BIM-based steel structure full life cycle digital management and control method and system

Abstract

The application discloses a BIM steel structure full-life cycle digital management and control method and system, which comprises the following steps: S1: a design stage, a main control BIM model is established, a global unique identifier (GUID) is allocated to each steel structure component, and a drawing model consistency check and a model model consistency check are performed, and any check is not passed to prohibit entering the next stage; actual point cloud data of an installed area is acquired through three-dimensional laser scanning, a continuous deviation field function is constructed after registration with the main control BIM model, and a position-sensitive dynamic tolerance field is generated, on the basis of which, optimal pose optimization is performed on the to-be-installed component, so that subsequent installation can adaptively compensate for the existing cumulative deviation, the method effectively avoids the problems that the component cannot be positioned or additional stress is introduced due to forced correction caused by the cumulative deviation being too large, and completely solves the installation 'dead knot', which is a core bottleneck restricting the steel structure construction efficiency and safety.

Description

Technical Field

[0001] This invention relates to the field of prefabricated steel structure technology, and in particular to a BIM-based digital management and control method and system for the entire life cycle of steel structures. Background Technology

[0002] Steel structures are widely used in large public buildings, super high-rise buildings and industrial plants due to their advantages such as high strength, light weight and high degree of industrialization. With the popularization of Building Information Modeling (BIM) technology, steel structure projects have begun to gradually adopt digital means for management in all stages of design, production, construction and operation and maintenance.

[0003] During steel structure construction, component installation typically uses theoretical coordinates determined in the design phase for positioning. However, the installed area inevitably accumulates deviations due to factors such as construction errors, uneven foundation settlement, temperature effects, and welding deformation. Under current technology, this actual on-site deviation information cannot be effectively fed back to the installation guidance stage, causing subsequent components to still be installed according to fixed theoretical design coordinates. When the accumulated deviation is large, the components cannot be correctly positioned. If forced correction is attempted, excessive additional stress may be introduced, endangering structural safety. At the same time, data flow between different stages is not smooth, and the design model, production data, and actual construction site conditions are disconnected, making it difficult to foresee and avoid deviation problems in the design or production stages. Therefore, there is an urgent need for a BIM-based digital management method and system for the entire lifecycle of steel structures to solve this problem. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art. To this end, this invention proposes a BIM-based digital management method and system for the entire lifecycle of steel structures, as detailed below: On the one hand, it provides a BIM-based digital management and control method for the entire lifecycle of steel structures. Includes the following steps: S1: During the design phase, establish the master BIM model, assign a globally unique identifier (GUID) to each steel structure component, and perform drawing-model consistency verification and model-model consistency verification. If either verification fails, the process is prohibited from proceeding to the next phase. S2: During the production stage, CNC cutting code, welding robot path, assembly sequence animation and component QR code are automatically generated based on the master control BIM model to realize drawing-free manufacturing. The production status data is linked to the master control BIM model in real time by scanning the QR code. After the component is completed, the CNC model consistency verification is performed. If the verification fails, the component is prohibited from leaving the factory. S3: During the construction phase, the actual position point cloud of the installed components is obtained through 3D scanning and registered with the main control BIM model via ICP. The actual deviation of the installed components is calculated, and a deviation field function D(x,y,z) and a dynamic tolerance field function T(x,y,z) are constructed. The deviation of the physical model is verified using T(x,y,z) as the criterion. If the actual deviation of any installed component at its corresponding spatial position exceeds T(x,y,z), the verification fails, the corresponding installation area is locked, and a rectification notice is issued. After the verification passes, the optimal installation pose of the component to be installed is calculated based on the dynamic tolerance field, which replaces the theoretical design coordinates for installation guidance. S4: After hoisting is completed, the actual installation deviation value of each component is compared and verified with the corresponding allowable deviation range in the dynamic tolerance field; if the verification is successful, the key milestone of hoisting completion is confirmed to be effective, and the weld inspection task is automatically triggered; if the verification fails, the status is locked and a rectification notice is issued. S5: Throughout the entire lifecycle, key milestones are set as design sealing, component delivery, hoisting completion, and weld inspection qualification. The next stage of the task can only be triggered after each milestone is verified and effective, and all milestone information is written to the blockchain for evidence storage. S6: During the operation and maintenance phase, a graph neural network model G=(V,E) is constructed with steel structure components as nodes and the connection relationships between components as edges. The installation deviation recorded during the construction phase is used as one of the node features. The remaining life of each component is predicted by aggregating neighborhood information through the graph neural network. When the predicted remaining life is lower than the preset threshold, a maintenance work order is automatically generated.

