A digital management method for engineering quality and safety integrating a large language model

Abstract

The application provides an engineering quality safety digital management method integrated with a large language model, comprising the following steps: S1, extracting engineering feature data from a basic database, and building an engineering quality safety knowledge graph according to the engineering feature data, wherein the engineering feature data comprises process feature data, material feature data and safety monitoring data; S2, obtaining engineering site data of a current engineering project; S3, performing engineering quality safety analysis on the engineering site data based on the large language model enhanced based on the knowledge graph according to the built engineering quality safety knowledge graph, and generating an engineering quality safety evaluation result. The application helps to improve the pertinence and application effect of the large language model in quality safety control of engineering projects, and improve the intelligent level of digital management of engineering projects.

Description

Technical Field

[0001] This invention relates to the field of construction project quality and safety management technology, and in particular to a digital management method for engineering quality and safety that incorporates a large language model. Background Technology

[0002] The quality and safety of construction projects are directly related to the safety of the nation and its people's lives and property, as well as social stability. However, traditional construction quality and safety management models have gradually revealed problems such as low efficiency in digital management, poor information flow, and insufficient decision support when faced with complex engineering environments, diverse stakeholders, and multimodal data information.

[0003] Artificial intelligence technologies, based on deep learning, image recognition, and natural language processing, are becoming increasingly widely used and recognized in construction project management as key technologies for digital management of construction project quality and safety. Meanwhile, large language model technology is playing a vital role in the digital management of enterprise engineering quality and safety.

[0004] However, in the current process of applying general large language models to solve the digital management of engineering quality and safety in enterprises, there are still problems such as insufficient intelligence and weak targeting, resulting in poor use effect. Summary of the Invention

[0005] To address the aforementioned problems, this invention aims to provide a digital management method for engineering quality and safety that incorporates a large language model.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] This invention proposes a digital management method for engineering quality and safety that incorporates a large language model, comprising the following steps:

[0008] S1 extracts engineering feature data from the basic database and builds an engineering quality and safety knowledge graph based on the engineering feature data. The engineering feature data includes process feature data, material feature data, and safety monitoring data.

[0009] S2 retrieves the current project's on-site data;

[0010] Based on the constructed engineering quality and safety knowledge graph, S3 uses the knowledge graph-enhanced large language model to perform engineering quality and safety analysis on on-site data and generate engineering quality and safety evaluation results.

[0011] Preferably, step S1 includes:

[0012] S11 retrieves historical project construction data from the basic database and extracts process feature data based on the construction data. The process feature data includes process name, construction period information, and process scheduling information.

[0013] S12 further extracts material characteristic data from the construction data, including material type, material performance record data, corresponding process and usage location, etc.

[0014] S13 extracts safety monitoring data based on the engineering records of historical projects. The safety monitoring data includes recorded quality and safety incidents, the corresponding work processes for the safety incidents, and the rules and regulations associated with the safety incidents.

[0015] S14 constructs an engineering quality and safety knowledge graph based on the acquired process characteristic data, material characteristic data, and safety monitoring data, and completes the setting of experience parameters in the engineering quality and safety knowledge graph.

[0016] Preferably, the constructed engineering quality and safety knowledge graph includes three types of nodes: process nodes, material nodes, and safety nodes; where each process node corresponds to a process, each material node corresponds to a material, and each safety node corresponds to a type of quality and safety event.

[0017] During the setup phase: Based on the process characteristic data of historical projects, obtain the duration information for each process node, and calculate the delay probability of each process node based on the duration information. ,in Let represent the probability of delay for process i. This represents the number of times process i is counted in a historical project. This represents the number of delays for process i in a historical project;

[0018] Based on the material characteristic data of historical projects, the performance degradation function corresponding to each material node is calculated. ,in This represents the property degradation function of material k at time t. This represents the initial property value of material k. This represents the attenuation rate parameter. Indicates ambient temperature. Indicates ambient humidity. Indicates the parameters of protective measures. This represents the environmental sensitivity parameter of material k;

[0019] Based on safety monitoring data from historical projects, the probability of quality and safety events at each safety milestone causing abnormalities in processes or materials is calculated. ,in Indicates a security incident The probability of triggering anomalies, Indicates security incidents in historical projects Total number of times recorded Indicates security incidents in historical projects The number of times that anomalies occur in the associated processes or materials while the record is being made.

[0020] Preferably, the current project site data includes construction plan information, construction logs, site monitoring data, and safety monitoring data; wherein, step S2 includes:

[0021] S21 obtains the corresponding planned process characteristic data based on the construction plan information of the current project. The planned process characteristic data includes process name, planned construction period information, planned process scheduling information, and planned process material information. Furthermore, it obtains the actual process characteristic data based on the construction log of the current project. The actual process characteristic data includes process name, actual construction period information, actual process scheduling information, and actual process material consumption information.

[0022] S22 acquires material performance parameters and on-site environmental data based on the on-site monitoring data of the current engineering project. The material performance parameters are obtained through the material test report, and the on-site environmental data are acquired through sensors installed at the construction site.

[0023] S23 obtains recorded quality and safety incident information based on the safety monitoring data of the current engineering project.

[0024] Preferably, step S3 includes:

[0025] S31 updates the engineering site data to the engineering quality and safety knowledge graph based on the acquired engineering site data;

[0026] S32 performs engineering quality and safety analysis based on the updated safety knowledge graph and the knowledge graph-enhanced large language model, and obtains the engineering quality and safety analysis results.

