structured data using key construction nodes and key indicators.

[0037] In this application embodiment, the format used for data storage and processing is determined, commonly including JSON and XML. Taking JSON as an example, it is characterized by its simplicity, readability, ease of parsing, and transmission. The structured data framework is designed, typically including a root node and child nodes for storing key construction nodes and key indicators, respectively. For example, the root node can be named "construction_plan_data", which contains two child nodes: "construction_key_nodes" (key construction nodes) and "key_indicators" (key indicators).

[0038] For each key construction node, collect and organize its relevant information, such as node name, node description, construction stage, required resources (manpower, materials, equipment, etc.), and estimated duration. Organize this information into key-value pairs to form a node object. Compile all the node objects for all key construction nodes into a list and store it under the "construction_key_nodes" sub-node.

[0039] For each key indicator corresponding to a critical construction milestone, specify the indicator name, indicator type (such as quality indicator, safety indicator, schedule indicator, etc.), indicator value (or value range), and unit of measurement.

[0040] Similarly, key indicator information is organized in key-value pairs to form indicator objects. All key indicator objects for all key construction milestones are compiled into a list and stored under the "key_indicators" sub-node.

[0041] The compiled list of key construction nodes and key indicators are integrated into the root node "construction_plan_data" according to the pre-designed structured data structure framework.

[0042] Step S104: Input the structured data into the trained large language model so that the large language model can make predictions based on the structured data to obtain a predicted construction plan, and then dynamically correct the deviation based on the predicted construction plan and construction standards to obtain a construction scheme.

[0043] In this embodiment, structured data is input into a trained large language model so that the large language model can make predictions based on the structured data to obtain a predicted construction plan. Then, based on the predicted construction plan and construction standards, dynamic correction is performed to obtain a construction scheme, including the following steps B1-B3: Step B1 involves using the prediction network in the large language model to predict the initial construction plan for key construction nodes based on the node descriptions in the structured data.

[0044] Specifically, such as Figure 2 As shown, the initial construction plan for key construction nodes is predicted based on node descriptions in structured data using a prediction network in a large language model. This includes: the prediction network parsing the node descriptions in the structured data and extracting key features for each key construction node; obtaining similar construction content in scenarios similar to the key construction node descriptions; and matching the key features with the similar construction content to obtain the initial construction plan corresponding to each key construction node.

[0045] Predictive networks employ natural language processing techniques, such as syntactic analysis and semantic analysis, to parse the node descriptions of each key construction node in structured data. Syntactic analysis determines the sentence structure and grammatical relationships within the text; semantic analysis helps to understand the meaning expressed by the text.

[0046] Key features are extracted based on the analysis results. These key features can be keywords or phrases describing construction methods, materials used, construction sequence, quality requirements, etc. For example, the description of the "foundation concrete pouring" node as "using pumping method, C30 concrete, poured in layers and vibrated to compact" may extract key features such as "pumping method", "C30 concrete", "layered pouring", and "vibrated to compact".

[0047] Collect and organize a large amount of construction content data from various construction projects in advance to build a comprehensive construction content database. This data should include detailed descriptions of various construction scenarios and stages.

[0048] For each key construction node description, a similarity calculation algorithm (such as cosine similarity, edit distance, etc.) is used to compare it with construction content in the database. By calculating the similarity score between the node description and the construction content text in the database, construction content with high similarity is selected as similar construction content in similar scenarios. For example, if a key construction node is described as "exterior wall insulation construction of high-rise residential buildings," similarity calculation is used to find relevant content in the database that is also related to high-rise residential buildings and involves exterior wall insulation construction.

[0049] The calculated similar construction content is further screened and organized to remove content that is irrelevant or has low similarity to the current key construction node, ensuring that the obtained similar construction content has high relevance and reference value. The key features of each extracted key construction node are matched one-to-one with the obtained similar construction content. Matching algorithms (such as exact matching, fuzzy matching, etc.) are used to determine whether the key features exist in the similar construction content and their frequency and location. For example, key features such as "pumping method" and "C30 concrete" are matched with similar concrete pouring construction content.

[0050] Based on the matching results, information related to key features, such as construction steps, technical requirements, and resource allocation, is extracted from similar construction content and integrated to form an initial construction plan for each key construction node. For each key node, the initial construction plan should include, but is not limited to, the construction process, required resources (manpower, materials, equipment, etc.), construction schedule, and quality control measures. For example, if detailed construction process and resource allocation information for "foundation concrete pouring" are matched in similar construction content, this information is organized into a specific initial construction plan, including steps such as formwork installation, concrete pumping, layered pouring, and vibration, as well as the required resources such as concrete workers and vibration equipment.