[0005] As an improvement to the above technical solution, the drawing-model consistency verification uses primitive feature matching and deep learning comparison methods to verify the consistency between the two-dimensional drawings and the master control BIM model. The model-model consistency verification uses IFC collision detection and topology relationship verification. The digital-model consistency verification automatically compares the component attributes with the model attributes by scanning the component's QR code. The physical model deviation verification uses three-dimensional laser scanning and ICP registration methods to calculate the deviation between the actual component and the model.

[0006] As an improvement to the above technical solution, after registering the actual location point cloud of the installed component with the master control BIM model via ICP, the deviation field function D(x,y,z) at the spatial point (x,y,z) is obtained. D(x,y,z) represents the distance deviation between the actual position and the theoretical design position of the component at that point. The dynamic tolerance field function T(x,y,z) is defined as: in, To standardize the basic tolerance value, Let be the Laplace operator for the bias field in the installed region, and represent the curvature of the bias variation. The value ranges from 0 to 0.5, and is dynamically adjusted bidirectionally based on the cumulative trend of the measured deviation. When the deviation shows an accelerating trend, α is increased to further widen the tolerance in the accelerated divergence region and further tighten the tolerance in the convergence and positivity region. When the deviation tends to converge globally, α is decreased to make the tolerance field as a whole approach the basic tolerance value T0. This is the absolute value of the maximum curvature across the entire field, used for normalization.

[0007] As an improvement to the above technical solution, the formula for calculating the optimal installation pose is: ; in, The theoretical design pose of the component to be installed; The distance between the candidate pose and the theoretical pose; denoted as , where is the additional stress value caused by the candidate pose; λ is the stress penalty coefficient. The yield strength of the component material.

[0008] As an improvement to the above technical solution, the neighborhood information aggregation includes a single-layer propagation formula and a loss function. The single-layer propagation formula of the graph neural network model is as follows: in, For node i in the th... The feature vector of the layer, Let i be the set of neighboring nodes. For neighbor node j in the th... The feature vector of the layer, Let W0, W1, and W2 be the feature vectors of edge (i, j), and let W0, W1, and W2 be the learnable weight matrices.

[0009] As an improvement to the above technical solution, the loss function of the graph neural network model is: ; in, This is the mean square error loss; This is a physical constraint regularization term that penalizes the case where the predicted lifetime of node i on the directed edge (i, j) exceeds the predicted lifetime of node j and the connection type tolerance. The sum of the cases; λ is the regularization coefficient.

[0010] As an improvement to the above technical solution, in step S4, the comparison and verification adopts a method combining numerical comparison and hash evidence storage: first, the actual installation deviation value δ is determined. actual Does it not exceed the allowable deviation range ε ​​corresponding to the component in the dynamic tolerance field? max If δ actual ≤ ε maxThen it is determined to be compliant, if δ actual >ε max If it is deemed non-compliant, the status will be locked; at the same time, the allowable deviation range ε ​​will be reduced. max Deviation from measured value δ actual Generate hash values ​​respectively allowed With Hash actual And write it into the blockchain.

[0011] As an improvement to the above technical solution, the automatic triggering of the next stage task in S5 is achieved through distributed automatic execution rules. All task triggering is premised on the effectiveness of the preceding milestone. The distributed automatic execution rules are configured as follows: after the design sealing milestone takes effect, the component list is automatically pushed to the production system; after the component leaving the factory milestone takes effect, the transportation task is automatically dispatched; after the hoisting completion milestone takes effect, the weld quality inspection work order is automatically generated; after the weld inspection qualified milestone takes effect, the as-built model is automatically updated.

