Construction design method for special-shaped structure based on AI simulation technology

Abstract

The application discloses a special-shaped structure construction design method based on an AI simulation technology, relates to the intelligent simulation technical field of building structures, and comprises the following steps: collecting geometric characteristics and physical attribute information of a target special-shaped structure, inputting a pre-trained structure behavior learning model to complete multi-physical field coupling simulation, generating an initial dynamic response data set containing structure internal stress field distribution, multi-order modal vibration characteristics and key node displacement time sequence changes, performing potential failure mode identification on the data set, locating a risk area and completing risk grading marking, starting adaptive component optimization to generate a preliminary reinforcement scheme in combination with a topological connection relationship, and iteratively evaluating construction process feasibility, and finally outputting optimized construction design drawings and process instructions meeting construction feasibility constraints. The method realizes special-shaped structure design automation by relying on the AI simulation, adapts to the complex characteristics of the special-shaped structure, and makes the design results more suitable for construction implementation requirements.

Description

Technical Field

[0001] This invention belongs to the field of intelligent simulation technology for building structures, specifically a construction design method for irregular structures based on AI simulation technology. Background Technology

[0002] In conventional irregular-shaped structure construction design, geometric parameter extraction and physical property input are largely done manually. Traditional finite element methods are used for single-physics field simulation analysis, geometric features rely solely on scattered control point acquisition, physical property parameters need to be manually broken down and entered, simulation calculations require step-by-step parameter adjustments, and structural dynamic response data needs to be calculated and extracted one by one. Potential failure modes are determined by the designer's experience, structural risk areas are marked by manual data comparison, component optimization schemes are adjusted based on manual experience, construction process feasibility is verified separately from the structural design scheme, and construction design drawings and process instructions need to be manually integrated and drawn.

[0003] Traditional design methods are prone to errors in the extraction of geometric and physical information, multiphysics coupling simulation cannot be automatically achieved, and data on internal stress fields, vibration characteristics, and nodal displacements cannot be generated synchronously, resulting in insufficient data completeness and correlation. Manual failure mode identification struggles to accurately locate areas of stress exceeding limits, concentrated vibration energy, and deformation-sensitive regions. Risk classification lacks unified standards, adaptive optimization cannot be automatically triggered, and reinforcement schemes and construction process evaluations are disconnected, making it difficult to form an integrated design outcome adapted to the characteristics of irregular structures.

[0004] To address the problem that traditional methods cannot automatically complete multi-physics coupling simulation and simultaneously generate a complete dynamic response dataset, and to address the inability to automatically achieve failure mode identification, risk classification and labeling, adaptive optimization, and construction process iterative evaluation, we will optimize and improve the construction design method for irregular structures. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a construction design method for irregular structures based on AI simulation technology, including: Collect geometric feature information and physical property information of the target irregular structure. The geometric feature information includes a set of control points on the structure surface, spatial curvature variation data and topological connection relationship. The physical property information includes material stiffness characteristics, material density distribution and connection node construction parameters. The geometric feature information and physical attribute information are input into the pre-trained structural behavior learning model, which drives the structural behavior learning model to perform multi-physics coupling simulation, generating the initial dynamic response data set of the target irregular structure under load. The initial dynamic response data set includes the internal stress field distribution, multi-mode vibration characteristics and temporal changes of key node displacements. The initial dynamic response data set is processed for potential failure mode identification to locate stress over-limit areas, vibration energy concentration areas and deformation sensitive nodes, and risk classification and labeling of the stress over-limit areas, vibration energy concentration areas and deformation sensitive nodes based on preset safety thresholds; Based on the risk classification marker and the topological connection relationship, an adaptive component optimization process is initiated to generate a preliminary structural reinforcement scheme. Based on the stress field distribution inside the structure and the stiffness characteristics of the material, the feasibility of the construction process of the preliminary structural reinforcement scheme is evaluated iteratively, and finally optimized construction design drawings and process instructions that meet the constructability constraints are generated.

[0006] Furthermore, the geometric feature information and physical attribute information are input into a pre-trained structural behavior learning model to drive the model to perform multi-physics coupled simulation, generating an initial dynamic response data set of the target irregular structure under load, including: The control point set of the structural surface and the spatial curvature change data are used to reconstruct the surface, and a parametric digital geometric model of the target irregular structure is established. The material stiffness characteristics and material density distribution are mapped to the voxel elements of the parameterized digital geometry model, and the connection node construction parameters are mapped to the node regions corresponding to the topological connection relationships, thereby generating a finite element simulation model with physical properties. The finite element simulation model is input into the structural behavior learning model, and a design load combination including gravity, wind load, seismic motion and temperature change is applied. The numerical solution kernel built into the structural behavior learning model is used to solve the equilibrium equations of the finite element simulation model under the combined design loads, and output the transient stress-strain field results and dynamic response results. The stress and strain tensors of the entire time history are extracted from the transient stress and strain field results to construct the stress field distribution inside the structure; the main vibration modes and frequencies are extracted from the dynamic response results to construct the multi-mode vibration characteristics; the maximum displacement value and variation law are extracted from the displacement time history of the nodes to construct the displacement time sequence variation of the key nodes.

[0007] Furthermore, the initial dynamic response data set is subjected to potential failure mode identification processing to locate stress over-limit regions, vibration energy concentration regions, and deformation-sensitive nodes, including: From the stress field distribution inside the structure, the principal stress value of each voxel unit is obtained, and the principal stress value is compared with the allowable stress value corresponding to the material stiffness characteristics. All voxel unit clusters whose principal stress values ​​continuously exceed the allowable stress value are marked as the stress over-limit region. From the multi-mode vibration characteristics, the strain energy density distribution of each mode is calculated, and spatial regions with strain energy density higher than the average density by a certain multiple are identified. These spatial regions are then marked as the vibration energy concentration regions. From the temporal changes of the displacement of the key nodes, analyze the displacement amplitude and the rate of change, identify nodes whose displacement amplitude exceeds the deformation threshold or whose displacement rate of change exceeds the rate threshold, and mark the nodes as the deformation-sensitive nodes. Based on the degree of principal stress exceeding the limit in the stress-over-limit region, the degree of strain energy density concentration in the vibration energy concentration region, and the degree of displacement exceeding the limit in the deformation-sensitive node, a risk level value is assigned to each stress-over-limit region, vibration energy concentration region, and deformation-sensitive node.

[0008] Furthermore, based on the risk classification marker and the topological connectivity, an adaptive component optimization process is initiated to generate a preliminary structural reinforcement scheme, including: The preliminary structural reinforcement scheme includes the spatial orientation, cross-sectional parameters and connection method of the newly added supporting components, as well as the adjustment strategy for the size of the original components; Using the aforementioned topological connectivity as constraints, a structural topology optimization design space for the target irregular structure is constructed; In the structural topology optimization design space, the optimization objectives are to maximize the overall structural stiffness and minimize the total structural mass. The risk classification marker is used as the penalty function, and the locations of the stress over-limit region, the vibration energy concentration region, and the deformation-sensitive node are set as optimization-sensitive regions. Perform topology optimization based on the variable density method to obtain the ideal distribution density cloud map of the material in the topology optimization design space of the structure; In the ideal distribution density cloud map, regions with material distribution density higher than the additive manufacturing threshold but originally designed as voids are identified, and these regions are designed as the new support components. The spatial pose, cross-sectional parameters and connection method of the new support components are determined. In the ideal distribution density cloud map, existing component areas with material distribution density below the material reduction threshold are identified, and adjustment strategies for the original component size are generated. The adjustment strategies for the original component size include size reduction or local hollowing out. By integrating all the new and adjusted design changes, a digital model of the preliminary structural reinforcement scheme is generated.