[0051] The initial construction plan for each key construction node is optimized, and its rationality and feasibility are checked. Simultaneously, considering the logical relationships and sequence between key construction nodes, the initial construction plans for each node are integrated to form a coherent and unified initial construction plan, providing a foundation for further improvement and adjustment of subsequent construction plans.

[0052] Step B2 involves integrating the initial construction plan based on the logical relationships between key construction nodes using the fusion network in the large language model, sorting out the sequence and connection process of each node, and obtaining the intermediate construction plan.

[0053] Specifically, based on the descriptions of key construction nodes, logical relationships such as sequence, parallelism, and causality between nodes are extracted. For example, for the two nodes "foundation excavation" and "foundation pouring," the model determines the sequence of "foundation excavation" preceding "foundation pouring" based on text descriptions and training experience. Furthermore, to make the logical relationships clearer, a logical relationship graph is established to graphically display each node and its interrelationships.

[0054] The initial construction plan corresponding to each critical construction node is transformed into structured data. Information such as construction steps, resource requirements, and timelines within the plan is clearly defined. For example, the "main structure construction" plan is broken down into specific steps such as scaffolding erection, rebar tying, formwork erection, and concrete pouring, with the required manpower, materials, and estimated time for each step determined. Each construction step is labeled with its corresponding critical construction node so that the fusion network can operate based on the logical relationships between nodes when integrating the plans.

[0055] The fusion network in this embodiment can be based on a neural network, such as a recurrent neural network (RNN) or a Transformer architecture, to handle sequential data and capture complex relationships between nodes. The fusion network is trained to learn how to integrate construction plans based on the logical relationships between nodes under different construction scenarios.

[0056] The logical relationship diagram of key construction nodes and the structured data of the initial construction plan are input into the fusion network. Based on the logical relationships between nodes, the fusion network reorders and combines the construction steps in the initial plan. For example, if two nodes are parallel, the fusion network will rationally arrange their positions in the plan so that they can be performed simultaneously; if they are sequential, they will be arranged in order. Simultaneously, resource requirements and time arrangements are coordinated to avoid conflicts and waste.

[0057] A verification algorithm can also be established to check the integrated intermediate construction plan. Verification includes the rationality of construction steps, the balance of resource allocation, and the feasibility of the time schedule. For example, it checks whether a construction step lacks prerequisites or whether resource requirements exceed actual supply capacity. If the intermediate construction plan meets the requirements after verification, it is output to provide a basis for subsequent comparison with construction standards and further optimization. If problems are found, feedback is given to the fusion network for adjustment and improvement.

[0058] Step B3 compares the intermediate construction plan with the construction standard, marks the content that does not meet the standard, and reorganizes the input instructions into the large language model based on the marked content in the intermediate construction plan, guiding the large language model to adjust the intermediate construction plan until the final construction plan matches the construction standard.

[0059] Specifically, the intermediate construction plan is compared with the construction standard, and any non-compliance is marked. The marked content in the intermediate construction plan is then reorganized and input into the large language model as instructions. This guides the large language model to adjust the intermediate construction plan until the final construction plan matches the construction standard. This process includes: comparing the intermediate construction plan and the construction standard to identify any non-compliance in the intermediate construction plan; analyzing the feasibility of the non-compliance within the intermediate construction plan; marking the non-compliance if it is not feasible; reorganizing the marked content and the construction standard into an instruction text according to the large language model's instruction rules; and inputting the instruction text into the large language model, which continuously optimizes and adjusts the marked content according to the construction standard until every element of the final construction plan matches the construction standard.

[0060] The intermediate construction plans and standards are structured to ensure consistency in data format and hierarchical structure, facilitating subsequent comparisons. For example, construction standards are categorized by construction stage and project, corresponding one-to-one with the relevant parts of the intermediate construction plans. Detailed comparison rules are developed based on expertise in the construction field. For instance, regarding construction techniques, standard operating procedures for different construction steps are clearly defined; regarding material usage, standard requirements for material specifications, models, and performance are specified.

[0061] Based on the comparison rules, each item in the intermediate construction plan and construction standards is compared one by one. Using text matching algorithms and numerical comparison algorithms, inconsistencies are identified and determined as abnormal content. For example, in building wall construction, if the masonry technique used in the construction plan does not conform to the standard specifications, or if the bricks used do not meet the standard requirements, these will be identified as abnormal content.