[0012] On the other hand, a BIM-based digital management and control system for the entire lifecycle of steel structures is provided, applying the methods described above. The master control BIM model module is used to establish a master control model shared throughout the entire lifecycle, assign a globally unique identifier to each steel structure component, and store the incremental change history. The verification module is used to perform consistency checks on drawing model, digital model, model-to-model, and physical model deviation. If any verification fails, the corresponding stage of the process will be prohibited from proceeding. The drawing-free manufacturing data generation module is used to automatically generate CNC cutting codes, welding robot paths, assembly sequence animations, and component QR codes from the master control BIM model. The dynamic tolerance calculation module is used to receive 3D scan point cloud data, perform point cloud registration, tolerance field construction and optimal installation pose solution, and replace theoretical design coordinates for installation guidance. The automatic verification and triggering module is used to perform deviation comparison verification after hoisting. If the verification passes, the weld inspection task is automatically triggered. If it fails, the weld is locked and a rectification notice is issued. At the same time, key milestone information is written into the blockchain and the next stage task is automatically triggered. The graph neural network health prediction module is used to construct a graph model with steel structure components as nodes, predict the remaining life of the components with installation deviation as a feature, and automatically generate maintenance work orders when the life falls below a threshold. On-site mobile terminals are used to scan component QR codes, upload production and acceptance data, and view installation instructions.

[0013] The beneficial effects of this invention are as follows: The actual point cloud data of the installed area is obtained by 3D laser scanning. After being registered with the master BIM model, a continuous deviation field function is constructed, which in turn generates a position-sensitive dynamic tolerance field. On this basis, the optimal pose of the component to be installed is optimized so that the subsequent installation can adaptively compensate for the existing accumulated deviation. This method effectively avoids the problem of components not being able to be positioned or the introduction of additional stress due to forced correction caused by excessive accumulated deviation. It completely solves the core bottleneck of installation "dead knot" that restricts the efficiency and safety of steel structure construction. By assigning a globally unique identifier (GUID) to each steel structure component, this identifier accompanies the component throughout the entire process from design to demolition. All subsequent data, such as design parameters, production status, installation deviations, and inspection records, are linked with the GUID as the main thread, achieving seamless data connection across stages and participants. Any component can trace its entire life cycle file with one click, providing complete data support for deviation analysis and quality traceability. A graph neural network health prediction model is constructed based on the as-built model. The initial feature vector of the node is integrated with the final installation deviation value of the component recorded during the construction stage. The loss function is designed with a physical constraint regularization term to penalize the case where the difference in the predicted life of connected components exceeds the tolerance. Detailed Implementation

[0014] The embodiments of the present invention are described in detail below.

[0015] Example 1 This embodiment provides a BIM-based digital management and control method for the entire life cycle of steel structures. Taking a large-span spatial steel structure of a large public building as an example, its main body is a space frame structure, and the nodes are connected by welded balls and steel pipes. The entire process is managed using the digital management and control system of this invention.

[0016] S1: Design Phase In the collaborative modeling platform, the design team used 3D BIM software to create a full-discipline master control model of the steel structure. During the model creation process, the system's built-in master control BIM model module automatically assigned a globally unique identifier (GUID) to each steel structure component, such as "STADIUM-TRUSS-01-NODE-0012". Once generated, the GUID is fixed and accompanies the entire life cycle of the component. All subsequent data associations are based on the GUID. The GUID follows the entire process of the component from design to demolition and serves as the core field for data association. After the master BIM model is built, the consistency verification process is initiated, and the following two sub-verifications are performed during the design phase: Drawing-Model Consistency Verification: The two-dimensional detailed drawings generated by the design are compared with the master BIM model. Feature vectors of graphic elements (such as axes, holes, and cross-sectional contours) in the drawings are extracted. At the same time, corresponding feature sets are generated from the IFC data of the model. A trained deep convolutional neural network is used for feature matching and difference detection. The dimensions, locations, or quantities that are inconsistent between the drawings and the model are automatically marked, and a difference report is output.

