Steel structure power machine room construction quality whole process management and control method based on digital twinning

Abstract

The invention discloses a steel structure power machine room construction quality whole process management and control method based on digital twinning, and belongs to the technical field of industrial intelligent manufacturing. The method comprises the following steps: obtaining construction object information, carrying out hierarchical division, and constructing and training a three-layer digital twinborn model; collecting and processing construction data; performing hierarchical quality analysis on the processed construction data through a three-layer digital twinborn model, and starting cross-layer correlation analysis to position a problem root when quality abnormality is detected; according to an analysis result of the layered quality analysis, triggering graded early warning, and generating a quantitative construction adjustment instruction based on analog simulation of a three-layer digital twin model; according to the method, a foundation-structure-construction three-layer digital twinborn model is constructed, and a cross-layer parameter association rule is established, so that high-fidelity virtual mapping of construction total elements is realized.

Description

Technical Field

[0001] This invention relates to the field of industrial intelligent manufacturing technology, specifically a method for full-process quality control of steel structure power room construction based on digital twins. Background Technology

[0002] As a core facility of the power system, the construction quality of steel structure power rooms directly affects the safety, stability, and service life of the overall structure. With the increasing scale and complexity of modern engineering construction, traditional construction quality management methods are no longer sufficient to meet the high standards of quality control. Currently, the main technical limitations are as follows: First, existing monitoring methods have significant shortcomings. Traditional quality control mainly relies on manual inspections and post-event testing, which not only results in a delayed response with the average time to discover quality problems exceeding 24 hours, but is also greatly affected by personnel experience, leading to a high rate of missed detections. Although some studies have attempted to use digital twin technology for monitoring, its monitoring dimensions are limited, focusing only on limited parameters such as structural displacement and stress, and failing to incorporate influencing factors such as geological conditions and construction operations into a unified analysis framework, making it impossible to accurately trace the root causes of quality problems.

[0003] Secondly, the disconnect between the model and the physical entity is a prominent problem. Most existing digital twin models adopt a static architecture with fixed parameter thresholds, which cannot adapt to dynamic factors such as fluctuations in material properties and changes in the environment. At the same time, the models lack an effective accuracy verification mechanism, making it difficult to ensure the consistency between the virtual model and the physical entity, resulting in a significant decrease in monitoring accuracy in the later stages.

[0004] Third, there is a lack of quantitative guidance for rectifying quality issues. Existing systems are usually limited to alarm alerts and fail to provide specific rectification plans. Rectification measures rely on manual experience, which is highly subjective, resulting in low compliance rates, high rework costs, and an inability to form effective closed-loop management.

[0005] Furthermore, the data traceability and knowledge reuse capabilities are insufficient. Traditional methods lack a full-chain quality data management mechanism, making data prone to loss and difficult to trace. Moreover, they cannot continuously optimize model parameters and analysis algorithms through historical data, leading to the recurrence of similar problems in different projects.

[0006] Therefore, there is an urgent need for an intelligent quality management method that can achieve full-process, multi-dimensional, and adaptive management, fundamentally solving problems such as one-sided monitoring, rigid models, ambiguous decision-making, and difficulty in traceability in existing technologies. Summary of the Invention

[0007] The purpose of this invention is to provide a method for full-process quality control of steel structure power room construction based on digital twins, so as to solve the problems mentioned in the background art.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A method for full-process quality control of steel structure power room construction based on digital twins, the method comprising the following steps: Obtain information about the construction objects and divide them into levels, then construct and train a three-layer digital twin model; Collect and process construction data; A three-layer digital twin model is used to perform hierarchical quality analysis on the processed construction data. When a quality anomaly is detected, cross-layer correlation analysis is initiated to locate the root cause of the problem. Based on the analysis results of the layered quality analysis, a graded early warning is triggered, and quantitative construction adjustment instructions are generated based on the simulation of the three-layer digital twin model. The construction quality traceability archive is constructed and uploaded to the blockchain, and the parameter thresholds and analysis algorithms of the three-layer digital twin model are dynamically iteratively optimized based on historical construction data.

[0009] Furthermore, the construction of the three-layer digital twin model specifically includes: Historical qualified data and experimental data from the construction archives of similar steel structure power rooms are extracted, along with experimental data from a welding environment built in a laboratory consistent with the site. The historical qualified data includes valid data corresponding to welding current, voltage, time, weld flaw detection pass rate, and mechanical property test results. The experimental data includes quality results of weld appearance quality, internal defects, and tensile strength. Information on the construction object is obtained, and an initial training set of welding parameter-quality result data pairs is constructed. The initial training set is then deduplicated and outlier is removed. Based on data type, the construction object information is divided into a foundation layer, a structural layer, and a construction layer; the foundation layer includes geological survey data, the structural layer includes design drawing data, and the construction layer includes construction plan data. Based on BIM technology, a three-layer digital twin model corresponding to the foundation layer, structural layer and construction layer is constructed, and parameter association rules are preset in the model; the parameter association rules include the linkage rules between component displacement data and foundation settlement data, as well as the mapping rules between welding temperature data and material mechanical property data; The three-dimensional coordinate data of the entity structure obtained by laser scanning is compared with the coordinates of the basic layer of the model. When the error exceeds the preset threshold, the model parameters are adjusted. The simulation results of the mechanical performance of the key components are verified with the mechanical test data of the entity components. When the consistency error between the three-layer digital twin model and the entity is less than the error threshold, the trained three-layer digital twin model is obtained.