[0062] A feasibility assessment network is trained based on historical construction data, engineering cases, and expert experience. This model can determine the feasibility of anomalies in actual construction based on factors such as the construction scenario, resource conditions, and technical requirements. The feasibility assessment network is provided with necessary information, including the actual conditions of the construction site (such as site space and geological conditions), available resources (human, material, and financial resources), and the current level of construction technology.

[0063] The identified anomalous content is input into the feasibility assessment network, and combined with the collected assessment criteria, its feasibility is determined. If the anomalous content cannot be implemented under existing conditions, or if its implementation may lead to quality or safety issues, it is deemed infeasible.

[0064] For any abnormal content that is deemed infeasible after assessment, mark it in the intermediate construction plan. The marking method should be clear and unambiguous to facilitate subsequent extraction and processing. Extract the marked content from the intermediate construction plan and obtain the corresponding standard requirements from the construction standards.

[0065] Following the instruction rules of the large language model, the marked content and construction standards are integrated to form instruction text. The instruction text should clearly describe the content that needs to be optimized and the expected standard, such as "Adjust [the marked masonry process] to conform to [the masonry process specified in the construction standard]".

[0066] The organized instruction text is input into the large language model, triggering the model to optimize and adjust the intermediate construction plan. Based on the instruction text, the large language model uses its learned construction knowledge and language logic to modify and improve the marked content, generating a new version of the construction plan. The newly generated construction plan is then compared with the construction standards again, repeating the process of anomaly identification, feasibility analysis, instruction text organization, and model optimization until every aspect of the construction plan matches the construction standards.

[0067] Based on this, in the initial construction plan generation stage, the predictive network deeply analyzes the node descriptions of structured data, extracts key features, and matches construction content with similar scenarios. This process draws on a wealth of past experience, overcoming the limitations of traditional solutions that rely on single experiences, making the initial plans for each key construction node more practical and scientific. Subsequently, the fusion network integrates the initial plans based on the logical relationships between key construction nodes, streamlining a reasonable construction sequence and connection process, avoiding chaotic construction steps, significantly improving construction efficiency, and reducing resource waste caused by unreasonable processes. In the construction plan optimization stage, intermediate construction plans are compared with construction standards to accurately identify mismatches and perform feasibility analysis. After marking infeasible anomalies, instruction text is generated according to the large language model's instruction rules, prompting the model to continuously optimize the plan. This dynamic correction mechanism ensures that every aspect of the construction plan conforms to standards, reducing the risk of rework due to discrepancies between the plan and standards, and avoiding project delays and increased costs.

[0068] As an example, suppose in a high-rise residential building project, the construction team identifies an anomaly: the original design requires the use of a new glass curtain wall installation technique on the exterior walls of the 20th floor, but this technique is rarely used in local construction cases.

[0069] During the data collection phase, construction data from 100 similar high-rise residential construction projects were gathered, including technical challenges encountered during curtain wall installation, resource allocation, and final results. Simultaneously, 50 engineering case studies involving the application of new construction techniques were compiled, and five industry experts were invited to share their experience in handling complex curtain wall installation processes. Based on this data, a feasibility assessment network based on a multilayer perceptron (MLP) was built using the TensorFlow deep learning framework. The data was divided into training, validation, and test sets at a ratio of 70%, 20%, and 10% respectively for model training, validation, and optimization.

[0070] Regarding the assessment data collection, on-site surveys revealed limited space at the construction site, restricting the parking and operation of large hoisting equipment; however, the geological conditions were stable, with minimal impact on the main building construction. In terms of available resources, the construction team had only two workers with experience in similar new construction techniques, and the budget was also tight. Regarding existing construction technology, the company was relatively mature in conventional glass curtain wall installation technology, but its mastery of this new technique was still limited.

[0071] The abnormal content and collected evaluation criteria are input into a pre-trained feasibility assessment network. The network first extracts features from the input data, transforming it into a vector form that the model can process. Then, each hidden layer of the multilayer perceptron performs a non-linear transformation on the input features using a weight matrix to uncover potential relationships within the data. Finally, the output layer provides a feasibility probability value between 0 and 1. In this example, the model outputs a low probability value. Considering the actual situation, due to insufficient worker experience and technical expertise, and the potential impact of site space limitations on construction efficiency and safety risks, implementing this new glass curtain wall installation process may lead to unreliable construction quality and even safety accidents. Therefore, it is determined that the abnormal content is not feasible under the current conditions.