[0027] S33 transforms the obtained engineering quality and safety analysis results to obtain visualized engineering quality and safety evaluation results.

[0028] Preferably, in step S31, updating the on-site engineering data to the engineering quality and safety knowledge graph specifically includes:

[0029] During the initialization phase: Based on the obtained planned process characteristic data, the corresponding process nodes are updated, activating and connecting them to form a process chain. Each process node contains the planned start and end times for the process, information on the associated process nodes, and required material information. The association weights between process nodes and their corresponding material nodes are adjusted based on the required material information. ,in Indicates process node With material nodes The correlation weight between them ;

[0030] During the dynamic update phase: Process nodes are updated based on the obtained actual process characteristic data, updating the process status and actual project duration information to the corresponding process nodes; further, the dependency strength between related process nodes in the process chain is obtained. ,in This indicates the dependence strength of process node j on process node i; This represents the probability of delay for process node i, when process node i has already experienced a delay compared to its planned end time. ,otherwise , This represents the minimum waiting time for process node j after process node i has completed; This represents the time length between the planned / actual completion time of process node i and the planned start time of process node j;

[0031] The material nodes are activated and updated based on the obtained field monitoring data. Material performance parameters and field environmental data are updated to the material nodes, and the real-time material performance degradation value is further estimated. ,in This represents the degradation value of material k's properties after time t, where t represents the time difference between the current moment and the initial property value recorded for the material. This represents the initial performance value recorded for material k, obtained from the test report. Indicates ambient temperature. This indicates ambient humidity, obtained from data collected by environmental sensors. This represents the protective measure parameters, obtained based on the current protective measures for material k. This represents the environmental sensitivity parameter of material k;

[0032] The safety nodes are updated based on the obtained safety monitoring data. Corresponding safety nodes are activated based on recorded quality and safety events, and the event information is recorded in the safety nodes. This information includes the event type, time of occurrence, corresponding material or process information, and relevant regulatory clauses. The safety hazard intensity of each safety node is further calculated. ,in Indicates a security incident The probability of triggering anomalies, Indicates the process currently in progress. Security incident The probability of triggering This indicates the activation parameter; when a security event is logged, otherwise .

[0033] Preferably, in step S32, the engineering quality and safety analysis based on the knowledge graph-enhanced large language model includes:

[0034] S321 detects the dependency strength between current process nodes, compares the dependency strength with the set dependency standard value, and obtains the delay risk analysis results;

[0035] S322 detects the material performance degradation value of the current material node, compares the material performance degradation value with the set material performance standard value, and obtains the material performance analysis result.

[0036] S323 detects the security trigger strength of the current security node, compares the security trigger strength with the set security standard value, and obtains the security event analysis results.

[0037] Based on the current engineering quality and safety knowledge graph and the engineering quality and safety analysis results, the process is evolved, and steps S321-S323 are repeated based on the evolved engineering quality and safety knowledge graph to obtain the corresponding evolution analysis results.

[0038] Preferably, in step S32, the evolution is based on the current engineering quality and safety knowledge graph and the engineering quality and safety analysis results, including:

[0039] When a delay occurs at the current process node, an evolutionary analysis is performed on the material nodes associated with the current process node and the next process node whose dependence strength exceeds the standard. The evolution yields the maximum waiting time for the corresponding material node under the condition that the material performance strength requirements are met, thus obtaining the required completion time for the current process and the required completion time for the next process node as the evolutionary analysis result.

[0040] Based on the planned completion time of the current process node, the related material node is evolved to obtain the material performance degradation value of the evolved material node when the plan is completed. Furthermore, the evolved material performance degradation value is compared with the standard to obtain the evolved friction material performance analysis result.

[0041] When a safety node is activated and the safety trigger strength exceeds the preset standard, the current process will be delayed and the situation when the current process is delayed will be evolved.

[0042] Preferably, the abnormal engineering quality analysis results are input into the knowledge graph-enhanced big language model. The knowledge graph-enhanced big language model triggers further evolutionary analysis based on the abnormal engineering quality analysis results, and further obtains the evolved engineering quality analysis results based on the evolutionary data, thereby forming an analysis result logic chain. Based on the engineering quality analysis results associated with the logic chain, the model generates corresponding judgment criteria and processing suggestions from the knowledge base, and generates a visualized analysis report.

[0043] The beneficial effects of this invention are as follows: This invention proposes a digital management method for engineering quality and safety, which constructs a targeted engineering quality and safety knowledge graph based on historical engineering feature data. During the actual engineering quality management process, the engineering quality and safety knowledge graph is dynamically updated based on real-time on-site data obtained from the engineering project. Finally, the knowledge graph is used to enhance the large language model to achieve intelligent engineering quality and safety analysis and evolution, thereby outputting visualized engineering quality and safety evaluation results. This helps to improve the pertinence and application effect of the large language model in the quality and safety control of engineering projects, and improves the intelligent level of digital management of engineering projects. Attached Figure Description

[0044] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0045] Figure 1 This is a schematic diagram illustrating the steps of a digital management method for engineering quality and safety that incorporates a large language model, as shown in an embodiment of the present invention.

[0046] Figure 2 for Figure 1 A schematic diagram illustrating the specific steps of step S3;

[0047] Figure 3 for Figure 2 A schematic diagram illustrating the specific steps of step S32. Detailed Implementation

[0048] The present invention will be further described in conjunction with the following application scenarios.