[0009] Furthermore, combining the internal stress field distribution of the structure with the material stiffness characteristics, an iterative evaluation of the construction process feasibility of the preliminary structural reinforcement scheme is conducted, including: The feasibility iterative assessment of the construction process includes component hoisting interference inspection, on-site welding / connection accessibility analysis, and temporary support structure requirement calculation. The specific steps are as follows: Based on the digital model of the preliminary structural reinforcement scheme, extract the geometric dimensions, weight, spatial coordinates and connection point information of all the newly added support components; In a three-dimensional virtual construction environment, the hoisting path of each newly added support component is simulated to check whether there is spatial interference between the newly added support component and existing structural components and temporary facilities during the hoisting process. If there is interference, the spatial pose of the newly added support component is adjusted or it is split into several sub-components that can be hoisted independently. For all locations in the preliminary structural reinforcement scheme that require on-site connection, a virtual welder / operator accessibility analysis is performed to assess whether the operating space of the connection tools is sufficient and whether the connection angle is within the allowable range of the process. If they are not accessible, the connection method is adjusted or auxiliary process holes are added. Based on the adjusted component hoisting and connection requirements, the required temporary support structure parameters are simulated and calculated, including the location, bearing capacity, and timing of removal of the temporary supports. The reinforcement scheme, after undergoing hoisting interference checks, accessibility analysis, and temporary support calculations, is re-simulated using finite element methods to verify whether its structural performance meets the requirements. If it does not, adjustments are made until all constraints are met, generating the final optimized construction design drawings and process instructions including hoisting sequence, connection process, and temporary support settings.

[0010] Furthermore, it also includes dynamic simulation of the construction process and schedule optimization steps: Based on the optimized construction design drawings and process instructions, all construction activities are decomposed, and the logical sequence, required resources and duration of each construction activity are defined. Establish a discrete event simulation model that includes constraints on construction machinery, personnel, and material supply, and input the construction activities into the discrete event simulation model; The discrete event simulation model is driven to run, simulating the complete construction process from start to finish, and recording the start time, end time, resource usage, and waiting time caused by process conflicts and insufficient resources for each construction activity during the simulation. Based on the simulation results, the key construction activity sequences and resource bottlenecks that restrict the overall construction progress were identified; With the goal of shortening the overall construction period and balancing resource load, the logical sequence and resource allocation of the key construction activities are optimized and adjusted, and simulation is performed again, iterating until an optimized construction schedule that meets the constraints of construction period and resources is generated.

[0011] Furthermore, the discrete event simulation model is driven to run, simulating the complete construction process from commencement to completion, including: Initialize the simulation clock and construction resource status, which includes the number of available machines, the number of workers, and the material inventory. From the available construction activities, select an activity to start execution according to priority rules, and occupy the corresponding construction resources; The simulation clock is advanced according to the duration of the construction activity, the construction resource status is updated, and it is checked whether any new construction activities meet the commencement conditions due to the completion of the preceding activities. When the construction activity is completed, the construction status update of the corresponding structural part in the parametric digital geometric model is recorded. The process of selecting activities and updating the construction status is executed cyclically until all construction activities are completed. The time displayed on the simulation clock is the total construction period.

[0012] Furthermore, it also includes digital twin calibration and early warning steps based on real-time on-site data: A sensor network is deployed at the construction site to collect structural and environmental monitoring data in real time during the construction process. The structural monitoring data includes strain, displacement, and vibration of key parts, while the environmental monitoring data includes wind speed, temperature, and humidity. A digital twin model is established that is associated with the parametric digital geometric model. The structural monitoring data and the environmental monitoring data drive the digital twin model in real time, so that the state of the digital twin model is synchronized with the construction state of the physical structure. Based on the current state of the digital twin model, the structural safety is calculated rapidly in real time, and the calculated real-time response value is compared with the design allowable value or the predicted value in the initial dynamic response data set. A construction safety warning is generated when the real-time response value continuously exceeds the allowable range or deviates from the predicted value by more than a threshold.

[0013] Furthermore, the structural monitoring data and environmental monitoring data are used to drive the digital twin model in real time, synchronizing the state of the digital twin model with the construction state of the physical structure, including: Establish a one-to-one mapping relationship between the sensor deployment locations and their geometric locations in the digital twin model; It receives monitoring data streams uploaded by sensors in real time and performs preprocessing such as filtering and outlier removal on the data. The preprocessed structural monitoring data is used as boundary conditions or state update parameters and input into a simplified computing kernel integrated with the digital twin model to drive the model state update. The preprocessed environmental monitoring data is converted into an equivalent load and applied to the digital twin model; The model state updated based on structural monitoring data is coupled with the load applied based on environmental monitoring data to perform calculations, so that the deformation and stress state of the digital twin model can reflect the actual state of the physical structure in real time and record the complete state evolution time history.

[0014] Furthermore, it also includes the construction and self-learning steps of the construction scheme knowledge base: The optimized construction design drawings, process instructions, dynamic simulation logs of the construction process, and construction safety early warning records are integrated into a complete project construction case. From the project construction cases, the characteristics of irregular structures, the construction difficulties encountered, the solutions adopted and their effects are extracted to form structured knowledge entries; The structured knowledge entries are stored in the construction scheme knowledge base and associated and compared with similar cases already existing in the knowledge base. When a new irregular structure construction design project is launched, similar cases are retrieved from the construction scheme knowledge base based on the geometric feature information and physical attribute information of the new project. The design points, potential risks and successful experiences in the retrieved similar cases are used as initial constraints and inspirational information and provided to the pre-trained structural behavior learning model and the construction process feasibility iterative evaluation process.

[0015] Compared with the prior art, the beneficial effects of the present invention are: The pre-trained structural behavior learning model can simultaneously receive geometric feature information such as the control point set of the structural surface, spatial curvature change data, and topological connection relationships, as well as physical property information such as material stiffness characteristics, material density distribution, and connection node construction parameters. The model directly drives multi-physics coupled simulation based on the above information autonomously, without the need for manual step-by-step adjustment of simulation parameters. It can directly generate an initial dynamic response data set containing the internal stress field distribution of the structure, multi-mode vibration characteristics, and temporal changes in the displacement of key nodes. Various dynamic response data are generated synchronously and maintain the correlation between data. Geometric and physical information do not need to be manually separated and organized. The simulation calculation and data generation links form a continuous processing flow. The connection between data acquisition input and simulation output remains stable, and the manual intervention in simulation calculation is greatly reduced.

[0016] The initial dynamic response dataset can directly perform potential failure mode identification processing, accurately locate stress over-limit areas, vibration energy concentration areas, and deformation-sensitive nodes, and complete risk classification and labeling of corresponding areas and nodes according to preset safety thresholds. The risk classification and labeling results, combined with topological connectivity, can directly initiate the adaptive component optimization process, automatically generate preliminary structural reinforcement schemes, and conduct iterative feasibility assessments of the preliminary structural reinforcement schemes based on the internal stress field distribution and material stiffness characteristics. During the assessment process, the adaptability of the schemes is adjusted, and finally, optimized construction design drawings and process instructions that meet constructability constraints are generated. Failure mode identification and risk labeling do not rely on human experience for judgment. Adaptive optimization and process feasibility assessment form a coherent processing logic. The design scheme and construction implementation constraints are mutually adapted, and the design output can directly correspond to the actual needs of construction execution. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the steps of the construction design method for irregular structures based on AI simulation technology described in this invention. Figure 2 A flowchart for potential failure mode identification and risk classification labeling; Figure 3 The curves showing resource balance and project duration variation; Figure 4 A simulation cloud map of stress distribution in a digital twin and a mapping map of sensor deployment locations; Figure 5 Comparison of the effects of optimizing construction plans. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0019] See Figure 1This invention provides a construction design method for irregular structures based on AI simulation technology. The method includes: collecting geometric feature information and physical property information of the target irregular structure; the geometric feature information includes a set of control points on the structural surface, spatial curvature variation data, and topological connection relationships; the physical property information includes material stiffness characteristics, material density distribution, and connection node construction parameters. The geometric feature information and physical property information are input into a pre-trained structural behavior learning model, driving the model to perform multi-physics coupled simulation to generate an initial dynamic response data set of the target irregular structure under load. This set includes the internal stress field distribution, multi-mode vibration characteristics, and temporal changes in displacement of key nodes. Potential failure mode identification processing is performed on the initial dynamic response data set to locate stress over-limit regions, vibration energy concentration regions, and deformation-sensitive nodes, and these regions and nodes are risk-classified and labeled based on preset safety thresholds. Based on the risk-classification labels and topological connection relationships, an adaptive component optimization process is initiated to generate a preliminary structural reinforcement scheme. By combining the internal stress field distribution and material stiffness characteristics of the structure, the feasibility of the construction process of the preliminary structural reinforcement scheme is evaluated iteratively, and finally optimized construction design drawings and process instructions that meet the constructability constraints are generated.