[0017] Model-to-Model Consistency Verification: The master BIM model is exported in IFC format, and then collision detection and topology verification are performed. An octree-based spatial segmentation algorithm is used to quickly identify intersecting members, abnormal gaps, or connection logic errors, and to verify the consistency between the model hierarchy and the actual assembly topology.

[0018] After both verifications pass, the system performs a SHA-256 hash operation on the model file and its attributes. The resulting hash value is signed by the design manager using a PKI digital certificate to generate a digital archive record. The hash value and signature are stored on the blockchain, and the model status is changed to "design archive status". Any modification must be unsealed and re-verified. At the same time, based on distributed automatic execution rules, after the design is archived, a component manufacturing list with GUID and geometric information is automatically pushed to the production management system, prohibiting model data that fails verification from entering the production stage. S2: Production Stage After receiving the component list, the multi-source data fusion interface deployed in the production workshop automatically extracts the three-dimensional geometric information of each component from the model and generates three types of data: first, DSTV / NC cutting code that can be directly recognized by CNC machine tools; second, the six-axis motion path file of the welding robot; and third, a three-dimensional assembly sequence animation based on the model to guide workers in assembly. At the same time, a unique QR code containing a GUID and access address is generated online for each component and attached to the surface of the component by laser marking or weather-resistant labeling. At the beginning and end of each process, such as cutting, welding, straightening, and pre-assembly, workers use on-site mobile terminals to scan the component's QR code and upload production status data, including operator, equipment number, parameters, and self-inspection photos. This data is written back to the corresponding GUID node of the main control BIM model in real time through the IoT gateway. Managers can see the component status represented by different colors in the dashboard view on the 3D model, such as "material cutting completed", "welding completed", and "pre-assembly qualified", achieving complete "drawing-free manufacturing" and process transparency. After the component is completed, a consistency verification between the digital model and the component is performed: Quality inspectors use a mobile terminal to scan the QR code on the component. The terminal automatically retrieves the component's attribute set (material, cross-sectional specifications, fire rating, etc.) from the cloud and automatically compares it with the corresponding attributes measured on site or recorded in the quality inspection records. If all attributes match, the verification passes and the component is allowed to leave the factory; otherwise, it is marked as abnormal and the factory process is locked until it is rectified and re-verified. S3: Construction Phase During the construction of the main structure, several installed main trusses and circumferential members formed an initial stable zone. A stationary 3D laser scanner was used to measure the installed area, acquiring point cloud data of approximately ten million points. After denoising and thinning, improved ICP registration was performed with the master BIM model. Based on the registration, the deviation distribution at spatial coordinates (x,y,z) was obtained, and a continuous deviation field function D(x,y,z) was constructed (radial basis function interpolation was used between points to ensure the deviation field was smooth and differentiable). Then, the dynamic tolerance field T(x,y,z) was calculated. After the dynamic tolerance field was constructed, a physical model deviation verification was performed. It should be noted that the verification object of the physical model deviation verification is each independent component in the installed area, while the construction of the dynamic tolerance field T(x,y,z) is based on... The deviation field D(x,y,z) reflects the overall deviation distribution trend of the installed area. By determining the local tolerance threshold through the overall trend and then using this threshold to examine individual components, abnormal components that deviate from the overall trend can be effectively identified, avoiding the excessive deviation of individual components being masked by the overall average effect. The deviation value between the actual position of the installed component and the theoretical position of the master control BIM model is calculated through ICP registration to determine whether the deviation exceeds the dynamic tolerance field value T(x,y,z) corresponding to the spatial position of the component. If it exceeds, the corresponding installation area is locked and a rectification notice is issued. After the rectification is qualified, the scan is re-verified. If it does not exceed, the following steps are continued based on the registration results. Based on the deviation distribution obtained from registration at spatial coordinates (x, y, z), a continuous deviation field function D(x, y, z) is constructed. Radial basis function interpolation is used between points to ensure that the deviation field is smooth and differentiable. Based on this deviation field, the dynamic tolerance field T(x,y,z) is calculated. Specifically, the Laplace operator of the deviation field is first obtained at discrete points in the installed area using the finite difference method. Then take the largest absolute value of the entire game. For normalization, the basic tolerance value T0 is taken as the allowable deviation specified in the standard, for example, T0 = 3mm