[0010] Furthermore, the mapping rule between the welding temperature data and the material mechanical property data, the mapping formula is as follows: in, σ represents the target mechanical properties of the weld metal, including tensile strength, yield strength, or impact toughness. I is the welding current; U is the welding voltage; V is the welding speed; t is a time coefficient related to plate thickness or thermal conduction; C, Mn, Si, ... are the specific chemical element contents of the base material and welding material.

[0011] Furthermore, the collection and processing of construction data specifically includes: Multiple data acquisition devices are deployed at the construction site to collect construction data; the data acquisition devices include displacement sensors installed at the connection points of load-bearing components, infrared temperature sensors installed in the welding operation area, and drones equipped with high-definition cameras; Outlier removal and data standardization are performed on the construction data. The outlier removal adopts the 3σ criterion, and construction data that exceeds the preset normal value range is automatically marked and triggers secondary data collection. The data standardization process includes converting construction data of different dimensions into standardized values ​​in the range of 0-1. A weighted fusion algorithm is used to fuse construction data of the same type from different acquisition devices to generate fused construction data; the same type of construction data includes at least component displacement data and welding quality data. The integrated construction data will be synchronized in real time to the corresponding level of the three-layer digital twin model.

[0012] Furthermore, the component displacement data is calculated using the following weighted fusion formula: in, The displacement values ​​of the fused components. The displacement values ​​of the component collected by the displacement sensor. The component displacement value is extracted by UAV image recognition, α is the displacement weight corresponding to the displacement sensor, and β is the displacement weight corresponding to UAV image recognition. For welding quality data, the following weighted fusion formula is used to generate a comprehensive welding quality evaluation index: in, As a comprehensive evaluation index for welding quality, These are standardized values ​​for welding temperature data. Standardized scoring of weld appearance image feature data, ω1 represents the pass rate of ultrasonic flaw detection data, ω2 represents the weight of welding temperature data, ω3 represents the weight of weld appearance image feature data, and ω3 represents the weight of ultrasonic flaw detection data.

[0013] Furthermore, the hierarchical quality analysis and cross-hierarchical correlation analysis specifically include: The basic layer analysis involves inputting real-time collected foundation settlement data into a pre-trained time series prediction model to obtain predicted foundation settlement values ​​for future periods. The predicted foundation settlement values ​​are then compared with the preset allowable settlement threshold in the basic layer of the model. If the predicted foundation settlement values ​​exceed the predicted threshold, they are marked as potential quality hazards in the basic layer. Structural layer analysis compares the real-time collected component installation coordinate data with the preset three-dimensional coordinate thresholds in the model's structural layer to calculate the deviation value; a multi-parameter coupling analysis algorithm is used to comprehensively analyze multiple parameters of the welding process to determine the welding quality level. Construction layer analysis involves analyzing real-time collected construction operation video streams using behavior recognition algorithms to identify whether the operational behaviors are compliant. Cross-level correlation analysis automatically retrieves and correlates historical and real-time data from other levels when a quality anomaly is detected at any level, comprehensively locating the root cause of the quality problem. The cross-layer correlation analysis specifically refers to: When a component displacement exceeds the tolerance in the structural layer, the system automatically retrieves the foundation settlement data of the area in the foundation layer at the previous time point, as well as the installation process record data of the component in the construction layer. If the settlement amount in the foundation settlement data exceeds the limit and the trend of change is synchronized with the displacement deviation, then the root cause of the problem is determined to be foundation settlement. If the foundation settlement data is normal, but the installation procedure record shows that the positioning clamps were not used in accordance with the specifications, then the root cause of the problem is determined to be an installation operation error.

[0014] Furthermore, the multi-parameter coupled analysis algorithm calculates the comprehensive welding quality score using the following formula: in, To determine the overall score for welding quality, This is a standardized value for the welding current. This is a standardized value for the welding voltage. This is a standardized evaluation value for the characteristics of the welding time versus temperature curve. This is a standardized value for weld height. The standardized value weight of the welding current. The standardized value weights for welding voltage, The standardized evaluation values ​​for the characteristics of the time-temperature curve are weighted. 7 represents the standardized weight of the weld height.

[0015] Furthermore, the triggering of tiered early warnings and the generation of quantified construction adjustment instructions specifically include: Based on the severity of the quality problems obtained from the stratified quality analysis, different levels of early warnings are triggered; the early warning levels include at least a Level 1 early warning requiring immediate work stoppage and rectification, a Level 2 important early warning requiring timely handling, and a Level 3 general early warning for record-keeping only; Based on the simulation analysis results of the three-layer digital twin model, executable adjustment instructions containing specific operation parameters are generated; the adjustment instructions include component correction instructions and welding parameter adjustment instructions. The component correction instructions specifically include: In the three-layer digital twin model, with the goal of correcting displacement deviation to be no greater than 2mm, the required corrective force is calculated using a mechanical simulation module; the corrective force is calculated using the following mechanical formula: F = (ΔL × E × S) / L in, F is the required corrective force, ΔL is the displacement to be corrected, E is the elastic modulus of the component material, S is the cross-sectional area of ​​the component, and L is the length of the component. Based on the required corrective force F, the specific steps for applying the corrective force in stages, the required tools and their installation positions are planned according to the preset planning standards, thus forming an executable corrective instruction; The welding parameter adjustment instructions specifically include: Retrieve the welding temperature-current-voltage correlation model and historical qualified data stored in the three-layer digital twin model; The required current and voltage values ​​are calculated by reverse engineering using the aforementioned correlation model; the correlation model is as follows: T = a × I + b × U + c in, T is the welding temperature, I is the welding current, U is the welding voltage, and a, b, c are model coefficients obtained by training with historical data; The calculated current and voltage values, along with their corresponding simulated welding strength results, are used to generate an adjustment instruction according to the adjustment instruction template.