[0020] In one embodiment of the present invention, the control point set of the structural surface and the spatial curvature variation data are reconstructed to establish a parametric digital geometric model of the target irregular structure. Material stiffness characteristics and material density distribution are mapped to voxel elements of the parametric digital geometric model, and connection node construction parameters are mapped to node regions corresponding to topological connections, thereby generating a finite element simulation model with physical properties. This finite element simulation model is input into a structural behavior learning model, and a design load combination including gravity, wind load, seismic motion, and temperature changes is applied. The equilibrium equations of the finite element simulation model under the design load combination are solved using the numerical solution kernel built into the structural behavior learning model, outputting transient stress-strain field results and dynamic response results. The full-time stress and strain tensors are extracted from the transient stress-strain field results to construct the internal stress field distribution of the structure; the main vibration modes and frequencies are extracted from the dynamic response results to construct multi-mode vibration characteristics; and the maximum displacement values ​​and variation patterns are extracted from the displacement time histories of the nodes to construct the temporal variation of key node displacements.

[0021] See Figure 2From the internal stress field distribution of the structure, the principal stress values ​​of each voxel element are obtained. These principal stress values ​​are compared with the corresponding allowable stress values ​​in the material stiffness characteristics. Clusters of voxel elements whose principal stress values ​​consistently exceed the allowable stress values ​​are marked as stress-over-limit regions. From the multi-mode vibration characteristics, the strain energy density distribution of each mode is calculated, and spatial regions with strain energy densities higher than a specific multiple of the average density are identified and marked as vibration energy concentration regions. From the temporal changes in the displacement of key nodes, the displacement amplitude and rate of change are analyzed, and nodes whose displacement amplitude exceeds the deformation threshold or whose displacement rate of change exceeds the rate threshold are identified and marked as deformation-sensitive nodes. Based on the degree of principal stress exceeding the limit in stress-over-limit regions, the degree of strain energy density concentration in vibration energy concentration regions, and the degree of displacement exceeding the limit in deformation-sensitive nodes, a risk level value is assigned to each stress-over-limit region, vibration energy concentration region, and deformation-sensitive node.

[0022] In the specific implementation, a complex saddle-shaped hyperbolic paraboloid roof structure was used as the target irregular structure. Geometric feature information was acquired through 3D laser scanning and photogrammetry. The obtained control point set for the structural surface contained tens of thousands of points with 3D coordinates. Spatial curvature variation data was recorded as a matrix of local curvature values ​​fitted from the point cloud. Topological connections were defined in the form of a node-element connection table. Physical property information included material stiffness characteristics such as the elastic modulus and Poisson's ratio of the steel used, a constant material density distribution, and connection node construction parameters including bolt specifications, weld dimensions, and node plate thickness. In the implementation, the structural surface control point set and spatial curvature variation data were input into a NURBS surface reconstruction algorithm to establish a parametric digital geometric model of the target irregular structure. This model was composed of a series of seamlessly stitched parametric surface patches. The elastic modulus and Poisson's ratio from the material stiffness characteristics, and the density value from the material density distribution, were assigned to each voxel element generated after meshing the parametric digital geometric model. The information in the construction parameters of the connecting nodes is mapped to each node region defined in the topology connection table, thereby generating a finite element simulation model with complete physical properties.

[0023] In some embodiments, the generated finite element simulation model is input into a structural behavior learning model pre-trained using a large amount of historical engineering data. During the application of the design load combination, gravity loads are applied globally in the form of gravitational acceleration; wind loads are converted into surface pressures acting on the model surface based on the building shape and local wind pressure distribution; seismic loads act on the structural base in the form of acceleration time histories; and temperature change loads are reflected as uniform temperature field changes across the model. The numerical solution kernel built into the structural behavior learning model calls an implicit dynamic solver to solve the equilibrium equations of the finite element simulation model under the design load combination, outputting transient stress-strain field results and dynamic response results. From the full-time history data of the transient stress-strain field results, the stress tensor and strain tensor at all integration points at different times are extracted to construct the internal stress field distribution of the structure, which is a function of spatial location and time. From the dynamic response results, the shapes and corresponding natural frequencies of the first ten main vibration modes are extracted through modal analysis to construct multi-mode vibration characteristics. From the displacement time history data of key nodes in the finite element model, the maximum displacement, root mean square value, and variation pattern are extracted to construct the displacement time series variation of key nodes.

[0024] In some embodiments, the process of identifying potential failure modes and locating stress-over-limit regions by performing potential failure mode identification processing on the initial dynamic response dataset involves obtaining the three principal stress values ​​of each voxel element throughout the entire time history from the stress field distribution inside the structure, comparing each principal stress value with the allowable stress value in the corresponding direction defined in the material stiffness characteristics point by point, identifying voxel elements whose principal stress values ​​continuously exceed the allowable stress value for more than 90% of the time history, and marking these spatially connected voxel element clusters as stress-over-limit regions. The location of vibration energy concentration regions starts from the multi-mode vibration characteristics, calculating the strain energy density of each element in the model under each mode, averaging the strain energy density of all elements, and identifying spatial regions with strain energy densities more than three times the average density, which are then marked as vibration energy concentration regions. The location of deformation-sensitive nodes involves analyzing the temporal changes in the displacement of key nodes, identifying nodes whose displacement amplitude exceeds the deformation threshold or whose displacement change rate exceeds the rate threshold; for example, if the maximum displacement of a node reaches 120% of the deformation threshold, it is marked as a deformation-sensitive node. Based on the above identification results, the degree of principal stress exceeding the limit in the stress-over-limit region, the degree of strain energy density concentration in the vibration energy concentration region, and the degree of displacement exceeding the limit at deformation-sensitive nodes are each quantified into a risk level value between one and ten. The more principal stress values ​​exceed the limit or the more significant the strain energy concentration, the higher the risk level value is assigned. It can be understood that the quantification of the stress-over-limit region can be calculated based on the stress exceedance coefficient, which... Defined as the normalized result of the integral of the ratio of the portion exceeding the allowable stress in the principal stress time history to the allowable stress over the total time, the expression is:

[0025] in: This indicates the total time the load is applied. Indicates the allowable stress of the material. This represents the actual principal stress value of the element at time t. This coefficient can be used to assist in assigning risk level values. Optionally, when calculating the strain energy density concentration in areas of concentrated vibration energy, the ratio of the strain energy density of each element to the global average strain energy density can be used as an intermediate indicator. The identification of deformation-sensitive nodes relies not only on single-point thresholds of displacement amplitude but also on analysis combining the spectral characteristics of displacement time history. The displacement change rate threshold can be set according to a proportion of the structure's natural vibration period. In practice, all risk level values ​​are attached to the corresponding finite element model elements or nodes, forming a label file with spatial location and risk level, which guides the subsequent optimization process.

[0026] In one embodiment of the present invention, the preliminary structural reinforcement scheme includes the spatial orientation, cross-sectional parameters, and connection methods of newly added supporting components, as well as an adjustment strategy for the dimensions of existing components. A structural topology optimization design space for the target irregular structure is constructed using topological connections as constraints. In this design space, the optimization objectives are to maximize the overall structural stiffness and minimize the total structural mass. Risk classification markers are used as penalty functions, and the locations of stress-over-limit areas, vibration energy concentration areas, and deformation-sensitive nodes are designated as optimization-sensitive areas. A topology optimization solution based on the variable density method is performed to obtain an ideal material distribution density cloud map in the structural topology optimization design space. In the ideal density cloud map, areas with material distribution density higher than the additive threshold but originally designed as voids are identified and designed as newly added supporting components. The spatial orientation, cross-sectional parameters, and connection methods of these newly added supporting components are determined. In the ideal density cloud map, existing component areas with material distribution density lower than the subtractive threshold are identified, and an adjustment strategy for the dimensions of the existing components is generated. This adjustment strategy includes size reduction or partial hollowing. All newly added and adjusted design changes are integrated to generate a digital model of the preliminary structural reinforcement scheme.