[0049] See Figure 1 This invention discloses a digital management method for engineering quality and safety that incorporates a large language model, comprising the following steps:

[0050] S1 extracts engineering feature data from the basic database and builds an engineering quality and safety knowledge graph based on the engineering feature data. The engineering feature data includes process feature data, material feature data, and safety monitoring data.

[0051] S2 retrieves the current project's on-site data;

[0052] Based on the constructed engineering quality and safety knowledge graph, S3 uses the knowledge graph-enhanced large language model to perform engineering quality and safety analysis on on-site data and generate engineering quality and safety evaluation results.

[0053] The above embodiments of the present invention propose a digital management method for engineering quality and safety. This method constructs a targeted engineering quality and safety knowledge graph based on historical engineering feature data. During actual engineering quality management, the knowledge graph is dynamically updated using real-time on-site data acquired from the engineering project. Finally, the knowledge graph is used to enhance a large language model, enabling intelligent analysis and evolution of engineering quality and safety. This results in visualized engineering quality and safety evaluation results, improving the targeting and application effectiveness of the large language model in quality and safety control of engineering projects, and enhancing the intelligence level of digital management of engineering projects.

[0054] The digital management method for engineering quality and safety proposed in the above embodiments can be implemented based on locally deployed smart terminals or cloud servers, thereby providing targeted and visualized engineering quality and safety assessment results according to the needs of actual scenarios.

[0055] Preferably, step S1 includes:

[0056] S11 retrieves historical project construction data from the basic database and extracts process feature data based on the construction data. The process feature data includes process name, construction period information, and process scheduling information.

[0057] S12 further extracts material characteristic data from the construction data, including material type, material performance record data, corresponding process and usage location, etc.

[0058] S13 extracts safety monitoring data based on the engineering records of historical projects. The safety monitoring data includes recorded quality and safety incidents, the corresponding work processes for the safety incidents, and the rules and regulations associated with the safety incidents.

[0059] S14 constructs an engineering quality and safety knowledge graph based on the acquired process characteristic data, material characteristic data, and safety monitoring data, and completes the setting of experience parameters in the engineering quality and safety knowledge graph.

[0060] The basic database records construction data from historical projects, including construction plans, real log data, and monitoring data. By extracting necessary construction data, material characteristic data, and safety monitoring data from these historical records, the coverage of the engineering quality and safety knowledge graph can be broadened, increasing its data richness. The data in the basic database can be represented as BIM model data, site management logs, or data collected from on-site monitoring equipment. Building a comprehensive database by storing relevant historical project data in the basic database improves data management and facilitates subsequent data retrieval and analysis. The basic database further includes a dedicated enterprise knowledge base for construction project quality and safety management. This base integrates reference knowledge related to construction project quality and safety management, such as laws, regulations, procedures, standards, and industry standards; construction process flow and standards; quality and safety construction plans and operation standards; emergency plans; identification and control measures for potential quality defects and safety risks; enterprise engineering quality and safety management processes and systems; job responsibilities and work standards; project management standards; quality and safety standardization management manuals; standardized management forms; quality defect and control lists; FMEA analysis database; hazard source and major project control lists; quality and safety management evaluation indicators and assessments; application cases of BIM technology in digital management of engineering quality and safety; and application cases of PDCA / AHP / FMEA + intelligent algorithm integrated models in digital management of engineering quality and safety. This facilitates the subsequent use of large language models.

[0061] Among them, the engineering quality and safety knowledge graph proposed in this invention specifically extracts the required construction data, material characteristic data, and safety monitoring data from the basic database. Based on this, a three-level knowledge graph based on process, material, and safety event is built, which helps to improve the effectiveness of further safety analysis and evolution based on the knowledge graph and improve the pertinence of engineering quality and safety management.

[0062] Preferably, the constructed engineering quality and safety knowledge graph includes three types of nodes: process nodes, material nodes, and safety nodes; where each process node corresponds to a process, each material node corresponds to a material, and each safety node corresponds to a type of quality and safety event.

[0063] During the setup phase: Based on the process characteristic data of historical projects, obtain the duration information for each process node, and calculate the delay probability of each process node based on the duration information. ,in Let represent the probability of delay for process i. This represents the number of times process i is counted in a historical project. This represents the number of delays for process i in a historical project;

[0064] Based on the material characteristic data of historical projects, the performance degradation function corresponding to each material node is calculated. ,in This represents the property degradation function of material k at time t. This represents the initial property value of material k. This represents the attenuation rate parameter. Indicates ambient temperature. Indicates ambient humidity. Indicates the parameters of protective measures. This represents the environmental sensitivity parameter of material k;

[0065] Based on safety monitoring data from historical projects, the probability of quality and safety events at each safety milestone causing abnormalities in processes or materials is calculated. ,in Indicates a security incident The probability of triggering anomalies, Indicates security incidents in historical projects Total number of times recorded Indicates security incidents in historical projects The number of times that anomalies occur in the associated processes or materials while the record is being made.

[0066] Based on relevant data from historical projects, the parameters of the corresponding nodes are set to form an initial engineering quality and safety knowledge graph.

[0067] At the same time, based on the recorded historical safety incident data and process anomaly records, the probability of a process being triggered by a safety incident is calculated. ,in This indicates that process i was affected by a safety event. The probability of triggering This indicates the total number of times an anomaly occurred in process i. Indicate process A security event occurred simultaneously with the log. The total number of times will As the association weight between process node i and safety node v.