[0016] Furthermore, the construction quality traceability archive is uploaded to the blockchain, and the parameter thresholds and analysis algorithms of the three-layer digital twin model are dynamically iteratively optimized based on historical construction data, specifically including: Construct a construction quality traceability file containing problem details, rectification instructions, and rectification results, and upload the construction quality traceability file to the blockchain; Each record in the construction quality traceability archive generates a unique hash value, which is stored in association with the hash value of the previous record, forming a chain structure; the construction quality traceability archive supports retrieval by issue ID or component number; Based on historical construction data, machine learning algorithms are used to optimize and update the parameter thresholds and analysis algorithms in the three-layer digital twin model; The optimization process for the parameter threshold includes: A random forest model is used, with welding parameters from historical construction processes as input features and the probability of quality compliance as output. The optimal temperature threshold is found through training the random forest model, and the lower limit threshold of welding temperature in the random forest model is updated to the optimal temperature threshold.

[0017] Compared with the prior art, the beneficial effects achieved by the present invention are: This invention achieves high-fidelity virtual mapping of all construction elements by constructing a three-layer digital twin model (foundation-structure-construction) and establishing cross-layer parameter association rules; it achieves high-frequency synchronization and interaction between physical entities and virtual models through multi-source heterogeneous data fusion processing and a virtual-real synchronization mechanism; it achieves precise root-cause localization of quality problems through layered intelligent analysis, including foundation settlement trend analysis based on time series prediction, multi-parameter coupled welding quality assessment, and cross-layer correlation diagnosis, such as the collaborative analysis of structural displacement anomalies and foundation settlement, and construction operations; and it generates quantitative decision support based on mechanical simulation and process parameter optimization. The system provides executable and precise adjustment instructions; through a blockchain-enabled, tamper-proof traceability system and a data-driven model self-optimization mechanism, it achieves reliable traceability in quality management and continuous evolution of system performance. This solution effectively addresses the technical pain points of traditional methods, such as single monitoring dimensions, one-sided problem diagnosis, reliance on experience for decision-making, and static and rigid models. It realizes a significant transformation in construction quality management, from passive post-construction inspection to proactive full-process control, from isolated local analysis to global collaborative diagnosis, from qualitative experience-based judgment to quantitative scientific decision-making, and from fixed threshold monitoring to adaptive optimization. This significantly improves the accuracy, reliability, and intelligence of quality control. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0019] Figure 1 This is a schematic diagram illustrating the steps of the present invention's method for full-process quality control of steel structure power room construction based on digital twins. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 In this embodiment: a method for full-process quality control of steel structure power room construction based on digital twin is provided, wherein the steel structure power room is the construction object, and the construction object information includes foundation layer information, structural layer information, and construction layer information; the method includes the following steps: S1. Obtain information about the construction objects and divide them into levels, then construct and train a three-layer digital twin model; S2. Collect and process construction data; S3. Perform layered quality analysis on the processed construction data through a three-layer digital twin model. When a quality anomaly is detected, initiate cross-layer correlation analysis to locate the root cause of the problem. S4. Based on the analysis results of the layered quality analysis, trigger a graded early warning, and generate quantitative construction adjustment instructions based on the simulation of the three-layer digital twin model. S5. Construct a construction quality traceability archive and upload it to the blockchain. Based on historical construction data, dynamically iterate and optimize the parameter thresholds and analysis algorithms of the digital twin model.

[0022] Through the above steps, this invention provides a systematic solution to the three core problems existing in traditional steel structure construction quality management: First, it solves the problem of single monitoring dimensions by realizing the digital mapping of all elements from geological environment to component status to construction operation through a three-layer model architecture of foundation-structure-construction; Second, it breaks through the technical bottleneck of one-sided problem diagnosis by tracing the root cause of quality anomalies from the system level through cross-layer correlation analysis mechanism; Third, it overcomes the limitation of rectification measures relying on experience by generating quantitative and executable precise adjustment plans based on the simulation and deduction capabilities of digital twin models.

[0023] The significant effects achieved in this embodiment are reflected in three aspects: at the technical level, a high-frequency interaction channel between physical entities and virtual models has been established, realizing real-time synchronization and bidirectional optimization of virtual and physical spaces; at the management level, a closed-loop control system from quality monitoring, analysis and early warning to decision execution has been constructed, transforming traditional passive quality inspection into proactive preventive quality control; at the system level, blockchain technology ensures the immutability and traceability of quality data, while relying on machine learning algorithms to enable the system to have continuous self-optimization capabilities, significantly improving the accuracy and reliability of quality management.

[0024] In this preferred embodiment, the construction of the three-layer digital twin model in step S1 specifically includes: S1.1 Obtaining the initial training set: Extract historical qualified data from the construction archives of similar steel structure power rooms and experimental data from a welding environment built in the laboratory that is consistent with the site. The historical qualified data includes valid data corresponding to welding current, voltage, time, weld flaw detection pass rate, and mechanical property test results. The experimental data includes quality results of weld appearance quality, internal defects, and tensile strength. Obtain construction object information to form an initial training set of welding parameter-quality result data pairs. Perform deduplication and outlier removal processing on the initial training set. S1.2 Data Layered Acquisition: Based on data type, the construction object information is divided into a foundation layer, a structural layer, and a construction layer; the foundation layer includes geological survey data, the structural layer includes design drawing data, and the construction layer includes construction plan data; the structural layer design drawing data includes allowable values ​​for component size deviations, material parameters, and connection method requirements; the construction layer construction plan data includes construction time windows for each process and personnel operation standards.