for the main truss splicing node; the adaptive coefficient α is initially set to 0.4, with an upper limit of 0.5; α is used to control the response gain of the tolerance field to the deviation trend of the installed area—the larger α is, the more significant the spatial difference of the tolerance field. When the deviations of multiple consecutive components are in the same direction and When the deviation continues to increase, it indicates that the deviation is accumulating at an accelerated rate, and the system automatically adjusts α upward to 0.5, so that: in , the divergence zone tolerance is appropriately relaxed to accommodate inevitable cumulative deviations, while in , the convergence zone tolerance is further tightened to promote deviation correction; when the deviation tends to global convergence and the absolute value continues to decrease, the system automatically adjusts α back to 0.4, so that the overall tolerance field approaches the basic tolerance value T0, and the dynamic tolerance field is finally formed according to the following formula: A position-sensitive tolerance field is thus formed: when is a positive value, it indicates that the spatial position deviation shows an accelerating expansion trend, at this time T>T0, and the dynamic tolerance is automatically appropriately relaxed; when is a negative value, it indicates that the spatial position deviation tends to converge and correct, at this time T<T0, and the dynamic tolerance is automatically tightened; the value of α determines the amplitude of the above relaxation and tightening, the larger α is, T is in the region is larger than T0, and in the region is smaller than T0, that is, the higher the degree of spatial differentiation of the tolerance field; For the component to be installed, at its theoretical design pose , a group of candidate poses P including translation and rotation are generated nearby. For each candidate pose, calculate the deviation distance between it and the theoretical pose , and quickly estimate the additional nodal stress generated if forced installation is performed according to this pose through a simplified mechanical model , in this embodiment, a stress penalty coefficient λ=0.015 is introduced, and combined with the dynamic tolerance value T(x,y,z) at the spatial position of the component, optimization is performed according to the following objective function: ; The obtained optimal installation pose will replace the theoretical design coordinates as the actual installation command value. On-site installation personnel receive the optimal pose data through mobile terminals or augmented reality interfaces to guide component hoisting and initial positioning, and complete this round of installation operations; S4: Verification and automatic triggering after hoisting After the component is hoisted and initially tightened, the 3D laser scanning is reused to obtain the point cloud of this area, which is registered to the main control BIM model, and the actual position of the key feature points of the component is extracted, and the actual installation deviation δ is calculated actual , the system queries the dynamic tolerance field value T(X0,Y0,Z0) corresponding to its spatial position (X0,Y0,Z0) according to the GUID of the component, and uses this as the allowable deviation range ε max ; if δ actual ≤ε maxIf the verification is successful, the milestone of hoisting completion is confirmed and the relevant data is written to the blockchain for notarization. Then, two automated processing steps are initiated: first, the deviation value is compared and hashed for notarization; second, a weld inspection work order is automatically generated and pushed to the corresponding personnel. If δ actual >ε max If the deviation exceeds the standard, the component status will be immediately locked as "deviation exceeds the standard", and a rectification notice will be automatically sent to the construction manager, quality inspector and design representative. The subsequent task flow will be locked until the rectification is qualified and the verification is passed again. S5 Full Lifecycle Milestone Management and Blockchain Evidence Storage This project strictly adheres to milestone-based pre-constraints throughout the entire process. If any milestone fails to be verified and effective, or if blockchain notually stored, the system will prohibit the initiation of any subsequent procedures. The four key milestones are, in order: design sealing, component delivery, hoisting completion, and weld inspection qualification. The hoisting completion milestone can only be confirmed as effective after hoisting deviation verification is passed. Upon reaching each milestone, the system collects and packages the following information: component GUID, milestone type code, timestamp, operator's digital identity, hash value of relevant reports (such as quality inspection reports), and the on-chain transaction ID of the previous milestone. This data package is written to the consortium blockchain after consensus among multiple participants, ensuring that each milestone is irreversible and traceable. Distributed automatic execution rules are encoded into a rule engine, whose key actions are as follows: After the design is sealed, the component manufacturing BOM is automatically pushed to the production MES system, and the start-up command is triggered. When components leave the factory, a waybill is automatically generated in the transportation management system to notify the site to prepare for receiving. Once the hoisting is completed and the deviation verification is passed, a weld quality inspection work order will be automatically generated and relevant personnel will be notified. After the weld inspection is qualified, the system automatically writes the completion status back to the main control BIM model and generates a snapshot of the completed model and uploads it to the blockchain.