[0027] The feasibility iterative assessment of the construction process includes component hoisting interference checks, on-site welding or connection accessibility analysis, and temporary support structure requirement calculations. Based on the digital model of the preliminary structural reinforcement scheme, the geometric dimensions, weight, spatial coordinates, and connection point information of all newly added support components are extracted. In a 3D virtual construction environment, the hoisting path of each newly added support component is simulated to check for spatial interference between the newly added support component and existing structural components and temporary facilities during hoisting. If interference exists, the spatial orientation of the newly added support component is adjusted or it is broken down into several independently hoistable sub-components. For all locations in the preliminary structural reinforcement scheme that require on-site connections, a virtual welder or operator accessibility analysis is performed to assess whether the operating space for connection tools is sufficient and whether the connection angle is within the allowable range of the process. If inaccessible, the connection method is adjusted or auxiliary process holes are added. Based on the adjusted component hoisting and connection operation requirements, the required temporary support structure parameters are simulated and calculated, including the location, bearing capacity, and removal timing of the temporary supports. The reinforcement scheme, after undergoing hoisting interference checks, accessibility analysis, and temporary support calculations, is re-simulated using finite element methods to verify whether its structural performance meets the requirements. If it does not, adjustments are made until all constraints are met, generating the final optimized construction design drawings and process instructions including hoisting sequence, connection process, and temporary support settings.

[0028] In practice, an adaptive component optimization process is initiated based on a finite element simulation model of an irregular structure that has undergone potential failure mode identification and risk classification. The generation of a preliminary structural reinforcement scheme begins with defining its components, including the spatial orientation, cross-sectional parameters, and connection methods of newly added support components, as well as adjustment strategies for the dimensions of existing components. The construction of the structural topology optimization design space strictly adheres to the constraints of topological connections, which define the connection possibilities and force transmission paths between structural components. The optimization design space is defined as the entire three-dimensional envelope containing all components of the original structure and areas where new components are permitted. The mathematical expression of the optimization objective is to maximize the overall structural stiffness while minimizing the total structural mass. Risk classification markers are integrated into a penalty function, which applies additional penalty weights to optimization-sensitive regions such as stress over-limit areas, vibration energy concentration areas, and deformation-sensitive nodes, forcing the optimization algorithm to tend to retain or increase material in these areas. The topology optimization solution based on the variable density method is performed, where the material density of each design element is treated as a design variable that varies continuously between zero and one. Through iterative calculation using the optimization algorithm, the ideal distribution density cloud map of the material in the structural topology optimization design space is finally output. The ideal distribution density cloud map displays the relative material density at each element location in grayscale form.

[0029] In some embodiments, post-processing of the ideal density distribution cloud map involves the application of additive and subtractive thresholds. The additive threshold is set to 0.3, identifying regions in the ideal density distribution cloud map where the material distribution density is higher than the additive threshold but the original design was void. These regions are designed as new support members. The spatial pose of the new support members is determined by the geometric center and principal axis of inertia of the high-density region. The cross-sectional parameters of the new support members are derived from the region's volume and density integral. The connection method of the new support members is designed based on their topological connection with surrounding existing members. The subtractive threshold is set to 0.1, identifying existing member regions in the ideal density distribution cloud map where the material distribution density is lower than the subtractive threshold. An adjustment strategy for the original member dimensions is generated, which may include proportionally reducing the cross-sectional dimensions of the corresponding region or locally hollowing out non-critical stress paths. The designs of all new support members and the adjustment strategies for existing members are integrated to generate a preliminary 3D model of the structural reinforcement scheme in a computer-aided design system.

[0030] In practice, the feasibility of the construction process for the preliminary structural reinforcement scheme is iteratively evaluated. Interference checks during component hoisting are conducted based on the digital 3D model of the preliminary structural reinforcement scheme. The geometric dimensions, weight, spatial coordinates, and connection point information of all newly added supporting components are extracted from the digital 3D model and imported into a 3D virtual construction environment. In the 3D virtual construction environment, hoisting path simulation is performed for each newly added supporting component, simulating the entire process of a crane lifting from the component stacking point and moving along a preset path to the installation point. This checks for spatial interference between the newly added supporting components and existing structural parts and temporary facilities. If interference is detected, the spatial orientation of the newly added supporting component is adjusted or it is broken down into several independently hoistable sub-components. On-site welding or connection accessibility analysis is performed for all locations in the preliminary structural reinforcement scheme that require on-site connections. Virtual welder or operator models and equipment models are placed in the 3D virtual construction environment to assess whether the operating space of the connection tools is sufficient and whether the connection angle is within the allowable range of the process. If the analysis determines that a certain connection location is inaccessible, the connection method is adjusted or auxiliary process holes are added at locations that do not affect structural performance. It is understandable that the calculation of temporary support structure requirements is part of the iterative assessment of construction process feasibility. Based on the adjusted component hoisting and connection operation requirements, the required temporary support structure parameters are simulated and calculated. These parameters include the location, bearing capacity, and removal timing of the temporary supports. The location of the temporary supports is determined based on the cantilever state of the component during hoisting and the structural stability during connection operations. The bearing capacity of the temporary supports is calculated through analysis of the construction stage loads. The removal timing of the temporary supports is related to the time point when the permanent structure achieves overall stiffness. The reinforcement scheme, after hoisting interference checks, accessibility analysis, and temporary support calculations, is re-imported into finite element analysis software for simulation to verify whether its structural performance meets the design requirements. If it does not meet the requirements, the process is adjusted back to the digital 3D model of the initial structural reinforcement scheme, and the process assessment and structural verification are repeated. This iterative cycle continues until all structural performance and construction process constraints are met. Finally, optimized construction design drawings and process instructions including hoisting sequence, connection process, and temporary support settings are output.

[0031] In some embodiments, the construction of the objective function in the structural topology optimization process needs to ensure dimensionality consistency. A method for constructing a dimensionless objective function... as follows:

[0032] in: This indicates the overall flexibility of the structure under the current design. This indicates the overall structural flexibility of the initial design. This indicates the total mass of the structure under the current design. This represents the total structural mass in the initial design. This represents the penalty item based on risk classification markers, and its calculation formula is as follows: ,in It is a design variable vector, representing the relative density of the material in each unit. It is a unit density, This represents the set of all cells marked as optimization-sensitive regions. It is a unit The weighting coefficients that map the corresponding risk level values. It is a punishment factor. , , This is a dimensionless weighting coefficient used to adjust the importance of various objectives. The first term in the formula... For normalized softness, the second term For the normalized quality, the third term The penalty value is dimensionless, therefore the entire objective function... It is a comprehensive evaluation value. The smaller the value, the better the overall performance of the design scheme in terms of stiffness, lightweighting and risk control.

[0033] In one embodiment of the present invention, all construction activities are decomposed based on the optimized construction design drawings and process instructions, and the logical sequence, required resources, and duration of each construction activity are defined. A discrete event simulation model including constraints on construction machinery, personnel, and material supply is established, and the construction activities are input into this discrete event simulation model. The discrete event simulation model is driven to run, simulating the complete construction process from commencement to completion, and recording the start time, end time, resource occupancy, and waiting time caused by process conflicts and resource shortages for each construction activity during the simulation. Based on the simulation results, the key construction activity sequences and resource bottlenecks that restrict the overall construction progress are identified. With the goal of shortening the total construction period and balancing resource load, the logical sequence and resource allocation of the key construction activity sequences are optimized and adjusted, and the simulation is re-run, iterating until an optimized construction schedule plan that meets the time and resource constraints is generated. The simulation clock and construction resource status are initialized, and the construction resource status includes the number of available machinery, the number of workers, and the material inventory. From the construction activities that can be started, activities are selected to start execution according to priority rules, occupying the corresponding construction resources. The simulation clock advances based on the duration of each construction activity, updates the status of construction resources, and checks if any new construction activities meet the commencement conditions due to the completion of preceding activities. It records the construction status update of the corresponding structural part in the parametric digital geometry model when a construction activity is completed. This process of selecting activities and updating construction status is repeated cyclically until all construction activities are completed; the time displayed on the simulation clock is the total project duration.