[0068] Based on historical security incident records and material anomaly records, the probability of materials being triggered by security incidents is statistically analyzed. ,in This indicates that material k was involved in a safety event. The probability of triggering an anomaly. This indicates the total number of times an anomaly occurred in material k. This indicates that a safety event occurred simultaneously with an anomaly in material k record. The total number of times will As material node k and safety node The correlation weight between them.

[0069] By leveraging data recorded in historical projects, the node architecture of the engineering quality and safety knowledge graph can be constructed. The nodes in this knowledge graph are categorized into three types: process nodes, material nodes, and safety nodes. Process nodes correspond to a specific construction process (e.g., concrete pouring, formwork installation, foundation treatment, waterproofing, site leveling, etc., with the process information also including the objects being constructed). The unidirectional paths between process nodes represent the entire process chain. Each process node records relevant knowledge and experience data. Knowledge data includes matching information from knowledge bases such as construction specifications and case studies (e.g., different specifications for the same process can be represented as different process nodes). Experience data includes the proposed delay probability. The process delay probability, obtained through statistical analysis of experience data, reflects the estimated delay situation of the process based on statistical experience, providing a reference for further analysis of process anomaly chains and overall engineering quality and safety management. Material nodes correspond to the materials required in the project (such as concrete, steel, timber, water pipes, paint, cement, etc.). Each material node records knowledge and experience data related to the corresponding material. The knowledge data includes material standards and instructions for use. The experience data includes the proposed performance decay function. Considering that the performance of materials will change with the storage time and environmental factors, the performance decay function corresponding to the material is obtained by combining the recorded changes in material performance data from historical projects with the environmental data at that time through fitting, simulation and other methods. This function reflects the performance change of the material over time and can serve as a basis for further reference in determining whether material performance will lead to abnormalities when combined with process delays or the evolution of engineering quality. A safety node corresponds to a safety event (such as impact, corrosion, high temperature, rainfall, snowfall, etc., which may affect the quality of the project; extracted from historical data of recorded safety events, events of different degrees can be divided into different safety nodes for further refinement). The safety node records the cases and standards of the safety event, as well as the probability that the proposed safety event will cause abnormalities in other processes or materials. The probability of triggering the event obtained through historical data statistics can provide a basis for further assessment of whether a chain reaction of processes, materials, or global impact occurs when a safety event occurs.

[0070] Based on the engineering quality and safety knowledge graph proposed in this invention, the empirical parameters of relevant nodes in the knowledge graph are improved by extracting and statistically analyzing historical data. This helps to truly reflect the mutual influence level of process, materials, and safety events during the construction of engineering projects. As a result, it can adapt to subsequent quality and safety analysis and further evolution based on abnormal situations, thus providing a basis for possible chain reactions and improving the pertinence and effectiveness of engineering quality and safety assessment.

[0071] The acquisition of specific data can be achieved through methods such as text parsing, key feature data extraction, and structured processing based on the recorded data; this invention does not impose specific limitations on these methods. For example, for structured data, corresponding historical data can be obtained through processing such as completion, interpolation, and feature extraction. For unstructured data, such as log text, corresponding feature data can be obtained through text parsing and text extraction.

[0072] Based on material data from historical projects, the performance degradation function is corrected using mathematical statistics. This is achieved by using recorded material data from historical projects, such as material performance values ​​at different times, and corresponding environmental temperature, humidity, and protection level data. By employing classification or fitting methods, the values ​​of key parameters such as degradation rate, protection measures, and environmental sensitivity parameters in the function can be empirically extracted, thereby obtaining parameter settings suitable for different situations.

[0073] Preferably, the current project site data includes construction plan information, construction logs, site monitoring data, and safety monitoring data; wherein, step S2 includes:

[0074] S21 obtains the corresponding planned process characteristic data based on the construction plan information of the current project. The planned process characteristic data includes process name, planned construction period information, planned process scheduling information, and planned process material information. Furthermore, it obtains the actual process characteristic data based on the construction log of the current project. The actual process characteristic data includes process name, actual construction period information, actual process scheduling information, and actual process material consumption information.

[0075] S22 acquires material performance parameters and on-site environmental data based on the on-site monitoring data of the current engineering project. The material performance parameters are obtained through the material test report, and the on-site environmental data are acquired through sensors installed at the construction site.

[0076] S23 obtains recorded quality and safety incident information based on the safety monitoring data of the current engineering project.

[0077] Based on the established engineering quality and safety knowledge graph, further real-time engineering site data of the target engineering project is collected, and the real real-time engineering site data is further integrated into the engineering quality and safety knowledge graph to obtain complete engineering quality and safety knowledge graph data for the current situation.

[0078] Preferably, in step S2, the acquired on-site engineering data is aligned to the same timestamp standard. For construction plan information, the overall project schedule and the construction schedule of specific procedures can be extracted. Through real-time construction log recording, the actual completion status and corresponding completion time of procedures can be updated in real time. At the same time, the material data required for each procedure and the environmental data of the construction site are also monitored in real time, and these are used as on-site monitoring data for real-time updates. When a safety incident is detected at the construction site, it is used as safety monitoring data for real-time updates.

[0079] The material performance parameters can be obtained by measuring them using specialized equipment or methods, and this invention does not specifically limit the scope of these parameters.

[0080] Preferred, see Figure 2 Step S3 includes:

[0081] S31 updates the engineering site data to the engineering quality and safety knowledge graph based on the acquired engineering site data;

[0082] S32 performs engineering quality and safety analysis based on the updated safety knowledge graph and the knowledge graph-enhanced large language model, and obtains the engineering quality and safety analysis results.