[0019] Through the above mechanism, we can ensure that each process is strictly driven by milestones and prevent violations such as "using before inspection". S6: Operation and Maintenance Phase At the beginning of the operation and maintenance phase, a graph neural network health prediction model is constructed based on the as-built model. In the graph G=(V,E), the set of nodes V corresponds to all steel components, and the initial feature vector h of each node i is... i This includes: the final installation deviation value (ratio) of the component recorded during the construction phase, component type (column / beam / brace, etc.), material grade, section type, design service life, static load and fatigue level, etc.; the edge set E represents the actual connection relationship between components; and the feature vector e of each edge (i,j) is... ijThis includes connection type (bolted connection, welding, pin, etc.), connection stiffness level, and environmental exposure level. It should be noted that for each physical connection, two directed edges (i,j) and (j,i) are created in the edge set E, and assigned the same edge feature vector e. ij =e ji This ensures that the physical constraint regularization term in the loss function can symmetrically penalize cases where the difference in predicted lifetimes of connected components in any direction exceeds the tolerance τij. Information aggregation is performed using a two-layer graph neural network. The single-layer propagation formula is: ; in, For node i in the th... The feature vector of the layer, Let i be the set of neighboring nodes. For neighbor node j, itself is in the th order. The feature vector of the layer, Let be the feature vector of edge (i, j), W0, W1, and W2 be learnable weight matrices, σ be activated by ReLU, and after two layers of propagation, the entire graph is read out and node features are concatenated. The remaining lifetime y is then predicted by outputting through a fully connected layer. i , During model training, the loss function is defined as: ; in, This is the mean square error loss; This is a physical constraint regularization term that penalizes the case where the predicted lifetime of node i on the directed edge (i, j) exceeds the predicted lifetime of node j and the connection type tolerance. The sum of the cases; λ is the regularization coefficient; The training data comes from over 10 years of regular inspection and maintenance records of similar large steel structures, and the tolerance for connection types is considered. The settings are as follows: 5 years for welded connections, 8 years for high-strength bolt friction connections, and 10 years for ordinary connections. The regularization coefficient λ is set to 0.05, and the mean absolute error on the validation set is controlled within 2.3 years. After the annual inspection data is entered into the database, the system automatically runs graph neural network inference to update the remaining life prediction values ​​of each component. The operation and maintenance management platform sets threshold rules: when the predicted remaining life of a component is less than 15% of the design service life or the absolute value is less than 4 years, a maintenance work order is automatically generated. The work order includes the component GUID, predicted life, suggested inspection items and maintenance instructions, and is pushed to the maintenance engineer. At the same time, the warning information is written to the blockchain as a basis for operation and maintenance decision-making.

[0020] Example 2 The BIM-based digital management and control system for the entire lifecycle of steel structures is applied to the method described in Example 1. The master control BIM model module is used to establish a master control model shared throughout the entire lifecycle, assign a globally unique identifier to each steel structure component, and store the incremental change history. The verification module is used to perform consistency checks on drawing model, digital model, model-to-model, and physical model deviation. If any verification fails, the corresponding stage of the process will be prohibited from proceeding. The drawing-free manufacturing data generation module is used to automatically generate CNC cutting codes, welding robot paths, assembly sequence animations, and component QR codes from the master control BIM model. The dynamic tolerance calculation module is used to receive 3D scan point cloud data, perform point cloud registration, tolerance field construction and optimal installation pose solution, and replace theoretical design coordinates for installation guidance. The automatic verification and triggering module is used to perform deviation comparison verification after hoisting. If the verification passes, the weld inspection task is automatically triggered. If it fails, the weld is locked and a rectification notice is issued. At the same time, key milestone information is written into the blockchain and the next stage task is automatically triggered. The graph neural network health prediction module is used to construct a graph model with steel structure components as nodes, predict the remaining life of the components with installation deviation as a feature, and automatically generate maintenance work orders when the life falls below a threshold. On-site mobile terminals are used to scan component QR codes, upload production and acceptance data, and view installation instructions.