[0034] In practical implementation, dynamic simulation and schedule optimization of the construction process are carried out based on the generated optimized construction design drawings and process instructions. According to the optimized construction design drawings and process instructions, all construction activities are decomposed, including component processing, component transportation, on-site hoisting, node connection, temporary support erection and dismantling, measurement and calibration, etc. The logical sequence, required resources, and duration of each construction activity are defined. For example, the construction activity "A-3 axis main beam hoisting" can only begin after the construction activity "A-3 axis temporary support erection" is completed. Its required resources include one 200-ton crawler crane, two crane operators, and four installers, with an estimated duration of four hours. A discrete event simulation model is established, incorporating constraints on construction machinery, personnel, and material supply. The quantity, work efficiency, and collaborative relationships between different types of resources are defined in the discrete event simulation model. The defined construction activities, their logic, resources, and duration information are input into the discrete event simulation model.

[0035] In some embodiments, a discrete event simulation model is driven to simulate the complete construction process from commencement to completion. The simulation clock advances in minutes, recording the start and end times, resource usage, and waiting times due to process conflicts and resource shortages for each construction activity during the simulation. For example, the simulation records show that the construction activity "Installation of Curved Panels in Area B" was delayed by 120 minutes due to waiting for a dedicated installation fixture, and the construction activity "Central Mast Hoisting" was delayed due to competition with another hoisting activity for the same large crane resource. Based on the simulation results, key construction activity sequences and resource bottlenecks that constrain the overall construction progress are identified. Key construction activity sequences are continuous activity chains with zero total float, and resource bottlenecks are resource types whose utilization rate consistently exceeds 85% during the simulation. With the goal of shortening the total construction period and balancing resource load, the logical order and resource allocation of key construction activity sequences are optimized and adjusted. For example, resources from activities on certain non-critical paths are reallocated to key activities, or some sequential operations are changed to parallel operations. The simulation is then repeated iteratively until an optimized construction schedule that meets the time and resource constraints is generated.

[0036] In practical implementation, a core loop of the discrete event simulation model includes initialization, event handling, and state updating. Initialization involves setting the simulation clock and construction resource status, which includes the number of available machines, the number of workers, and material inventory. For example, initialization might include two large cranes, fifteen welders, and 500 tons of steel of a specific specification. From the available construction activities, activities are selected to start execution according to priority rules, such as earliest start time priority, occupying corresponding construction resources. The simulation clock advances based on the duration of the construction activity, updating the construction resource status. This update includes releasing resources occupied by completed activities and checking if any new construction activities meet the start conditions due to the completion of preceding activities. The construction status update of the corresponding structural part in the parametric digital geometry model is recorded when a construction activity is completed. For example, when the "Northeast Corner Truss Unit Hoisting" activity is completed, the status of the truss unit in the parametric digital geometry model is updated from "Not Under Construction" to "Installed." This process of selecting activities and updating construction status is repeated until all construction activities are completed. The time of the simulation clock is the total project duration (see Table 1).

[0037] Table 1: Breakdown of Construction Activities

[0038] It is understandable that identifying resource bottlenecks and achieving resource balancing are key to the optimization process. After a simulation run, the utilization rate changes of various resources throughout the entire project duration can be analyzed. Resource load imbalance index. It can be used to quantify a certain type of resource The usage fluctuations during the simulation period are calculated as follows:

[0039] in: Representing resources The load imbalance index This indicates the number of time intervals into which the total project duration is equally divided. Representing resources In the Utilization rate within a time interval Representing resources Average utilization rate throughout the entire simulation cycle. One of the optimization objectives is to reduce the utilization rate of critical resources. The value is optimized to ensure smoother operation. Optionally, a scheduling optimization method based on a genetic algorithm can be used when optimizing the logical sequence of key construction activities. Chromosome encoding represents the order and start time of activities, and the fitness function considers both the total project duration and the resource load imbalance index. A discrete event simulation model serves as the evaluation function, simulating the scheduling scheme obtained after decoding each generation of chromosomes, calculating its total project duration and resource load imbalance index, thereby guiding the search towards a scheme with shorter project duration and more balanced resource usage. The termination condition for iterative simulation can be reaching a preset maximum number of iterations, or the optimal solution no longer improving after several consecutive generations. The final output optimized construction schedule includes precise, conflict-free start and end times for all construction activities, as well as a detailed resource usage plan.

[0040] See Figure 3 In the iterative optimization process of construction schedule based on genetic algorithm, the resource load imbalance index and the total construction period show a coordinated convergence optimization trend, which intuitively verifies the effectiveness of the multi-objective construction scheduling optimization method. Specifically, the curve is plotted with the number of genetic algorithm optimization iterations on the horizontal axis, the left vertical axis representing the resource load imbalance index (used to quantify the fluctuation of the use of key construction resources throughout the entire construction period; the lower the index, the more balanced the resource allocation and the less conflict between idle and overload), and the right vertical axis representing the total construction period (unit: hours, reflecting the overall construction efficiency of the project). From the evolution of the curve, as the number of iterations increased from 1 to 10, the resource load imbalance index decreased from approximately 4.2 to 1.0, a reduction of 76.2%. This indicates that through the iterative optimization of the logical sequence of construction activities and resource allocation using the genetic algorithm, the fluctuations in the use of key resources (such as large hoisting machinery and specialized personnel) were continuously smoothed out, and the resource bottleneck problem was systematically solved. At the same time, the total construction period decreased from approximately 168 hours to 128 hours, a reduction of 23.8%. This achieved the synergistic optimization of the dual objectives of "shortening the total construction period" and "balancing resource load," rather than a sacrificial optimization of a single objective. Analysis of convergence characteristics shows that both indicators exhibit a rapid downward trend in the early stages of iteration (1-6 iterations), corresponding to the global optimization phase of the genetic algorithm for the initial scheduling scheme, where deteriorated scheduling schemes are quickly eliminated through chromosome crossover and mutation operations. In the later stages of iteration (7-10 iterations), the curves tend to flatten, and the resource load imbalance index and the total construction period gradually converge to stable values, corresponding to the local fine-tuning phase of the algorithm. Finally, the optimal convergence state is reached in the 10th iteration, generating the optimal construction schedule that meets the requirements of schedule constraints and resource balance. Through the coupled iteration of genetic algorithm and discrete event simulation, multi-objective optimization of construction schedule is achieved, which not only effectively shortens the total construction period of irregular structure construction but also significantly improves the utilization efficiency of construction resources, providing quantitative optimization basis and visual verification for schedule control and resource allocation in complex irregular structure construction.

[0041] In one embodiment of the invention, a sensor network is deployed at the construction site to collect structural and environmental monitoring data in real time during the construction process. The structural monitoring data includes strain, displacement, and vibration of key components, while the environmental monitoring data includes wind speed, temperature, and humidity. A digital twin model is established and associated with a parametric digital geometric model. The structural and environmental monitoring data drive the digital twin model in real time, synchronizing the state of the digital twin model with the construction state of the physical structure. Based on the current state of the digital twin model, rapid structural safety calculations are performed in real time, and the calculated real-time response values ​​are compared with the design allowable values ​​or predicted values ​​in the initial dynamic response data set. When the real-time response value continuously exceeds the allowable range or the deviation from the predicted value exceeds a threshold, a construction safety warning is generated. A one-to-one mapping relationship is established between the sensor deployment locations and their geometric locations in the digital twin model. The monitoring data stream uploaded by the sensors is received in real time, and the data undergoes preprocessing, including filtering and outlier removal. The preprocessed structural monitoring data is used as boundary conditions or state update parameters and input into a simplified calculation kernel integrated with the digital twin model to drive model state updates. The preprocessed environmental monitoring data is transformed into equivalent loads and applied to the digital twin model. The model state updated based on the structural monitoring data is coupled with the loads applied based on the environmental monitoring data for calculation, so that the deformation and stress state of the digital twin model reflects the actual state of the physical structure in real time, and records the complete state evolution time history.