[0025] S1.3, Layered Model Construction: Based on BIM technology, a three-layer digital twin model corresponding to the foundation layer, structural layer, and construction layer is constructed, and parameter association rules are preset in the model; the parameter association rules include the linkage rules between component displacement data and foundation settlement data, and the mapping rules between welding temperature data and material mechanical property data; the foundation layer is bound to geological survey data, the structural layer is associated with the three-dimensional coordinates of components and material performance parameters, and the construction layer is embedded with process flow and quality standard thresholds.

[0026] S1.4 Model accuracy verification: The three-dimensional coordinate data of the solid structure obtained by laser scanning is compared with the coordinates of the model's base layer. When the error exceeds the preset threshold of 0.5mm, the model parameters are adjusted. The simulation results of the mechanical performance of the key components are verified with the mechanical test data of the solid components to ensure that the consistency error between the model and the solid is ≤1%.

[0027] The mapping rule between the welding temperature data and the material mechanical property data preset in step S1.3 is specifically implemented through a multivariate nonlinear regression model. This model is used to characterize the quantitative relationship between welding heat input, cooling rate, and the final mechanical properties of the weld metal; the formula of the model is: in, σ represents the target mechanical properties of the weld metal, including tensile strength, yield strength, or impact toughness. I is the welding current (A); U is the welding voltage (V); V is the welding speed (cm / min); t is the time coefficient (s) related to plate thickness or thermal conduction; C, Mn, Si, ... are the specific chemical element contents (%) of the base material and welding material. The specific functional form f and coefficients of the model are obtained by machine learning training on more than 12,000 sets of welding parameter-quality result data in the initial training set. The model is then embedded in the structural layer of the three-layer digital twin model for real-time prediction of weld mechanical properties under different welding parameters.

[0028] Through the detailed implementation of step S1 above, this embodiment addresses three key issues in the construction of digital twin models: First, it solves the problem of incomplete data dimensions and lack of correlation between process parameters and quality results in traditional modeling processes by constructing a high-quality training set through a combination of historical and experimental data; second, it overcomes the limitation that a single model cannot reflect the multi-dimensional characteristics of construction by achieving full-element digital mapping through a three-layer architecture of foundation-structure-construction; and third, it breaks through the technical bottleneck of static model parameter solidification and disconnection from actual construction conditions by establishing parameter association rules to enable the model to have dynamic response capabilities.

[0029] The significant effects achieved in this embodiment are as follows: First, the multivariate nonlinear regression model, trained with over 12,000 sets of welding parameter-quality result data, enables precise quantitative mapping between welding process parameters and weld mechanical properties, allowing the model to predict final quality from process parameters. Second, the dual verification mechanism of laser scanning and mechanical testing controls the consistency error between the model and the physical entity to within 1%, ensuring the high fidelity of the digital twin model. Finally, by pre-setting multi-dimensional correlation rules for component displacement and foundation settlement, welding temperature and material properties, a cross-layer data linkage mechanism is established, laying a solid foundation for subsequent real-time monitoring and precise analysis. These innovations transform the digital twin model from a simple three-dimensional visualization model into an intelligent analysis platform integrating physical laws and process knowledge.

[0030] In a preferred embodiment, step S2 specifically includes: S2.1 Multi-source equipment deployment: Multiple data acquisition devices are deployed at the construction site to collect construction data; the data acquisition devices include displacement sensors installed at the connection points of load-bearing components, infrared temperature sensors installed in the welding operation area, and drones equipped with high-definition cameras; the measurement accuracy of the displacement sensors is ±0.01mm; the temperature measurement range of the infrared temperature sensors is 0-1500℃; the drones are equipped with 4K resolution high-definition cameras and inspect the overall structure every 30 minutes to collect data on the flatness of component installation.

[0031] S2.2 Data Preprocessing: Outlier removal and data standardization are performed on the received construction data. Outlier removal adopts the 3σ criterion, and construction data that exceeds the preset normal value range is automatically marked and triggers secondary acquisition. Data standardization converts construction data of different dimensions into standardized values ​​in the range of 0-1. For welding temperature data, the preset normal value range is 800-1300℃.

[0032] S2.3 Multi-source data fusion: A weighted fusion algorithm is used to fuse data of the same type from different acquisition devices to generate fused data; the weighted fusion algorithm adopts different weight configuration strategies for different types of monitoring data; the same type of data includes at least component displacement data and welding quality data; S2.4, Virtual-Real Synchronization: The fused data is synchronized to the corresponding level of the three-layer digital twin model in real time, and the time difference between the three-layer digital twin model and the actual construction status is ensured not to exceed the set threshold.