[0072] This application first automatically extracts key nodes, node descriptions, and corresponding key indicators of the construction content, ensuring the accuracy of information extraction and changing the current situation where manual analysis struggles to extract key aspects and set accurate indicators. Second, it generates structured data from this key information, providing clear and standardized input for a large language model, enabling it to generate preliminary construction plans based on the data and improving the effectiveness of information integration. Third, by dynamically correcting the predicted construction plan against construction standards, it promptly identifies and corrects any non-compliance, greatly reducing the probability of discrepancies between the construction plan and standards, and avoiding delays and increased costs due to unreasonable solutions.

[0073] In this embodiment, the method further includes: acquiring risk data associated with the construction project; using the risk data to assess the potential safety risks currently existing in the construction plan; acquiring corresponding countermeasures for the potential safety risks; and deploying these countermeasures at the construction site corresponding to the construction project.

[0074] Specifically, construction companies should establish comprehensive data recording systems to collect data on various risk events that have occurred in past construction projects, including the time, location, type (such as falls from heights, falling objects, electrical faults, etc.), losses caused, and causes. For example, after each safety accident, a detailed record of the accident process and related information should be kept. Data sharing mechanisms should be established with industry associations, government regulatory departments, and safety research institutions to obtain industry-standard risk data and statistical information. Web scraping technology can also be used to collect information related to construction safety risks from relevant safety news websites and professional forums.

[0075] Various monitoring devices, such as cameras and sensors (used to monitor temperature, humidity, pressure, vibration, etc.), are installed at the construction site to collect environmental and equipment operation data that may pose safety risks in real time. For example, sensors can be used to monitor the deformation of scaffolding to detect potential collapse risks in advance.

[0076] The collected risk data is cleaned to remove duplicates, errors, or incomplete data. The data is then categorized and organized according to certain classification criteria, such as risk type, construction stage, or location of occurrence. The organized data is stored in a dedicated database for later retrieval and use. A relational database (such as MySQL) or a non-relational database (such as MongoDB) can be used, choosing the appropriate storage method based on the data characteristics and usage requirements.

[0077] A risk assessment model is constructed based on historical risk data and expert experience. A combination of qualitative and quantitative methods can be used, such as the Analytic Hierarchy Process (AHP), Fault Tree Analysis (FTA), and Bayesian networks. The assessment indicators and weights in the model are determined. For example, the skill level of construction personnel, the condition of equipment, and the complexity of the construction environment can be used as assessment indicators, and appropriate weights can be assigned based on their impact on safety risks.

[0078] Relevant information about the construction plan (such as construction process, equipment and materials used, and personnel arrangements) and collected risk data are input into the risk assessment model. The model analyzes the construction plan according to the established algorithms and rules, calculating the risk value of each risk factor and the overall potential safety risk level. For example, based on the frequency of work at heights in the construction plan and the historical probability of falls from heights, the risk value of the work at heights segment is assessed.

[0079] A detailed analysis of the assessment results should be conducted to identify areas and factors with high potential safety risks. A risk assessment report should be generated, which should include the risk assessment methods, results, risk factor analysis, and suggestions for improving the construction plan.

[0080] Organize construction safety experts, technicians, and managers to develop response strategies for different types of safety risks based on historical experience in handling risk events and industry standards. Compile these strategies into a database, categorize them, and manage them for easy retrieval and use. The response strategy database should be continuously updated and improved to adapt to newly emerging risk situations.

[0081] Based on the risk assessment results, search the response strategy library for corresponding countermeasures to potential safety risks. Matching can be done using keywords such as risk type and risk level. For complex risk situations, multiple response strategies may need to be combined to effectively mitigate the risk. For example, for construction areas with both fire and electrical fault risks, it may be necessary to simultaneously implement fire prevention measures and electrical equipment maintenance and inspection measures.

[0082] Based on the specific circumstances of the construction project (such as site conditions, schedule requirements, and resource allocation), the matched response strategies are adjusted and optimized to develop a customized solution suitable for the project. The solution clearly defines the responsible parties and timelines for each measure to ensure effective execution of the response strategies.