[0042] In practical implementation, the irregular structure under construction is used as the physical entity. Digital twin calibration and early warning steps based on real-time on-site data are carried out. A sensor network is deployed at the construction site, including vibrating wire strain sensors, total station prisms, accelerometers, anemometers, and temperature and humidity sensors. This network collects structural and environmental monitoring data in real time during construction. Structural monitoring data includes strain, displacement, and vibration at key locations, while environmental monitoring data includes wind speed, temperature, and humidity. A digital twin model is established, linked to the parameterized digital geometric model created before construction. This digital twin model has the same geometric topology and material property definitions but possesses the ability to receive real-time data and perform rapid calculations. The structural and environmental monitoring data drive the digital twin model in real time, synchronizing its state with the physical structure's construction status. Based on the current state of the digital twin model, rapid real-time structural safety calculations are performed using a simplified finite element solver integrated into the digital twin platform. The calculated real-time response values ​​are compared with the design allowable values ​​or predicted values ​​from the initial dynamic response data set. When the real-time response value continuously exceeds the allowable range or the deviation from the predicted value exceeds the threshold, a construction safety warning is generated. The warning information is issued through the monitoring platform interface, SMS, or audible and visual alarms, as shown in Table 2.

[0043] Table 2: Sensor Network Configuration Table

[0044] In some embodiments, sensor monitoring data and the digital twin model are driven and synchronized in real time. A one-to-one mapping relationship is established between the sensor deployment locations and their geometric locations in the digital twin model. This mapping is achieved by associating the physical installation coordinates of the sensors with the coordinates of the nearest node in the digital twin model. The monitoring data stream uploaded by the sensors through the IoT gateway is received in real time. The data undergoes preprocessing, including filtering and outlier removal. The filtering uses a moving average filtering algorithm, and outlier removal is based on the Laida criterion. The preprocessed structural monitoring data is used as boundary conditions or state update parameters and input into a simplified computational kernel integrated with the digital twin model to drive model state updates. For example, the measured displacement value of a node is applied as a forced displacement boundary condition to the corresponding node in the digital twin model. The preprocessed environmental monitoring data is converted into equivalent loads and applied to the digital twin model. For example, real-time wind speed values ​​are converted into distributed wind pressure acting on the model surface using a wind pressure formula. The model state updated based on structural monitoring data is coupled with the load applied based on environmental monitoring data. The coupled calculation is completed within a shortened simulation time step, so that the deformation and stress state of the digital twin model can reflect the actual state of the physical structure in real time and record the complete state evolution time history.

[0045] In practical implementation, rapid structural safety calculation and early warning determination are based on the real-time state of the digital twin model. The core of the calculation is a simplified solver integrated into the digital twin model. This solver uses the same constitutive relations as the initial high-fidelity simulation model but with a reduced mesh size to achieve rapid calculation. Real-time response values ​​include the real-time equivalent stress of key elements, the real-time displacement of key nodes, and the real-time fundamental frequency of the structure calculated by the digital twin model. Design allowable values ​​are obtained from structural design codes, and the predicted values ​​in the initial dynamic response dataset refer to the response values ​​at the corresponding locations and under the same design load combination, obtained from pre-construction simulation. The comparison process is continuous. When the real-time response value continuously exceeds the allowable range or the deviation from the predicted value exceeds a threshold, a construction safety early warning is generated. The criterion for continuous exceedance is that the percentage of times the value exceeds the allowable range or threshold exceeds a set proportion within a set time window. The early warning generation logic can be implemented using a deviation index. To trigger formally, the deviation index Defined as:

[0046] in: This represents the overall deviation index at time t. This indicates the number of key response parameters being monitored. Indicates the first The real-time response values ​​of each parameter at time t. Indicates the first The predicted values ​​of each parameter under the design load. Indicates the first The design allowable values ​​for each parameter. When Exceeding the preset warning threshold An early warning will be generated immediately.

[0047] It is understandable that the state synchronization of a digital twin model is a continuous process. A simplified computational kernel employs a fast algorithm based on matrix condensation or modal superposition to ensure that model state updates and calculations are completed within second-level or even sub-second time intervals. The conversion of wind speed in environmental monitoring data is based on the wind pressure calculation formula in building structure load codes, converting wind speed values ​​into wind pressure loads acting on the corresponding surfaces of the digital twin model. The impact of temperature data may be reflected by assigning a temperature field to the digital twin model corresponding to the measured temperature distribution, and then calculating temperature stress. Optionally, the issuance of early warning information can be tiered, with the deviation index... Exceeding the warning threshold But below the higher alarm threshold When the deviation index is high, a yellow alert is issued to remind on-site personnel to pay attention to monitoring. Exceeding alarm threshold In such cases, a red alert is issued, which may be linked to the on-site broadcast system and the mobile terminals of engineering management personnel. It is recommended to suspend high-risk operations and conduct an inspection. All real-time monitoring data, digital twin model calculation status, and early warning records are stored in a time-series database for subsequent trend analysis and case review.

[0048] See Figure 4This paper presents the spatial mapping relationship between the stress field distribution characteristics of the entire structure based on the multiphysics coupling simulation results and the precise placement of sensors. From the perspective of stress field distribution, the color scale on the right side of the figure, in MPa, clearly quantifies the stress value range of the structural surface: the blue area corresponds to the low stress area (0-60MPa), mainly distributed in the X coordinate 95-105, Y coordinate 25-40 and X coordinate 110-125, Y coordinate 25-35 ranges, showing a low stress characteristic with a green gradient; the yellow to red area is the high stress concentration area (80-160MPa), concentrated in the core area of ​​X coordinate 110-125, Y coordinate 40-50, among which the stress value in the red area reaches 160MPa, which is the peak zone of the structural stress response, intuitively reflecting the mechanical bearing state and deformation potential of this area under load. From the perspective of sensor spatial mapping, the two types of sensor identifiers in the figure form a precise spatial correspondence with the stress field distribution: the ST-01 strain sensor, marked with a red pentagram, is deployed in the high-stress concentration core area at X-coordinate 120.5 and Y-coordinate 45.2, which corresponds to the high-stress gradient region of 140-160 MPa in the cloud map, meeting the core requirement of strain sensors for micro-strain monitoring in high-stress areas; the DS-05 displacement sensor, marked with a blue square, is deployed in the low-stress transition area at X-coordinate 98.7 and Y-coordinate 33.4, corresponding to the stress range of 40-60 MPa in the cloud map, adapting to the displacement monitoring requirement for locating structural deformation-sensitive nodes. The cloud map, through color-layered visualization, concretizes the continuous distribution pattern of the transient stress field inside the structure, and simultaneously achieves a one-to-one mapping between the physical deployment locations of sensors and the geometric coordinates of the digital twin model. This provides a visualized mechanical state basis for subsequent digital twin model calibration based on real-time monitoring data, risk identification of stress-exceeding areas, and feasibility assessment of construction processes.

[0049] In one embodiment of the present invention, optimized construction design drawings, process instructions, dynamic simulation logs of the construction process, and construction safety early warning records are integrated into a complete project construction case. From the project construction case, the characteristics of the irregular structure, the construction difficulties encountered, the solutions adopted, and their effects are extracted to form structured knowledge entries. These structured knowledge entries are stored in a construction scheme knowledge base and associated and compared with similar cases already existing in the knowledge base. When a new irregular structure construction design project is initiated, similar cases are retrieved from the construction scheme knowledge base based on the geometric feature information and physical attribute information of the new project. The design points, potential risks, and successful experiences from the retrieved similar cases are used as initial constraints and heuristic information, provided to the pre-trained structural behavior learning model and the iterative evaluation process of construction process feasibility.

[0050] In practice, the construction scheme knowledge base building and self-learning steps are initiated after the completion of an irregular structure project. The optimized construction design drawings, process instructions, dynamic simulation logs of the construction process, and construction safety early warning records of this project are integrated into a complete project construction case with a unified project identifier through the project data integration interface. From the project construction case, the characteristics of the irregular structure, the construction difficulties encountered, the solutions adopted, and their effects are extracted to form structured knowledge entries. The characteristics of the irregular structure include the main shape classification, maximum span, main curvature range, and height-to-width ratio. The construction difficulties encountered may include "interference between the hoisting path and temporary support in large cantilever sections" and "insufficient accessibility for on-site welding of complex nodes." The corresponding solutions are "disassembling the overall component into three sub-units and adjusting the hoisting points" and "changing the welded connection to a high-strength bolt connection and adding process holes." The effects are recorded as "elimination of hoisting interference, installation accuracy meeting standards" and "improved connection operation efficiency, and qualified quality inspection."