[0083] S33 transforms the obtained engineering quality and safety analysis results to obtain visualized engineering quality and safety evaluation results.

[0084] When conducting further quality and safety analysis based on on-site engineering data, the preliminary engineering data is first updated according to the classification of processes, materials, and safety events. This updates the information of the corresponding process nodes, material nodes, and safety nodes in the engineering quality and safety knowledge graph, resulting in a real-time, dynamic engineering quality and safety knowledge graph. Based on this real-time knowledge graph, engineering quality and safety analysis (e.g., real-time analysis, evolutionary analysis, local analysis, and overall analysis) is performed using an enhanced language model. This yields corresponding engineering quality and safety analysis results, which are then converted into text output to obtain visualized engineering quality and safety evaluation results. By completing real-time, targeted engineering quality and safety analysis and evaluation of the current engineering project using the above methods, the level and effectiveness of digital management of engineering quality and safety can be improved.

[0085] Preferably, in step S31, updating the on-site engineering data to the engineering quality and safety knowledge graph specifically includes:

[0086] During the initialization phase: Based on the obtained planned process characteristic data, the corresponding process nodes are updated, activating and connecting them to form a process chain. Each process node contains the planned start and end times for the process, information on the associated process nodes, and required material information. The association weights between process nodes and their corresponding material nodes are adjusted based on the required material information. ,in Indicates process node With material nodes The correlation weight between them ;

[0087] During the dynamic update phase: Process nodes are updated based on the obtained actual process characteristic data, updating the process status and actual project duration information to the corresponding process nodes; further, the dependency strength between related process nodes in the process chain is obtained. ,in This indicates the dependence strength of process node j on process node i; This represents the probability of delay for process node i, when process node i has already experienced a delay compared to its planned end time. ,otherwise , This represents the minimum waiting time for process node j after process node i has completed; This represents the time length between the planned / actual completion time of process node i and the planned start time of process node j;

[0088] The material nodes are activated and updated based on the obtained field monitoring data. Material performance parameters and field environmental data are updated to the material nodes, and the real-time material performance degradation value is further estimated. ,in This represents the degradation value of material k's properties after time t, where t represents the time difference between the current moment and the initial property value recorded for the material. This represents the initial performance value recorded for material k, obtained from the test report. Indicates ambient temperature. This indicates ambient humidity, obtained from data collected by environmental sensors. This represents the protective measure parameters, obtained based on the current protective measures for material k. This represents the environmental sensitivity parameter of material k;

[0089] The safety nodes are updated based on the obtained safety monitoring data. Corresponding safety nodes are activated based on recorded quality and safety events, and the event information is recorded in the safety nodes. This information includes the event type, time of occurrence, corresponding material or process information, and relevant regulatory clauses. The safety hazard intensity of each safety node is further calculated. ,in Indicates a security incident The probability of triggering anomalies, Indicates the process currently in progress. Security incident The probability of triggering This indicates the activation parameter; when a security event is logged, otherwise .

[0090] In this context, when the association weight between two nodes is greater than 0, it indicates that there is an association between the two nodes.

[0091] The updating of the engineering quality and safety knowledge graph involves two stages. The first stage, after completing the project's construction plan, activates the corresponding process nodes in the knowledge graph based on the plan information. Simultaneously, based on specific construction information, it further activates the corresponding material nodes, thus obtaining an initial process chain. The second stage, after actual construction begins, updates the corresponding nodes based on real-time on-site data, ensuring each node contains the latest data and information. Real-time updates of process nodes involve updating the completion status of the current process node to determine if there are delays or their specific durations. The dependency strength of the process nodes is updated based on the specific time information. Higher dependency strength indicates a stronger interdependence between related process nodes; therefore, a delay in an earlier process node could potentially lead to a chain reaction of abnormal situations. For material nodes, the current performance parameters are updated based on the latest monitored material performance parameters (such as concrete strength, concrete stress, steel corrosion resistance, and cement strength), as well as environmental information. This allows material nodes to reflect the current status of their corresponding materials, providing a reliable basis for further analysis and evolution. For safety nodes, when a corresponding safety event is detected based on on-site engineering data, the relevant safety node is activated, and the event information is recorded. This approach enables the engineering quality and safety knowledge graph to be updated in real-time and dynamically, improving the relevance and reliability of subsequent engineering quality and safety analysis and assessment.

[0092] Preferred, see Figure 3 In step S32, engineering quality and safety analysis is performed based on a knowledge graph-enhanced large language model, including:

[0093] S321 detects the dependency strength between current process nodes, compares the dependency strength with the set dependency standard value, and obtains the delay risk analysis results;

[0094] S322 detects the material performance degradation value of the current material node, compares the material performance degradation value with the set material performance standard value, and obtains the material performance analysis result.

[0095] S323 detects the security trigger strength of the current security node, compares the security trigger strength with the set security standard value, and obtains the security event analysis results.

[0096] In one scenario, the knowledge graph-enhanced large language model can call the built-in engineering specification database and match the corresponding standards based on the node information to be analyzed, so as to complete the corresponding comparison operation and obtain the corresponding analysis results.

[0097] In one scenario, the process dependency strength of the current process node is obtained through process node analysis. This indicates that the current two processes are too dependent on each other, so it is necessary to focus on monitoring other nodes related to the two processes.

[0098] In one scenario, the current concrete strength is obtained through material node information. If the strength is less than the set threshold of 0.7, the analysis result indicates that the current concrete strength is insufficient.