[0021] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it.

Claims

1. A BIM-based digital management and control method for the entire lifecycle of steel structures, characterized in that: Includes the following steps: S1: During the design phase, establish the master BIM model, assign a globally unique identifier (GUID) to each steel structure component, and perform drawing-model consistency verification and model-model consistency verification. If either verification fails, the process is prohibited from proceeding to the next phase. S2: During the production stage, CNC cutting code, welding robot path, assembly sequence animation and component QR code are automatically generated based on the master control BIM model to realize drawing-free manufacturing. The production status data is linked to the master control BIM model in real time by scanning the QR code. After the component is completed, the CNC model consistency verification is performed. If the verification fails, the component is prohibited from leaving the factory. S3: During the construction phase, the actual position point cloud of the installed components is obtained through 3D scanning and registered with the main control BIM model via ICP. The actual deviation of the installed components is calculated, and a deviation field function D(x,y,z) and a dynamic tolerance field function T(x,y,z) are constructed. The deviation of the physical model is verified using T(x,y,z) as the criterion. If the actual deviation of any installed component at its corresponding spatial position exceeds T(x,y,z), the verification fails, the corresponding installation area is locked, and a rectification notice is issued. After the verification passes, the optimal installation pose of the component to be installed is calculated based on the dynamic tolerance field, which replaces the theoretical design coordinates for installation guidance. S4: After hoisting is completed, the actual installation deviation value of each component is compared and verified with the corresponding allowable deviation range in the dynamic tolerance field; if the verification is successful, the key milestone of hoisting completion is confirmed to be effective, and the weld inspection task is automatically triggered; if the verification fails, the status is locked and a rectification notice is issued. S5: Throughout the entire lifecycle, key milestones are set as design sealing, component delivery, hoisting completion, and weld inspection qualification. The next stage of the task can only be triggered after each milestone is verified and effective, and all milestone information is written to the blockchain for evidence storage. S6: During the operation and maintenance phase, a graph neural network model G=(V,E) is constructed with steel structure components as nodes and the connection relationships between components as edges. The installation deviation recorded during the construction phase is used as one of the node features. The remaining life of each component is predicted by aggregating neighborhood information through the graph neural network. When the predicted remaining life is lower than the preset threshold, a maintenance work order is automatically generated.

2. The BIM-based digital management and control method for the entire lifecycle of steel structures according to claim 1, characterized in that, The consistency verification between the drawing and the model uses primitive feature matching and deep learning comparison methods to verify the consistency between the two-dimensional drawings and the master BIM model. The consistency verification between the model and the model uses IFC collision detection and topology relationship verification. The consistency verification between the digital model and the model is achieved by scanning the component's QR code to automatically compare the component attributes with the model attributes. The deviation verification between the physical model and the model uses three-dimensional laser scanning and ICP registration methods to calculate the deviation between the actual component and the model.

3. The BIM-based digital management and control method for the entire lifecycle of steel structures according to claim 1, characterized in that, After registering the actual location point cloud of the installed component with the master control BIM model via ICP, the deviation field function D(x,y,z) at the spatial point (x,y,z) is obtained. D(x,y,z) represents the distance deviation between the actual position and the theoretical design position of the component at that point. The dynamic tolerance field function T(x,y,z) is defined as follows: in, To standardize the basic tolerance value, The Laplace operator for the deviation field in the installed area represents the curvature of the deviation change, and α is an adaptive coefficient ranging from 0 to 0.5, dynamically adjusted according to the cumulative trend of the measured deviation in the field. This is the absolute value of the maximum curvature across the entire field, used for normalization.