[0051] In practice, structured knowledge entries are stored in a construction scheme knowledge base, which combines relational and graph databases. The relational database stores the attribute information of the entries, while the graph database stores the relationships between entries, such as "similar shape," "using the same materials," and "having similar challenges." During storage, new structured knowledge entries are associated and compared with existing similar cases in the knowledge base. The association is based on the similarity of structural feature attributes, and the comparison includes the type of construction challenge, differences in solutions, and evaluation of the final effect. When a new irregular structure construction design project is initiated, similar cases are retrieved from the construction scheme knowledge base based on the geometric feature information and physical attribute information of the new project. The retrieval process first converts the features of the new project into feature vectors isomorphic to the entries in the knowledge base.

[0052] In some embodiments, similar case retrieval is achieved by calculating the similarity between feature vectors. Feature vectors of new projects With the knowledge base Feature vectors of historical cases similarity between It can be calculated using an improved cosine similarity method, and the formula is as follows:

[0053] in: This represents the total dimension of the feature vector, with each dimension corresponding to a structural feature attribute. The feature vector of the new project is represented at the th... Dimension value, Indicates the first The feature vector of the historical case in the th Dimension value, It is the first The weight coefficients of the feature attributes reflect the importance of different feature attributes in the matching process. For example, "subject morphology classification" and "maximum span" may be assigned higher weights. After calculation, the similarity is... Historical cases that exceed a preset threshold are output as similar cases.

[0054] It is understandable that the construction of feature vectors needs to cover key dimensions of geometric feature information and physical attribute information. The values ​​of each dimension of the vector may be continuous numerical values, discrete classification codes, or normalized values. The design points, potential risks, and successful experiences in similar cases retrieved are extracted and formatted into initial constraints and heuristic information. Design points may include "For this type of single-layer reticulated shell, it is recommended to focus on wind-induced vibration in the initial analysis," potential risks may include "Stress concentration at the corners of this type of bending and torsional member is prone to occur under temperature gradient loads," and successful experiences may include "The construction sequence of segmented hoisting and aerial patching can effectively control deformation." This information is provided to the pre-trained structural behavior learning model and the construction technology feasibility iterative evaluation process as initial input or background knowledge for analysis and evaluation. For example, when the structural behavior learning model is simulating, high-risk load conditions marked in similar cases can be loaded first.

[0055] In some embodiments, the construction scheme knowledge base has a self-learning and updating mechanism. When a new project is completed and a construction case is generated, the case is added to the knowledge base. If the solution of the new case significantly improves the effectiveness compared to the solutions of similar cases already in the knowledge base—for example, by more than 10% in shortening the construction period or more than 15% in reducing costs—the solution of the new case may be marked as a recommended solution and an "optimized alternative" association is established with relevant old cases for priority recommendation during subsequent searches. Optionally, the similarity calculation can combine features at different levels, dividing the feature vector into multiple sub-vectors such as geometric features, material features, load features, and construction environment features, calculating the similarity of each sub-vector separately, and then performing a weighted summation. The knowledge base can also record the feedback results of each search, i.e., which suggestions from similar cases the user ultimately adopted and the actual effects after adoption. This feedback information is used to dynamically adjust the weight coefficients of feature attributes. This allows the knowledge base's matching and recommendation capabilities to continuously improve as the number of use cases increases.

[0056] See Figure 5The figure shows a box plot comparing the optimization effects of construction schemes. It is used to evaluate the application effectiveness of AI simulation-based construction design methods for irregular structures in the knowledge base case study phase. The core comparison dimensions are two key performance indicators: schedule optimization rate and cost reduction rate. The vertical axis represents the optimization percentage (%), and the horizontal axis represents the optimization type. The boxes in the box plot represent the interquartile range (IQR), i.e., the interval from the 25th percentile (Q1) to the 75th percentile (Q3). The horizontal line inside the box represents the median, reflecting the central tendency of the data. The upper and lower whiskers correspond to the maximum and minimum values ​​of the data (in scenarios without outliers), completely covering the full distribution range of the sample. The median of the schedule optimization rate is approximately 10.3%, with a box range of approximately 8.9% to 12.2%, and the upper and lower whiskers cover the full range of 5.3% to 15.7%. The results show that in the knowledge base case library, the median improvement in construction schedule achieved by the optimized methods exceeds 10%, with 75% of the cases achieving a schedule reduction of over 8.9%, and the highest improvement reaching 15.7%, validating the universality and effectiveness of the method in schedule control. The median cost reduction rate is approximately 7.5%, with a range of 5.7% to 9.7%, and the upper and lower bounds covering the entire range of 3.2% to 12.4%. This indicates that the solutions also possess stable optimization capabilities in cost control, with a median cost reduction of 7.5%, 75% of the cases achieving cost savings of over 5.7%, and the highest reduction exceeding 12%, achieving the dual optimization goals of schedule and cost. From the distribution characteristics, the median of the schedule optimization rate and the overall position of the box are significantly higher than the cost reduction rate, indicating that the method has a more prominent effect on improving the schedule. At the same time, there are no outliers in the boxes of both types of indicators, indicating that the optimization effect has stability and consistency in different irregular structure projects. This verifies the positive empowering effect of the self-learning mechanism of the construction scheme knowledge base on the scheme effect. That is, through the reuse of experience from similar cases, the construction scheme has achieved systematic optimization in terms of schedule and cost.

[0057] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A construction design method for irregular structures based on AI simulation technology, characterized in that, include: Collect geometric feature information and physical property information of the target irregular structure. The geometric feature information includes a set of control points on the structure surface, spatial curvature variation data and topological connection relationship. The physical property information includes material stiffness characteristics, material density distribution and connection node construction parameters. The geometric feature information and physical attribute information are input into the pre-trained structural behavior learning model, which drives the structural behavior learning model to perform multi-physics coupling simulation, generating the initial dynamic response data set of the target irregular structure under load. The initial dynamic response data set includes the internal stress field distribution, multi-mode vibration characteristics and temporal changes of key node displacements. The initial dynamic response data set is processed for potential failure mode identification to locate stress over-limit areas, vibration energy concentration areas and deformation sensitive nodes, and risk classification and labeling of the stress over-limit areas, vibration energy concentration areas and deformation sensitive nodes based on preset safety thresholds; Based on the risk classification marker and the topological connection relationship, an adaptive component optimization process is initiated to generate a preliminary structural reinforcement scheme. Based on the stress field distribution inside the structure and the stiffness characteristics of the material, the feasibility of the construction process of the preliminary structural reinforcement scheme is evaluated iteratively, and finally optimized construction design drawings and process instructions that meet the constructability constraints are generated.

2. The construction design method for irregular structures based on AI simulation technology according to claim 1, characterized in that, The geometric feature information and physical attribute information are input into a pre-trained structural behavior learning model to drive the model to perform multi-physics coupled simulation, generating an initial dynamic response data set of the target irregular structure under load, including: The control point set of the structural surface and the spatial curvature change data are used to reconstruct the surface, and a parametric digital geometric model of the target irregular structure is established. The material stiffness characteristics and material density distribution are mapped to the voxel elements of the parameterized digital geometry model, and the connection node construction parameters are mapped to the node regions corresponding to the topological connection relationships, thereby generating a finite element simulation model with physical properties. The finite element simulation model is input into the structural behavior learning model, and a design load combination including gravity, wind load, seismic motion and temperature change is applied. The numerical solution kernel built into the structural behavior learning model is used to solve the equilibrium equations of the finite element simulation model under the combined design loads, and output the transient stress-strain field results and dynamic response results. The stress and strain tensors of the entire time history are extracted from the transient stress and strain field results to construct the stress field distribution inside the structure; the main vibration modes and frequencies are extracted from the dynamic response results to construct the multi-mode vibration characteristics; the maximum displacement value and variation law are extracted from the displacement time history of the nodes to construct the displacement time sequence variation of the key nodes.