[0099] In one scenario, when the security trigger strength is obtained based on security node information... If the threshold of 0.1 is exceeded, the material nodes associated with the current process node or safety node will be further monitored and activated for further evolution analysis.

[0100] Based on the deductive capabilities of the large language model, further deductions are made based on the current engineering quality and safety knowledge graph, thereby obtaining further evolutionary analysis results.

[0101] In one scenario, the deduction method includes considering the possibility of a delay in the currently ongoing process, the maximum waiting time of materials associated with the process, and whether the maximum waiting time of materials meets the completion time requirements of the corresponding process in the event of a delay.

[0102] Preferably, step S32 further includes:

[0103] S324 evolves based on the current engineering quality and safety knowledge graph and engineering quality and safety analysis results, and repeats steps S321-S323 based on the evolved engineering quality and safety knowledge graph to obtain the corresponding evolution analysis results.

[0104] Preferably, in step S32, the evolution is based on the current engineering quality and safety knowledge graph and the engineering quality and safety analysis results, including:

[0105] When a delay occurs at the current process node, an evolutionary analysis is performed on the material nodes associated with the current process node and the next process node whose dependence strength exceeds the standard. The evolution yields the maximum waiting time for the corresponding material node under the condition that the material performance strength requirements are met, thus obtaining the required completion time for the current process and the required completion time for the next process node as the evolutionary analysis result.

[0106] Based on the planned completion time of the current process node, the related material node is evolved to obtain the material performance degradation value of the evolved material node when the plan is completed. Furthermore, the evolved material performance degradation value is compared with the standard to obtain the evolved friction material performance analysis result.

[0107] When a safety node is activated and the safety trigger strength exceeds the preset standard, the current process will be delayed and the situation when the current process is delayed will be evolved.

[0108] The aforementioned evolutionary analysis methods can also be used to conduct further evolutionary analysis based on the results obtained from the evolutionary analysis, resulting in a comprehensive analysis.

[0109] In one scenario, depending on the actual situation, evolutionary analysis may also include:

[0110] The material nodes are evolved to obtain the maximum waiting time for the corresponding material nodes under the condition that the material performance and strength requirements are met, thereby determining whether the maximum waiting time meets the planned construction time of the related process.

[0111] By using a large language model to perform evolutionary analysis on the current engineering quality and safety knowledge graph, logical chains can be formed to accurately simulate the evolution of potential cascading situations, thereby obtaining predicted evolutionary analysis results. Compared with traditional engineering quality and safety analysis methods, the logical chain-based analysis method can improve the adaptability and accuracy of evolutionary analysis, contributing to the improvement of the logical and intelligent level of engineering quality and safety analysis and assessment. Furthermore, the obtained evolutionary analysis results are specific to real engineering projects, exhibiting strong adaptability and relevance, further enhancing the effectiveness of engineering quality and safety analysis.

[0112] Meanwhile, the evolutionary approach based on logical chains can improve the generation of early warning and response plans for abnormal situations, and enhance the effectiveness of digital management of engineering quality and safety based on large language models.

[0113] Preferably, step S32 includes:

[0114] A comprehensive risk analysis is conducted based on the process nodes included in the process chain, as well as the material and safety nodes associated with those process nodes, to calculate the total risk value of the project. ,in This represents the estimated total risk value of the engineering project. Represented as consecutive process nodes in a process chain. This represents the dependence strength of process node j on process node i. Let k be the material node associated in the process chain. This represents the degradation value of material properties of material k at the current moment. This represents the initial performance value recorded for material k. Represented as a safety node associated in the process chain. , Indicates a safe node The intensity of the safety incident, , , These represent the set weighting factors;

[0115] The comprehensive risk analysis results are obtained by comparing the total risk value of the project with the established risk standards.

[0116] Furthermore, based on the evolutionary or real-time engineering quality and safety knowledge graph, it is possible to further conduct a comprehensive risk analysis of the entire engineering project, providing a basis for engineering quality and safety management.

[0117] In one scenario, the specific steps of performing engineering quality and safety analysis based on the updated safety knowledge graph in step S32 can be implemented based on the internal reasoning module of the large language model, or by using the knowledge graph to enhance the large language model, or by setting up and training a dedicated deep learning model. This invention does not specifically limit these steps.

[0118] Preferably, step S33 includes:

[0119] Based on the obtained analysis results, engineering quality and safety analysis results, and corresponding node information, input parameters are generated. These input parameters are then input into a knowledge graph-enhanced large language model to obtain visualized engineering quality and safety evaluation results. The engineering quality and safety analysis results include at least one of the following: delay risk analysis results, material performance analysis results, safety incident analysis results, and comprehensive risk analysis results.

[0120] Preferably, step S33 includes:

[0121] Anomalies in the engineering quality analysis results are input into the knowledge graph-enhanced language model. The knowledge graph-enhanced language model then triggers further evolutionary analysis based on these anomalies, and further obtains the evolved engineering quality analysis results based on the evolutionary data, thus forming a logical chain of analysis results. Based on the engineering quality analysis results associated with the logical chain, the model generates further judgment criteria and processing suggestions from the knowledge base, and generates a visualized analysis report.

[0122] Based on the large language model, a visualized engineering quality and safety assessment report is further generated. Based on the engineering quality and safety analysis results, the corresponding knowledge can be retrieved from the standard database to further interpret and display the analysis results. This helps managers to propose engineering quality and safety improvement plans based on the assessment report, and realize digital management of engineering quality and safety.