4. The BIM-based digital management and control method for the entire lifecycle of steel structures according to claim 1, characterized in that, The formula for calculating the optimal installation pose is: ； in, The theoretical design pose of the component to be installed; The distance between the candidate pose and the theoretical pose; denoted as , where is the additional stress value caused by the candidate pose; λ is the stress penalty coefficient. The yield strength of the component material.

5. The BIM-based digital management and control method for the entire lifecycle of steel structures according to claim 1, characterized in that, The neighborhood information aggregation includes a single-layer propagation formula and a loss function. The single-layer propagation formula of the graph neural network model is as follows: ； in, For node i at the th The feature vector of the layer, Let be the set of neighboring nodes of node i. For neighbor node j in the th The feature vector of the layer, Let W0, W1, and W2 be the feature vectors of edge (i, j), and let W0, W1, and W2 be the learnable weight matrices.

6. The BIM-based digital management and control method for the entire lifecycle of steel structures according to claim 5, characterized in that, The loss function of the graph neural network model is: ； in, This is the mean square error loss; This is a physical constraint regularization term that penalizes the case where the predicted lifetime of node i on the directed edge (i, j) exceeds the predicted lifetime of node j and the connection type tolerance. The sum of the cases; λ is the regularization coefficient.

7. The BIM-based digital management and control method for the entire lifecycle of steel structures according to claim 1, characterized in that, In step S4, the comparison and verification adopts a method combining numerical comparison and hash evidence storage: first, the actual installation deviation value δ is determined. actual Does it not exceed the allowable deviation range ε ​​corresponding to the component in the dynamic tolerance field? max If δ actual ≤ ε max Then it is determined to be compliant, if δ actual > ε max If it is deemed non-compliant, the status will be locked; at the same time, the allowable deviation range ε ​​will be reduced. max Deviation from measured value δ actual Generate hash values ​​respectively allowed With Hash actual And write it into the blockchain.

8. The BIM-based digital management and control method for the entire lifecycle of steel structures according to claim 1, characterized in that, The automatic triggering of the next stage task in S5 is achieved through distributed automatic execution rules. All task triggering is premised on the effectiveness of the preceding milestone. The distributed automatic execution rules are configured as follows: after the design sealing milestone takes effect, the component list is automatically pushed to the production system; after the component leaving the factory milestone takes effect, the transportation task is automatically dispatched; after the hoisting completion milestone takes effect, the weld quality inspection work order is automatically generated; after the weld inspection qualified milestone takes effect, the as-built model is automatically updated.

9. A BIM-based digital management and control system for the entire lifecycle of steel structures, applied to the method described in any one of claims 1-8, characterized in that, The master control BIM model module is used to establish a master control model shared throughout the entire lifecycle, assign a globally unique identifier to each steel structure component, and store the incremental change history. The verification module is used to perform consistency checks on drawing model, digital model, model-to-model, and physical model deviation. If any verification fails, the corresponding stage of the process will be prohibited from proceeding. The drawing-free manufacturing data generation module is used to automatically generate CNC cutting codes, welding robot paths, assembly sequence animations, and component QR codes from the master control BIM model. The dynamic tolerance calculation module is used to receive 3D scan point cloud data, perform point cloud registration, tolerance field construction and optimal installation pose solution, and replace theoretical design coordinates for installation guidance. The automatic verification and triggering module is used to perform deviation comparison verification after hoisting. If the verification passes, the weld inspection task is automatically triggered. If it fails, the weld is locked and a rectification notice is issued. At the same time, key milestone information is written into the blockchain and the next stage task is automatically triggered. The graph neural network health prediction module is used to construct a graph model with steel structure components as nodes, predict the remaining life of the components with installation deviation as a feature, and automatically generate maintenance work orders when the life falls below a threshold. On-site mobile terminals are used to scan component QR codes, upload production and acceptance data, and view installation instructions.

Images