[0033] In step S2.3, the component displacement data is calculated using the following weighted fusion formula: in, The displacement value after fusion. The displacement value collected by the displacement sensor. The displacement value is extracted from the UAV image recognition, where α is the weight of the displacement sensor data with a value of 0.7, and β is the weight of the UAV image recognition displacement data with a value of 0.3. In step S2.3, the following weighted fusion formula is used to generate a comprehensive evaluation index for welding quality data: in, As a comprehensive evaluation index for welding quality, These are standardized values ​​from infrared temperature sensor data. The standardized numerical values ​​for the feature score of the weld appearance image. The ultrasonic flaw detection pass rate is given by ω1, which is the weight of temperature data, ω2, which is the weight of weld appearance image feature data, with a value of 0.3, and ω3, which is the weight of ultrasonic flaw detection data, with a value of 0.2. The synchronization mentioned in step S2 specifically involves establishing a dual-channel synchronization mechanism for real-time transmission and dynamic updates. When the data transmission delay is detected to exceed a set threshold, the system automatically switches to a backup communication link to ensure that the time difference between the digital twin model and the actual construction status does not exceed 5 minutes.

[0034] Furthermore, before synchronization, a weighted fusion algorithm is used to fuse the same type of data obtained from multiple acquisition devices to obtain fused data with higher precision.

[0035] In a preferred embodiment, step S3 specifically includes: S3.1, Basic Layer Analysis: Input the real-time collected foundation settlement data into the pre-trained time series prediction model to obtain the foundation settlement prediction value for future periods; compare the foundation settlement prediction value with the preset allowable settlement threshold in the basic layer of the model; if the foundation settlement prediction value exceeds the threshold, it is marked as a potential quality hazard in the basic layer. The time series prediction model is the ARIMA model; the input of the ARIMA model is the daily data sequence of foundation settlement over the past 30 days, and the output is the predicted daily settlement value for the next 7 days; the allowable settlement threshold is an average daily settlement of no more than 0.167 mm.

[0036] S3.2 Structural Layer Analysis: The real-time collected component installation coordinate data is compared with the preset three-dimensional coordinate threshold in the model structural layer to calculate the deviation value; a multi-parameter coupling analysis algorithm is used to comprehensively analyze multiple parameters of the welding process to determine the welding quality level. S3.3 Construction Layer Analysis: The real-time acquired construction operation video stream is analyzed using a behavior recognition algorithm to identify whether the operation behavior is compliant. The behavior recognition algorithm uses a YOLOv8 object detection model combined with an action classification model. The algorithm first identifies the construction personnel, tools, and components in the video stream, then analyzes their movement trajectory and action characteristics, and compares them with a preset standard action feature library to determine whether the operation is compliant.

[0037] S3.4 Cross-level correlation analysis: When a quality anomaly is detected at any level, the system automatically retrieves and correlates historical and real-time data from other levels to comprehensively pinpoint the root cause of the quality problem. The cross-layer correlation analysis specifically refers to: When a component displacement exceeds the tolerance in the structural layer, the system automatically retrieves the foundation settlement change data of that area in the foundation layer over the past 24 hours, as well as the installation procedure record data of that component in the construction layer. If the foundation settlement data shows that the settlement exceeds the limit and the trend of change is synchronized with the displacement deviation, then the root cause of the problem is determined to be foundation settlement. If the foundation settlement data is normal, but the installation procedure record shows that the positioning clamps were not used in accordance with the specifications, then the root cause of the problem is determined to be an installation operation error.

[0038] In step S3.2, the multi-parameter coupling analysis algorithm calculates the comprehensive welding quality score using the following formula: in, To determine the overall score for welding quality, This is a standardized value for the welding current. This is a standardized value for the welding voltage. This is a standardized evaluation value for the characteristics of the welding time versus temperature curve. This is a standardized value for weld height. The standardized value weight of the welding current. The standardized value weights for welding voltage, The standardized evaluation values ​​for the characteristics of the time-temperature curve are weighted. 7 represents the standardized weight of the weld height; like If the value is ≥ 0.8, the welding quality is deemed acceptable; otherwise, it is deemed unacceptable. The specific problem type is marked according to the deviation of each standardized parameter value. The problem types include incomplete fusion and insufficient weld size. The condition for incomplete fusion is that the duration of the weld temperature curve below 800°C exceeds 5 seconds.

[0039] Through the implementation of steps S2 and S3 above, this embodiment systematically solves the key technical problems in construction quality monitoring: First, it breaks through the limitations of traditional monitoring methods, which have single data sources and insufficient accuracy. By using multi-source heterogeneous sensors to collect data collaboratively and weighted fusion processing, it achieves the complementarity and enhancement of monitoring data in the spatiotemporal dimension. Second, it overcomes the problems of isolated quality analysis dimensions and difficulty in tracing the source. By establishing a three-layer linkage analysis mechanism of foundation-structure-construction, it achieves a systematic diagnosis from surface phenomena to root causes.

[0040] The innovative effects achieved in this embodiment are mainly reflected in three aspects: At the data acquisition level, the accuracy of monitoring data is improved by multi-source collaboration of displacement sensors, infrared temperature sensors, and UAV inspections, combined with a weighted fusion algorithm based on specific weight allocation; At the analysis and early warning level, the ARIMA time series model is used to achieve predictive analysis of foundation settlement, the welding quality is comprehensively evaluated through a multi-parameter coupling algorithm, and the YOLOv8 behavior recognition technology is combined to achieve all-round monitoring of the construction process, so that the accuracy rate of identifying quality hazards reaches more than 95%; At the problem diagnosis level, an innovative cross-layer correlation analysis mechanism is established. When structural displacement exceeds the tolerance, it can automatically correlate and analyze the foundation settlement trend and construction operation records, accurately distinguish the root causes of quality problems such as foundation settlement or installation errors, and fundamentally solve the pain point of traditional methods that "only report abnormalities without investigating the root cause". These technological innovations together constitute a construction quality control system that integrates precise monitoring, intelligent diagnosis, and predictive early warning.