[0083] Develop a detailed deployment plan based on the individualized response plan. The plan should include the implementation steps for each response measure, the required resources (human, material, and financial resources), and the timeline. Based on the deployment plan, prepare the necessary resources, such as safety equipment (helmets, safety belts, safety nets, etc.), fire-fighting equipment, and maintenance tools. Train construction personnel to familiarize them with the response strategies and operational requirements. Training content may include safety knowledge, emergency response skills, and equipment operation methods.

[0084] According to the deployment plan, various response measures are being implemented step by step at the construction site. For example, guardrails and warning signs are being installed in areas where work is being carried out at heights, electrical equipment is being regularly inspected and maintained, and fire safety drills are being organized for construction workers. A monitoring mechanism has been established to track and inspect the implementation of these measures to ensure that all measures are effectively implemented.

[0085] Regularly evaluate the effectiveness of implemented countermeasures by examining indicators such as the incidence of safety incidents and changes in risk factors to determine the effectiveness of the response strategies. If the countermeasures are found to be ineffective or new risks emerge, adjust and optimize the plan in a timely manner and redeploy the countermeasures.

[0086] This embodiment also provides a construction plan generation device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0087] This embodiment provides a device for generating construction plans, such as... Figure 3 As shown, it includes: The acquisition module 301 is used to acquire the construction requirements uploaded by the user, wherein the construction requirements include the construction content and construction standards of the construction project; Analysis module 302 is used to analyze the construction content to obtain each key construction node and node description, and to generate key indicators corresponding to the key construction nodes based on construction standards. The generation module 303 is used to generate structured data using key construction nodes and key indicators; The prediction module 304 is used to input structured data into the trained large language model so that the large language model can make predictions based on the structured data, obtain a predicted construction plan, and dynamically correct the construction plan and construction standards to obtain a construction scheme.

[0088] In this embodiment, the analysis module 302 is used to obtain the project type corresponding to the construction project; analyze the construction content using the knowledge graph corresponding to the project type to obtain the key construction nodes; and generate node descriptions based on the node attributes of the key construction nodes and the association paths of the key construction nodes in the knowledge graph.

[0089] In this embodiment of the application, the analysis module 302 is used to decompose the construction content into multiple sub-tasks and task descriptions corresponding to the sub-tasks, and extract keywords corresponding to the sub-tasks from the task descriptions; use the keywords to match corresponding candidate nodes from the knowledge graph corresponding to the project type; if the number of candidate nodes is 1, then the candidate node is taken as the construction key node; or, if the number of candidate nodes is greater than 1, then the construction key node is determined from the candidate nodes through the relationship weights between nodes in the knowledge graph and the upstream and downstream logic.

[0090] In this embodiment, the prediction module 304 is used to predict the initial construction plan of key construction nodes based on the node descriptions in the structured data through the prediction network in the large language model; to integrate the initial construction plan based on the logical relationship between the key construction nodes through the fusion network in the large language model, sort out the sequence and connection process of each node, and obtain the intermediate construction plan; to compare the intermediate construction plan with the construction standard, mark the content that does not meet the standard, and reorganize the instructions input into the large language model according to the marked content in the intermediate construction plan, so as to guide the large language model to adjust the intermediate construction plan until the final construction plan matches the construction standard.

[0091] In this embodiment, the prediction module 304 is used to predict the network to parse the node descriptions in the structured data, extract the key features of each construction key node; obtain similar construction content in scenarios similar to the construction key node descriptions; and match the key features with the similar construction content to obtain the initial construction plan corresponding to each construction key node.

[0092] In this embodiment, the prediction module 304 is used to compare the intermediate construction plan and the construction standard to obtain abnormal content in the intermediate construction plan that does not match the construction standard; analyze whether the abnormal content is feasible in the intermediate construction plan; if it is not feasible, mark the abnormal content to obtain marked content; reorganize the marked content and the construction standard into an instruction text according to the instruction rules of the big language model; input the instruction text into the big language model, and the big prediction model continuously optimizes and adjusts the marked content according to the construction standard until every item of the construction plan matches the construction standard.

[0093] In this embodiment of the application, the device further includes: an evaluation module, used to acquire risk data associated with the construction project; to evaluate the potential safety risks currently existing in the construction plan using the risk data; to acquire the corresponding response strategies for the potential safety risks; and to deploy the response strategies at the construction site corresponding to the construction project.

[0094] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an optional embodiment of the present invention, such as... Figure 4As shown, the electronic device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise as required. The processors can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system).

[0095] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0096] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0097] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device as displayed on a mini-program landing page. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0098] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0099] The electronic device also includes a communication interface 30 for communicating with other devices or communication networks.