3. The construction design method for irregular structures based on AI simulation technology according to claim 2, characterized in that, The initial dynamic response data set is processed for potential failure mode identification to locate stress over-limit regions, vibration energy concentration regions, and deformation-sensitive nodes, including: From the stress field distribution inside the structure, the principal stress value of each voxel unit is obtained, and the principal stress value is compared with the allowable stress value corresponding to the material stiffness characteristics. All voxel unit clusters whose principal stress values ​​continuously exceed the allowable stress value are marked as the stress over-limit region. From the multi-mode vibration characteristics, the strain energy density distribution of each mode is calculated, and spatial regions with strain energy density higher than the average density by a certain multiple are identified. These spatial regions are then marked as the vibration energy concentration regions. From the temporal changes of the displacement of the key nodes, analyze the displacement amplitude and the rate of change, identify nodes whose displacement amplitude exceeds the deformation threshold or whose displacement rate of change exceeds the rate threshold, and mark the nodes as the deformation-sensitive nodes. Based on the degree of excess of principal stress values ​​in the stress-over-limit region, the degree of strain energy density concentration in the vibration energy concentration region, and the degree of displacement exceeding the limit of the deformation-sensitive node, a risk level value is assigned to each stress-over-limit region, vibration energy concentration region, and deformation-sensitive node.

4. The construction design method for irregular structures based on AI simulation technology according to claim 3, characterized in that, Based on the risk classification markers and the topological connectivity, an adaptive component optimization process is initiated to generate a preliminary structural reinforcement scheme, including: The preliminary structural reinforcement scheme includes the spatial orientation, cross-sectional parameters and connection method of the newly added supporting components, as well as the adjustment strategy for the size of the original components; Using the aforementioned topological connectivity as constraints, a structural topology optimization design space for the target irregular structure is constructed; In the structural topology optimization design space, the optimization objectives are to maximize the overall structural stiffness and minimize the total structural mass. The risk classification marker is used as the penalty function, and the locations of the stress over-limit region, the vibration energy concentration region, and the deformation sensitive node are set as the optimization sensitive region. Perform topology optimization based on the variable density method to obtain the ideal distribution density cloud map of the material in the topology optimization design space of the structure; In the ideal distribution density cloud map, regions with material distribution density higher than the additive manufacturing threshold but originally designed as voids are identified, and these regions are designed as the new support components. The spatial pose, cross-sectional parameters and connection method of the new support components are determined. In the ideal distribution density cloud map, existing component areas with material distribution density below the material reduction threshold are identified, and adjustment strategies for the original component size are generated. The adjustment strategies for the original component size include size reduction or local hollowing out. By integrating all the new and adjusted design changes, a digital model of the preliminary structural reinforcement scheme is generated.

5. The construction design method for irregular structures based on AI simulation technology according to claim 4, characterized in that, Based on the internal stress field distribution of the structure and the stiffness characteristics of the material, an iterative evaluation of the construction process feasibility of the preliminary structural reinforcement scheme is conducted, including: The feasibility iterative assessment of the construction process includes component hoisting interference inspection, on-site welding / connection accessibility analysis, and temporary support structure requirement calculation. The specific steps are as follows: Based on the digital model of the preliminary structural reinforcement scheme, extract the geometric dimensions, weight, spatial coordinates and connection point information of all the newly added support components; In a three-dimensional virtual construction environment, the hoisting path of each newly added support component is simulated to check whether there is spatial interference between the newly added support component and existing structural components and temporary facilities during the hoisting process. If there is interference, the spatial pose of the newly added support component is adjusted or it is split into several sub-components that can be hoisted independently. For all locations in the preliminary structural reinforcement scheme that require on-site connection, a virtual welder / operator accessibility analysis is performed to assess whether the operating space of the connection tools is sufficient and whether the connection angle is within the allowable range of the process. If they are not accessible, the connection method is adjusted or auxiliary process holes are added. Based on the adjusted component hoisting and connection requirements, the required temporary support structure parameters are simulated and calculated, including the location, bearing capacity, and timing of removal of the temporary supports. The reinforcement scheme, after undergoing hoisting interference checks, accessibility analysis, and temporary support calculations, is re-simulated using finite element methods to verify whether its structural performance meets the requirements. If it does not, adjustments are made until all constraints are met, generating the final optimized construction design drawings and process instructions including hoisting sequence, connection process, and temporary support settings.

6. The construction design method for irregular structures based on AI simulation technology according to claim 5, characterized in that, It also includes dynamic simulation and schedule optimization steps during the construction process: Based on the optimized construction design drawings and process instructions, all construction activities are decomposed, and the logical sequence, required resources and duration of each construction activity are defined. Establish a discrete event simulation model that includes constraints on construction machinery, personnel, and material supply, and input the construction activities into the discrete event simulation model; The discrete event simulation model is driven to run, simulating the complete construction process from start to finish, and recording the start time, end time, resource usage, and waiting time caused by process conflicts and insufficient resources for each construction activity during the simulation. Based on the simulation results, the key construction activity sequences and resource bottlenecks that restrict the overall construction progress were identified; With the goal of shortening the overall construction period and balancing resource load, the logical sequence and resource allocation of the key construction activities are optimized and adjusted, and simulation is performed again, iterating until an optimized construction schedule that meets the constraints of construction period and resources is generated.

7. The construction design method for irregular structures based on AI simulation technology according to claim 6, characterized in that, The discrete event simulation model is driven to run, simulating the complete construction process from commencement to completion, including: Initialize the simulation clock and construction resource status, which includes the number of available machines, the number of workers, and the material inventory. From the available construction activities, select an activity to start execution according to priority rules, and occupy the corresponding construction resources; The simulation clock is advanced according to the duration of the construction activity, the construction resource status is updated, and it is checked whether any new construction activities meet the commencement conditions due to the completion of the preceding activities. When the construction activity is completed, the construction status update of the corresponding structural part in the parametric digital geometric model is recorded. The process of selecting activities and updating the construction status is executed cyclically until all construction activities are completed. The time displayed on the simulation clock is the total construction period.

8. The construction design method for irregular structures based on AI simulation technology according to claim 7, characterized in that, It also includes digital twin calibration and early warning steps based on real-time on-site data: A sensor network is deployed at the construction site to collect structural and environmental monitoring data in real time during the construction process. The structural monitoring data includes strain, displacement, and vibration of key parts, while the environmental monitoring data includes wind speed, temperature, and humidity. A digital twin model is established that is associated with the parametric digital geometric model. The structural monitoring data and the environmental monitoring data drive the digital twin model in real time, so that the state of the digital twin model is synchronized with the construction state of the physical structure. Based on the current state of the digital twin model, the structural safety is calculated rapidly in real time, and the calculated real-time response value is compared with the design allowable value or the predicted value in the initial dynamic response data set. A construction safety warning is generated when the real-time response value continuously exceeds the allowable range or deviates from the predicted value by more than a threshold.

9. The construction design method for irregular structures based on AI simulation technology according to claim 8, characterized in that, The structural monitoring data and environmental monitoring data are used to drive the digital twin model in real time, synchronizing the state of the digital twin model with the construction state of the physical structure, including: Establish a one-to-one mapping relationship between the sensor deployment locations and their geometric locations in the digital twin model; It receives monitoring data streams uploaded by sensors in real time and performs preprocessing such as filtering and outlier removal on the data. The preprocessed structural monitoring data is used as boundary conditions or state update parameters and input into a simplified computing kernel integrated with the digital twin model to drive the model state update. The preprocessed environmental monitoring data is converted into an equivalent load and applied to the digital twin model; The model state updated based on structural monitoring data is coupled with the load applied based on environmental monitoring data to perform calculations, so that the deformation and stress state of the digital twin model can reflect the actual state of the physical structure in real time and record the complete state evolution time history.

10. The construction design method for irregular structures based on AI simulation technology according to claim 9, characterized in that, It also includes the construction of a construction scheme knowledge base and self-learning steps: The optimized construction design drawings, process instructions, dynamic simulation logs of the construction process, and construction safety early warning records are integrated into a complete project construction case. From the project construction cases, the characteristics of irregular structures, the construction difficulties encountered, the solutions adopted and their effects are extracted to form structured knowledge entries; The structured knowledge entries are stored in the construction scheme knowledge base and associated and compared with similar cases already existing in the knowledge base. When a new irregular structure construction design project is launched, similar cases are retrieved from the construction scheme knowledge base based on the geometric feature information and physical attribute information of the new project. The design points, potential risks and successful experiences in the retrieved similar cases are used as initial constraints and inspirational information and provided to the pre-trained structural behavior learning model and the construction process feasibility iterative evaluation process.

Images