[0041] In a preferred embodiment, step S4 specifically includes: S4.1, Tiered Early Warning: Based on the severity of the quality problems obtained from the tiered quality analysis in step S3, different levels of early warning are triggered; the early warning levels include at least a Level 1 early warning requiring immediate work stoppage and rectification, an important Level 2 early warning requiring timely handling, and a general Level 3 early warning for record-keeping only; The triggering conditions for the first-level early warning are that the component displacement exceeds the tolerance by ≥10mm or the welding temperature remains below 700℃. After triggering, the system will issue an audible and visual alarm within ≤10 seconds and push a notification containing the precise location and severity of the problem to the mobile terminal of the management personnel. The triggering condition for the Level 2 early warning is a component displacement exceeding the tolerance by 5-10mm or a weld height deviation of 2-3mm; after triggering, the system pushes early warning information containing problem details and preliminary analysis conclusions to the on-site management terminal within ≤30 seconds. The trigger condition for the Level 3 early warning is that the component displacement exceeds the tolerance by 2-5mm; after triggering, the system only records and generates a daily early warning summary report in the background.

[0042] S4.2 Quantitative Adjustment Instruction Generation: Based on the simulation analysis results of the digital twin model, generate executable adjustment instructions containing specific operation parameters; the adjustment instructions include component correction instructions and welding parameter adjustment instructions; The component correction instructions specifically include: In the digital twin model, with the goal of correcting a displacement deviation of no more than 2mm, the required corrective force is calculated using a mechanical simulation module; the corrective force is calculated using the following mechanical formula: F = (ΔL × E × S) / L in, F is the required corrective force (N), ΔL is the displacement to be corrected (m), and E is the elastic modulus of the component material (Pa). S is the cross-sectional area of ​​the component (m²), and L is the length of the component (m). Based on the calculated corrective force F, the specific steps for applying the corrective force in stages, the required tools and their installation locations are planned to form an executable corrective instruction. The application of the corrective force in stages specifically means that the calculated corrective force F is applied in three stages, with each stage having a force of F / 3; the instruction explicitly specifies the use of two hydraulic jacks with a rated load capacity of 50kN connected in parallel for correction.

[0043] The welding parameter adjustment instructions specifically include: Retrieve the welding temperature-current-voltage correlation model and historical qualified data stored in the three-layer digital twin model; With the goal of adjusting the welding temperature to the optimal range, the required current (I) and voltage (U) values ​​are calculated backward using the aforementioned correlation model; the correlation model is as follows: T = a × I + b × U + c in, T is the welding temperature (°C), I is the welding current (A), U is the welding voltage (V), and a, b, c are model coefficients obtained by training with historical data; The calculated current and voltage values, along with their corresponding simulated welding strength results, are used to generate adjustment instructions.

[0044] In a preferred embodiment, step S5 specifically includes: S5.1 Construction of full-chain traceability archive: Construct a construction quality traceability archive containing problem details, rectification instructions and rectification results, and store the construction quality traceability archive in a blockchain database to ensure that the data is tamper-proof; Each record in the construction quality traceability archive generates a unique hash value and stores it in association with the hash value of the previous record, forming a chain structure; the construction quality traceability archive supports retrieval by issue ID or component number, and the retrieval response time is no more than 2 seconds; The aforementioned historical qualified data refers to the set of effective data that corresponds one-to-one with the final quality inspection results of welding parameters, including current, voltage, and speed, selected from the archives of similar steel structure power room projects that have been accepted and qualified in the past, such as the last 5 years. The purpose is to enable the digital twin model to learn from the industry's proven and reliable process experience and achievement standards. The experimental data refers to the precise quality results obtained in a laboratory environment using the same materials as on-site, such as Q355 steel specimens and equipment, by systematically adjusting welding parameters and conducting repeatable tests under each set of parameters. The purpose is to supplement the parameter combinations and results that may be missing in historical data under a controlled environment, paying particular attention to fault boundary conditions, thereby giving the model a more scientific and comprehensive causal mapping relationship. During the construction of this project, real-time data such as displacement, temperature, images, and operational behavior are continuously collected through multi-source acquisition devices. After preprocessing and fusion, this data is synchronized to the digital twin model and accumulates over time to form the project's unique historical construction data. This historical construction data fully records all statuses, operations, warnings, and rectification results from the start of construction to the present moment, and is a digital archive of the entire life cycle of the project. S5.2 Dynamic Iterative Optimization of the Model: Based on historical construction data, machine learning algorithms are used to optimize and update the parameter thresholds and analysis algorithms in the three-layer digital twin model; The parameter threshold optimization specifically involves: A random forest model is adopted, with welding parameters in the historical construction process as input features and the quality pass probability as output. The optimal temperature threshold that makes the pass probability not less than 99% is found through training the random forest model, and the lower limit threshold of welding temperature in the random forest model is updated to the optimal temperature threshold. The optimization of the analysis algorithm is as follows: For analysis scenarios where the recognition accuracy is below 90%, the multi-parameter coupled analysis algorithm is retrained by adding historical data samples for that scenario; cross-validation is used to adjust the parameter weights and decision thresholds in the algorithm, thereby improving the recognition accuracy of the test set to over 95%.