[0100] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0101] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for generating a construction plan, characterized in that, The method includes: Obtain construction requirements uploaded by users, wherein the construction requirements include the construction content and construction standards of the construction project; The construction content is analyzed to obtain each key construction node and node description, and key indicators corresponding to the key construction nodes are generated based on the construction standards. Structured data is generated using the aforementioned key construction nodes and key indicators; The structured data is input into the trained large language model so that the large language model can make predictions based on the structured data to obtain a predicted construction plan, and then dynamically correct the predictions based on the predicted construction plan and construction standards to obtain a construction scheme.

2. The method according to claim 1, characterized in that, The analysis of the construction content yields key construction nodes and node descriptions, including: Obtain the project type corresponding to the construction project; The construction content is analyzed using the knowledge graph corresponding to the project type to obtain the key construction nodes; The node description is generated based on the node attributes of the key construction nodes and their associated paths in the knowledge graph.

3. The method according to claim 2, characterized in that, The analysis of the construction content using the knowledge graph corresponding to the project type to obtain key construction nodes includes: The construction content is broken down into multiple sub-tasks and task descriptions corresponding to the sub-tasks, and keywords corresponding to the sub-tasks are extracted from the task descriptions. Using the keywords, match the corresponding candidate nodes from the knowledge graph corresponding to the project type; If there is only one candidate node, then the candidate node is designated as the key construction node; or, if there is more than one candidate node, then the key construction node is determined from the candidate nodes by using the relationship weights and upstream / downstream logic between nodes in the knowledge graph.

4. The method according to claim 1, characterized in that, The step of inputting the structured data into a trained large language model, enabling the large language model to make predictions based on the structured data to obtain a predicted construction plan, and then dynamically correcting the predicted construction plan against construction standards to obtain a construction scheme, includes: The prediction network in the large language model predicts the initial construction plan for key construction nodes based on node descriptions in structured data. The initial construction plan is integrated by the fusion network in the large language model based on the logical relationship between the key construction nodes, and the sequence and connection process of each node are sorted out to obtain the intermediate construction plan. The intermediate construction plan is compared with the construction standard, and the contents that do not meet the standard are marked. The instructions are then reorganized and input into the large language model based on the marked contents in the intermediate construction plan, guiding the large language model to adjust the intermediate construction plan until the final construction plan matches the construction standard.

5. The method according to claim 4, characterized in that, The initial construction plan for predicting key construction nodes based on node descriptions in structured data using the prediction network in the large language model includes: The prediction network parses the node descriptions in the structured data and extracts key features for each critical construction node. Obtain similar construction content in scenarios similar to the described key construction nodes; The key features are matched with similar construction content to obtain an initial construction plan corresponding to each key construction node.

6. The method according to claim 4, characterized in that, The process of comparing the intermediate construction plan with the construction standard, marking the content that does not meet the standard, and reorganizing the input instructions into the large language model based on the marked content in the intermediate construction plan, guiding the large language model to adjust the intermediate construction plan until the final construction plan matches the construction standard, includes: By comparing the intermediate construction plan with the construction standard, the abnormal content in the intermediate construction plan that does not match the construction standard is obtained; Analyze whether the aforementioned anomalies are feasible within the intermediate construction plan; If it is not feasible, the abnormal content is marked to obtain the marked content; According to the instruction rules of the large language model, the marked content and the construction standards are reorganized into instruction text; The instruction text is input into the large language model, which continuously optimizes and adjusts the marked content according to the construction standards until every item of the construction plan matches the construction standards.

7. The method according to claim 1, characterized in that, The method further includes: Obtain the risk data associated with the construction project; The risk data is used to assess the potential safety risks currently existing in the construction plan; Obtain the corresponding response strategies for the potential security risks, and deploy them at the construction site corresponding to the construction project in accordance with the response strategies.

8. A device for generating a construction plan, characterized in that, The device includes: The acquisition module is used to acquire construction requirements uploaded by users, wherein the construction requirements include the construction content and construction standards of the construction project; The analysis module is used to analyze the construction content to obtain each key construction node and node description, and to generate key indicators corresponding to the key construction nodes based on the construction standards. A generation module is used to generate structured data using the key construction nodes and key indicators; The prediction module is used to input the structured data into the trained large language model, so that the large language model can make predictions based on the structured data to obtain a predicted construction plan, and dynamically correct the predicted construction plan and construction standards to obtain a construction scheme.

9. An electronic device, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

Images