[0123] In an exemplary scenario, based on the current engineering quality and safety knowledge graph, when a safety event (such as a 24-hour rainfall of 60mm that has lasted for 15 minutes) is detected, if the safety initiation intensity obtained from this safety node exceeds the standard value, it is determined that the probability of the current process (such as concrete pouring) being delayed due to an abnormal impact is increased. That is, based on the current engineering quality and safety knowledge graph, the scenario of delay in the current process is evolved, and the associated material nodes are further evolved, with the maximum waiting time for the current material node (concrete) being 30 minutes. Simultaneously, the dependency strength of the current process node is further analyzed. If the value is less than the set standard of 0.6, it indicates that the next process node (curing) has a low dependence on the current process node, and the next process node is not significantly affected by the delay of the current process node. Based on the above evolution results, the evolution analysis results corresponding to the logical chain are: rainstorm occurs → concrete pouring is affected → concrete pouring is delayed → maximum waiting time for concrete is 30 minutes. The resulting handling suggestions are: 1. Take rain protection measures; 2. Improve the requirements for concrete quality monitoring; 3. Complete the pouring within 30 minutes; 4. Prepare spare concrete materials.

[0124] In another exemplary scenario, based on the current engineering quality and safety knowledge graph, when a safety event (such as insufficient concrete strength in area B) is detected, if the safety-inducing intensity obtained from this safety node exceeds the standard value, then the probability of the current process (such as concrete pouring in area B) being delayed due to an abnormality is increased. That is, based on the current engineering quality and safety knowledge graph, the scenario of a delay in the current process (a delay of 30 days) is evolved, and the dependence of the current process node on subsequent processes (such as concrete pouring in area D) is further analyzed. , ..., If all values ​​exceed the set standard of 0.6, it indicates that subsequent process nodes are affected by delays. For process node q, its associated material node (cement) is further obtained. Evolution of this material node reveals a maximum waiting time of 20 days. Based on the above evolution results, the evolutionary analysis of the logical chain is as follows: Insufficient concrete strength in area B → 30-day delay in concrete pouring in area B → Overall project delay → Maximum waiting time for cement in area D concrete pouring: 20 days… Therefore, the recommended actions are: 1. Immediately stop work (according to the "Code for Acceptance of Construction Quality of Concrete Structures": "Insufficient concrete strength requires immediate work stoppage and testing"), and complete the reinforcement of the concrete in area B (see attached accident and reinforcement case studies); 2. Adjust the subsequent process plan (see attached intelligent adjustment results); 3. Complete the concrete pouring in area D within 20 days; 4. Prepare spare cement materials.

[0125] From the above description of the embodiments, those skilled in the art will understand that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For hardware implementation, the processor can be implemented in one or more of the following units: application-specific integrated circuit (ASIC), digital signal processor (DSP), digital signal processing device (DSPD), programmable logic device (PLD), field-programmable gate array (FPGA), processor, controller, microcontroller, microprocessor, other electronic units designed to implement the functions described herein, or combinations thereof. For software implementation, some or all of the processes of the embodiments can be implemented by a computer program instructing the associated hardware. During implementation, the program described above can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media can be any available medium that is accessible to a computer. Computer-readable media can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible to a computer.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should be able to analyze that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A method for integrating a large language model into engineering quality safety digital management, characterized in that, Comprise the following steps: S1 extracts engineering feature data from the basic database, and builds an engineering quality and safety knowledge graph according to the engineering feature data, wherein the engineering feature data comprises process feature data, material feature data and safety monitoring data; S2 obtains engineering site data of a current engineering project; S3 performs engineering quality and safety analysis on the engineering site data based on the built engineering quality and safety knowledge graph and a knowledge graph enhanced large language model, and generates an engineering quality and safety evaluation result; Wherein, step S1 comprises: S11 obtains construction data of historical projects from the basic database, and extracts process feature data from the construction data, wherein the process feature data comprises process name, construction period information and process scheduling information; S12 further extracts material feature data from the construction data, wherein the material feature data comprises material type, material performance record data, corresponding process and use position; S13 extracts safety monitoring data from the engineering records of the historical projects, wherein the safety monitoring data comprises recorded quality and safety events, safety event corresponding processes and safety event associated rule clauses; S14 builds an engineering quality and safety knowledge graph according to the obtained process feature data, material feature data and safety monitoring data, and completes the setting of experience parameters in the engineering quality and safety knowledge graph.

2. The method of claim 1, wherein the method is characterized by, The built engineering quality and safety knowledge graph comprises three types of nodes: process nodes, material nodes and safety nodes; wherein each process node corresponds to a process, each material node corresponds to a material, and each safety node corresponds to a type of quality and safety event; In the construction stage: according to the process characteristic data of historical projects, the duration information of each process node is obtained, and the delay probability of each process node is calculated according to the duration time information wherein represents the delay probability of process i, represents the statistical number of process i in historical projects, represents the number of delays of process i in historical projects; According to the material characteristic data of the historical project, the performance attenuation function corresponding to each material node is counted wherein represents the performance attenuation function of the material k at time t, represents the initial performance value of the material k, represents the attenuation rate parameter, represents the ambient temperature, represents the ambient humidity, represents the protection measure parameter, represents the environmental sensitivity parameter of the material k; According to the safety monitoring data in the historical project, the probability of abnormality of the process or material caused by the quality safety event corresponding to each safety node is counted wherein represents the safety event the abnormality probability, represents the total number of times of the safety event recorded in the historical project, represents the number of times of abnormality of the process or material associated with the safety event recorded in the historical project.