[0045] Through the implementation of steps S4 and S5 above, this embodiment innovatively solves two key problems in the quality control closed loop: First, it breaks through the limitations of traditional early warning systems, such as delayed response, ambiguous hierarchical structure, and lack of precise execution guidance, and establishes a quantitative decision-making mechanism based on digital twin simulation; Second, it overcomes the defects of poor data traceability, fixed model parameters, and lack of self-optimization capability in traditional quality management, and constructs a continuously evolving intelligent control system.

[0046] The significant effects achieved in this embodiment are as follows: In terms of early warning response, the establishment of a three-level early warning mechanism enables precise control of quality issues through hierarchical classification, resulting in a significant improvement in response efficiency; in terms of rectification execution, the precise calculation of corrective force based on mechanical formulas and the adoption of a three-stage application strategy, combined with the back-calculation optimization of welding parameter correlation models, has achieved a major shift from experience-based judgment to quantitative execution, significantly improving the first-time pass rate; in terms of quality traceability, blockchain hash chain storage ensures data immutability, supports rapid retrieval within 2 seconds, and establishes a fully trusted traceability system; in terms of system optimization, a random forest model is used to optimize parameter thresholds with a target pass rate of 99%, and cross-validation improves the identification accuracy from 90% to over 95%, enabling the system to have continuous self-evolution capabilities. These innovations together constitute a complete quality control closed loop integrating intelligent early warning, precise execution, trusted traceability, and continuous optimization.

[0047] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0048] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for full-process quality control of steel structure power room construction based on digital twin, characterized in that, The method includes the following steps: Obtain information about the construction objects and divide them into levels, then construct and train a three-layer digital twin model; Collect and process construction data; A three-layer digital twin model is used to perform hierarchical quality analysis on the processed construction data. When a quality anomaly is detected, cross-layer correlation analysis is initiated to locate the root cause of the problem. Based on the analysis results of the layered quality analysis, a graded early warning is triggered, and quantitative construction adjustment instructions are generated based on the simulation of the three-layer digital twin model. The construction quality traceability archive is constructed and uploaded to the blockchain, and the parameter thresholds and analysis algorithms of the three-layer digital twin model are dynamically iteratively optimized based on historical construction data.

2. The method for full-process quality control of steel structure power room construction based on digital twin as described in claim 1, characterized in that, The construction of the three-layer digital twin model specifically includes: Historical qualified data and experimental data from the construction archives of similar steel structure power rooms are extracted, along with experimental data from a welding environment built in a laboratory consistent with the site. The historical qualified data includes valid data corresponding to welding current, voltage, time, weld flaw detection pass rate, and mechanical property test results. The experimental data includes quality results of weld appearance quality, internal defects, and tensile strength. Information on the construction object is obtained, and an initial training set of welding parameter-quality result data pairs is constructed. The initial training set is then deduplicated and outlier is removed. Based on data type, the construction object information is divided into a foundation layer, a structural layer, and a construction layer; the foundation layer includes geological survey data, the structural layer includes design drawing data, and the construction layer includes construction plan data. Based on BIM technology, a three-layer digital twin model corresponding to the foundation layer, structural layer and construction layer is constructed, and parameter association rules are preset in the model; the parameter association rules include the linkage rules between component displacement data and foundation settlement data, as well as the mapping rules between welding temperature data and material mechanical property data; The three-dimensional coordinate data of the entity structure obtained by laser scanning is compared with the coordinates of the basic layer of the model. When the error exceeds the preset threshold, the model parameters are adjusted. The simulation results of the mechanical performance of the key components are verified with the mechanical test data of the entity components. When the consistency error between the three-layer digital twin model and the entity is less than the error threshold, the trained three-layer digital twin model is obtained.

3. The method for full-process quality control of steel structure power room construction based on digital twin as described in claim 2, characterized in that, The mapping rule and formula between the welding temperature data and the material mechanical property data are as follows: in, σ represents the target mechanical properties of the weld metal, including tensile strength, yield strength, or impact toughness. I is the welding current; U is the welding voltage; V is the welding speed; Represents a mapping function; t is a time coefficient related to plate thickness or thermal conduction; C, Mn, Si, ... are the specific chemical element contents of the base material and welding material.

4. The method for full-process quality control of steel structure power room construction based on digital twin as described in claim 1, characterized in that, The collection and processing of construction data specifically includes: Multiple data acquisition devices are deployed at the construction site to collect construction data; the data acquisition devices include displacement sensors installed at the connection points of load-bearing components, infrared temperature sensors installed in the welding operation area, and drones equipped with high-definition cameras; Outlier removal and data standardization are performed on the construction data. The outlier removal adopts the 3σ criterion, and construction data that exceeds the preset normal value range is automatically marked and triggers secondary data collection. The data standardization process includes converting construction data of different dimensions into standardized values ​​in the range of 0-1. A weighted fusion algorithm is used to fuse construction data of the same type from different acquisition devices to generate fused construction data; the same type of construction data includes at least component displacement data and welding quality data. The integrated construction data will be synchronized in real time to the corresponding level of the three-layer digital twin model.

5. The method for full-process quality control of steel structure power room construction based on digital twin as described in claim 4, characterized in that, The component displacement data is calculated using the following weighted fusion formula: in, The displacement values ​​of the fused components. The component displacement values ​​collected by the displacement sensor. The component displacement value is extracted by UAV image recognition, α is the displacement weight corresponding to the displacement sensor, and β is the displacement weight corresponding to UAV image recognition. For welding quality data, the following weighted fusion formula is used to generate a comprehensive welding quality evaluation index: in, As a comprehensive evaluation index for welding quality, These are standardized values ​​for welding temperature data. Standardized scoring of weld appearance image feature data, ω1 represents the pass rate of ultrasonic flaw detection data, ω2 represents the weight of welding temperature data, ω3 represents the weight of weld appearance image feature data, and ω3 represents the weight of ultrasonic flaw detection data.

6. The method for full-process quality control of steel structure power room construction based on digital twin as described in claim 1, characterized in that, The hierarchical quality analysis and cross-hierarchical correlation analysis specifically include: The basic layer analysis involves inputting real-time collected foundation settlement data into a pre-trained time series prediction model to obtain predicted foundation settlement values ​​for future periods. The predicted foundation settlement values ​​are then compared with the preset allowable settlement threshold in the basic layer of the model. If the predicted foundation settlement values ​​exceed the predicted threshold, they are marked as potential quality hazards in the basic layer. Structural layer analysis compares the real-time collected component installation coordinate data with the preset three-dimensional coordinate thresholds in the model's structural layer to calculate the deviation value; a multi-parameter coupling analysis algorithm is used to comprehensively analyze multiple parameters of the welding process to determine the welding quality level. Construction layer analysis involves analyzing real-time collected construction operation video streams using behavior recognition algorithms to identify whether the operational behaviors are compliant. Cross-level correlation analysis automatically retrieves and correlates historical and real-time data from other levels when a quality anomaly is detected at any level, comprehensively locating the root cause of the quality problem. The cross-layer correlation analysis specifically includes: When a component displacement exceeds the tolerance in the structural layer, the system automatically retrieves the foundation settlement data of the area in the foundation layer at the previous time point, as well as the installation process record data of the component in the construction layer. If the settlement amount in the foundation settlement data exceeds the limit and the trend of change is synchronized with the displacement deviation, then the root cause of the problem is determined to be foundation settlement. If the foundation settlement data is normal, but the installation procedure record shows that the positioning clamps were not used in accordance with the specifications, then the root cause of the problem is determined to be an installation operation error.

7. The method for full-process quality control of steel structure power room construction based on digital twin as described in claim 6, characterized in that, The multi-parameter coupled analysis algorithm calculates the comprehensive welding quality score using the following formula: in, To determine the overall score for welding quality, This is a standardized value for the welding current. This is a standardized value for the welding voltage. This is a standardized evaluation value for the characteristics of the welding time versus temperature curve. This is a standardized value for weld height. The standardized value weight of the welding current. The standardized value weights for welding voltage, The standardized evaluation values ​​for the characteristics of the time-temperature curve are weighted. 7 represents the standardized weight of the weld height.

8. The method for full-process quality control of steel structure power room construction based on digital twin as described in claim 1, characterized in that, The triggering of tiered early warnings and the generation of quantitative construction adjustment instructions specifically include: Based on the severity of the quality problems obtained from the stratified quality analysis, different levels of early warnings are triggered; the early warning levels include at least a Level 1 early warning requiring immediate work stoppage and rectification, a Level 2 important early warning requiring timely handling, and a Level 3 general early warning for record-keeping only; Based on the simulation analysis results of the three-layer digital twin model, executable adjustment instructions containing specific operation parameters are generated; the adjustment instructions include component correction instructions and welding parameter adjustment instructions. The component correction instructions specifically include: In the three-layer digital twin model, with the goal of correcting displacement deviation to be no greater than 2mm, the required corrective force is calculated using a mechanical simulation module; the corrective force is calculated using the following mechanical formula: F = (ΔL × E × S) / L in, F is the required corrective force, ΔL is the displacement to be corrected, E is the elastic modulus of the component material, S is the cross-sectional area of ​​the component, and L is the length of the component. Based on the required corrective force F, the specific steps for applying the corrective force in stages, the required tools and their installation positions are planned according to the preset planning standards, thus forming an executable corrective instruction; The welding parameter adjustment instructions specifically include: Retrieve the welding temperature-current-voltage correlation model and historical qualified data stored in the three-layer digital twin model; The required current and voltage values ​​are calculated by reverse engineering using the aforementioned correlation model; the correlation model is as follows: T = a × I + b × U + c in, T is the welding temperature, I is the welding current, U is the welding voltage, and a, b, c are model coefficients obtained by training with historical data; The calculated current and voltage values, along with their corresponding simulated welding strength results, are used to generate an adjustment instruction according to the adjustment instruction template.

9. The method for full-process quality control of steel structure power room construction based on digital twin as described in claim 1, characterized in that, The construction quality traceability archive is uploaded to the blockchain, and the parameter thresholds and analysis algorithms of the three-layer digital twin model are dynamically iteratively optimized based on historical construction data. Specifically, this includes: Construct a construction quality traceability file containing problem details, rectification instructions, and rectification results, and upload the construction quality traceability file to the blockchain; Each record in the construction quality traceability archive generates a unique hash value, which is stored in association with the hash value of the previous record, forming a chain structure; the construction quality traceability archive supports retrieval by issue ID or component number; Based on historical construction data, machine learning algorithms are used to optimize and update the parameter thresholds and analysis algorithms in the three-layer digital twin model; The optimization process for the parameter threshold includes: A random forest model is used, with welding parameters from historical construction processes as input features and the probability of quality compliance as output. The optimal temperature threshold is found through training the random forest model, and the lower limit threshold of welding temperature in the random forest model is updated to the optimal temperature threshold.

Images