3. The method of claim 2, wherein the method is characterized by, The engineering site data of the current engineering project comprises construction plan information, construction log, site monitoring data and safety monitoring data; wherein, step S2 comprises: S21 obtains corresponding planned process feature data according to the construction plan information of the current engineering project, wherein the planned process feature data comprises process name, planned construction period information, planned process scheduling information and planned process required material information; further obtains actual process feature data according to the construction log of the current engineering project, wherein the actual process feature data comprises process name, actual construction period information, actual process scheduling information and actual process consumed material information; S22 obtains material performance parameters and site environment data according to the site monitoring data of the current engineering project, wherein the material performance parameters are obtained through the detection report of the material, and the site environment data are obtained through the sensors set in the construction site; S23 obtains recorded quality and safety event information according to the safety monitoring data of the current engineering project.

4. The method of claim 3, wherein the method is characterized by, Step S3 comprises: S31 updates the engineering site data to the engineering quality and safety knowledge graph according to the obtained engineering site data; S32 performs engineering quality and safety analysis based on the knowledge graph enhanced large language model according to the updated safety knowledge graph, and obtains engineering quality and safety analysis result; S33 converts the obtained engineering quality and safety analysis result to obtain visual engineering quality and safety evaluation result.

5. The method of claim 4, wherein the method is characterized by, In step S31, the engineering site data is updated to the engineering quality and safety knowledge graph, specifically comprising: In the initialization stage: according to the obtained planning process feature data, update to the corresponding process node, activate and connect the corresponding process node to form a process chain, wherein each process node contains the planning start time and the planning end time corresponding to the process, the information of the front and rear associated process nodes and the required material information; according to the required material information, adjust the association weight between the process node and the corresponding material node , wherein represents the association weight between the process node and the material node , ; In the dynamic updating stage: according to the obtained actual process characteristic data, the process node is updated, and the process state and actual duration information are updated to the corresponding process node; further, the dependency strength between the associated process nodes in the process chain is acquired wherein represents the dependency strength of the process node j to the process node i; represents the delay probability of the process node i, when the process node i has appeared delay compared with the planned end time, , otherwise , represents the minimum waiting time length of the process node j after the completion of the process node i; represents the time length between the planned / actual completion time of the process node i and the planned start time of the process node j; According to the obtained field monitoring data, the material node is activated and updated, the material performance parameters and the field environment data are updated to the material node, and the real-time material performance attenuation value is further estimated wherein represents the material performance attenuation value of the material k after time t, wherein t represents the time difference between the current time and the initial performance value recorded by the material, represents the initial performance value recorded by the material k, which is obtained according to the detection report, represents the environmental temperature, represents the environmental humidity, which is obtained according to the data collected by the environmental sensor, represents the protection measure parameter, which is obtained according to the current protection measure for the material k, represents the environmental sensitivity parameter of the material k; According to the obtained safety monitoring data, the safety node is updated, the corresponding safety node is activated according to the recorded quality safety event, and the quality safety event information is recorded into the safety node, wherein the quality safety event information includes a safety event type, a time of occurrence, corresponding material or process information, and associated regulation clauses; and the safety initiation intensity of the safety node is further calculated wherein represents the initiation abnormal probability of a safety event , represents a process currently being performed by a safety event , represents an activation parameter, and when a safety event is recorded, otherwise .

6. The method of claim 5, wherein the method is characterized by, In step S32, the engineering quality and safety analysis based on the knowledge graph enhanced large language model includes: S321 detects the dependency strength between the current process nodes, compares the dependency strength with the set dependency standard value, and obtains the delay risk analysis result; S322 detects the material performance attenuation value of the current material node, compares the material performance attenuation value with the set material performance standard value, and obtains the material performance analysis result; S323 detects the safety triggering strength of the current safety node, compares the safety triggering strength with the set safety standard value, and obtains the safety event analysis result; S324 evolves based on the current engineering quality and safety knowledge graph and the engineering quality and safety analysis result, and repeats steps S321-S323 based on the engineering quality and safety knowledge graph obtained by evolution to obtain the corresponding evolution analysis result.

7. The method of claim 6, wherein the method is characterized by, In step S32, the evolution based on the current engineering quality and safety knowledge graph and the engineering quality and safety analysis result includes: When the current process node is delayed, the material nodes associated with the current process node and the next process node with a dependency strength exceeding the standard are analyzed for evolution, and the maximum waiting time of the corresponding material node meeting the material performance strength requirement is obtained by evolution, so as to obtain the required completion time of the current process and the required completion time of the next process node as the evolution analysis result; According to the planned completion time of the current process node, the associated material node is evolved to obtain the evolved material performance attenuation value of the material node at the planned completion time, and further compared with the standard to obtain the evolved material performance analysis result; When the safety node is activated and the safety triggering strength exceeds the preset standard, the current process is evolved for delay, and the case when the current process is delayed is evolved.

8. The method according to claim 7, characterized in that, The abnormal engineering quality analysis result is input into the knowledge graph enhanced large language model, the knowledge graph enhanced large language model triggers further evolution analysis according to the abnormal engineering quality analysis result, and further obtains the evolved engineering quality analysis result according to the evolved data, so as to form an analysis result logic chain; according to the engineering quality analysis result associated by the logic chain, the corresponding judgment basis and processing suggestion are retrieved from the knowledge base to generate a visual analysis report.

Images