BIM high-rise suspension structure monitoring method and system based on machine learning

Abstract

The invention relates to the technical field of building structure monitoring, and discloses a BIM high-rise suspension structure monitoring method and system based on machine learning, and the method comprises the steps: constructing a physical constraint variational causal reasoning frame, and fusing BIM model data and sensing monitoring data; training a dynamic physical causal graph learning module, and identifying a causal relationship network in the high-rise suspension structure; constructing and training a hierarchical attention perception module, and capturing a structural anomaly mode; developing a physical constraint Shenchang differential equation module to accurately predict the dynamic behavior of the high-rise structure; establishing bidirectional real-time mapping between the monitoring data and BIM model elements; according to the method, physical constraint variational causal reasoning, dynamic physical causal graph learning, hierarchical attention perception, a physical constraint normal differential equation and BIM semantic mapping are fused, so that the technical problem of weak correlation characteristic monitoring of the high-rise suspension structure is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of building structure monitoring, and more specifically, it relates to a BIM high-rise suspended structure monitoring method and system based on machine learning. Background Art

[0002] The high-rise suspended structure is a new type of structural system in modern architecture. It supports the lower floors through the top truss and vertical steel tie rods to achieve the building effect of large span and large space. This structural form is increasingly widely used in buildings such as commercial complexes and cultural venues. However, the high-rise suspended structure has typical "weak correlation" characteristics. The loads are mainly transmitted between floors through steel tie rods, making the structural behavior complex and difficult to predict.

[0003] The existing structural monitoring technologies mainly include physical model-based methods and data-driven methods. The physical model-based methods rely on accurate structural models and are difficult to adapt to the complexity and uncertainty of high-rise suspended structures; although the data-driven methods can learn patterns from data, they lack the ability of physical interpretation and perform poorly in the case of sparse or abnormal data. In addition, the existing monitoring systems generally have the following problems: First, it is difficult to capture the "weak correlation" characteristics between floors in high-rise suspended structures; second, the monitoring results lack effective integration with the BIM model, resulting in difficult data interpretation; third, it is impossible to achieve predictive risk warning, mostly post-event alarms.

[0004] Therefore, there is an urgent need to develop an intelligent monitoring method and system that can effectively handle the "weak correlation" characteristics of high-rise suspended structures, integrate physical knowledge and data-driven methods, and be deeply integrated with the BIM system. Summary of the Invention

[0005] The present invention provides a BIM high-rise suspended structure monitoring method and system based on machine learning to solve the technical problem in related technologies that it is difficult to effectively monitor the "weak correlation" characteristics of high-rise suspended structures.

[0006] The present invention provides a BIM high-rise suspended structure monitoring method based on machine learning, including:

[0007] Construct a physical constraint variational causal inference framework to integrate BIM model data and sensing monitoring data;

[0008] Based on the integration result of BIM model data and sensing monitoring data, train a dynamic physical causal graph learning module to identify the causal relationship network in the high-rise suspended structure;

[0009] Use the identified causal relationship network to construct and train a hierarchical attention perception module to capture structural abnormal patterns;

[0010] Develop a physically constrained neural ordinary differential equation module based on the captured structural anomaly patterns to achieve accurate prediction of the dynamic behavior of high-rise structures;

[0011] Based on the predicted structural dynamic behavior, implement a BIM semantic mapping and visualization module to establish a two-way real-time mapping between the monitoring data and the BIM model elements.

[0012] Furthermore, the steps of constructing the physically constrained variational causal inference framework include:

[0013] Extract the geometric model, material properties, and component relationships of the high-rise suspended structure from the BIM system;

[0014] Deploy a multi-modal sensor network to collect real-time monitoring data of the structure;

[0015] Construct physical constraints based on the structural dynamics equation;

[0016] Construct a physically constrained variational causal inference framework by combining the structural dynamics equation, Bayesian causal inference, and deep learning through a unified model.

[0017] Furthermore, the steps of training the dynamic physical causal graph learning module include:

[0018] Initialize the dynamic causal graph, using the sensor monitoring points as nodes to construct the initial topological structure;

[0019] Use the physically constrained Bayesian network learning algorithm to learn the dynamic causal relationships between nodes based on the time-series monitoring data;

[0020] Apply structural sparsity regularization to optimize the causal graph based on the "weak correlation" characteristics of the high-rise suspended structure;

[0021] Dynamically update the causal graph structure, and periodically update the causal relationship network as the structural state and external environment change.

[0022] Furthermore, the steps of constructing and training the hierarchical attention perception module include:

[0023] Construct a three-level hierarchical attention network, including a time attention layer, a spatial attention layer, and a component attention layer;

[0024] Train the time attention layer, and for the monitoring data at different time scales, use the multi-head self-attention mechanism to identify the key time points;

[0025] Train the spatial attention layer, and use the graph attention network to learn the spatial correlation between different monitoring points based on the spatial distribution of the sensors and the structural topological relationship;

[0026] Component attention layer training, according to the component type and functional importance information in the BIM model, assigns dynamic attention weights to different structural components.

[0027] Further, the steps of developing the physical constraint neural ordinary differential equation module include:

[0028] Establish a neural ordinary differential equation model, representing the time evolution of the structural state as a combined form of a neural network and a physical model;

[0029] Design a hybrid loss function to optimize data fitting accuracy and physical law consistency;

[0030] Adopt an adaptive step-size numerical integration method to solve the neural ordinary differential equation and dynamically adjust the calculation step size;

[0031] By integrating the results of multiple numerical solution methods, improve the robustness of the model to noise and abnormal data.

[0032] Further, the steps of implementing the BIM semantic mapping and visualization module include:

[0033] Establish a BIM semantic mapping function to map monitoring data, causal relationships, and uncertainty information to BIM model elements;

[0034] Develop a hierarchical visualization interface to support the display of structural states at multiple scales and dimensions;

[0035] Implement adaptive computing resource allocation, and dynamically adjust computing resource allocation according to the output of the dynamic causal graph and hierarchical attention;

[0036] Construct a predictive maintenance decision support system, and automatically generate maintenance suggestions based on the model prediction results.

[0037] Further, the structural sparsity regularization is achieved through a graph structure loss function, which includes a probability logarithm term and a graph structure regularization term, and is used to promote the sparsity of the graph.

[0038] Further, the three-level hierarchical attention network combines temporal attention, spatial attention, and component attention in a cascaded manner, and sequentially performs attention calculations on the input monitoring data.

[0039] Further, the neural ordinary differential equation model combines neural network representation and physical constraint terms to realize the dynamic evolution prediction of the structural state over time.

[0040] The present invention provides a BIM high-rise suspension structure monitoring system based on machine learning, which is used to execute the above-mentioned BIM high-rise suspension structure monitoring method based on machine learning, including:

[0041] A physical constraint variational causal inference module for fusing BIM model data and sensing monitoring data to form a unified data processing basis;

[0042] A dynamic physical causal graph learning module for identifying the causal relationship network in high-rise suspended structures;

[0043] A hierarchical attention perception module for capturing structural anomaly patterns from multiple temporal and spatial scales;

[0044] A physical constraint neural ordinary differential equation module for accurately predicting the dynamic behavior of high-rise suspended structures;

[0045] A BIM semantic mapping and visualization module for establishing two-way real-time mapping between monitoring data and BIM model elements.

[0046] The beneficial effects of the present invention are as follows: By integrating technologies such as physical constraint variational causal inference, dynamic physical causal graph learning, hierarchical attention perception, physical constraint neural ordinary differential equations, and BIM semantic mapping, the technical problem of monitoring the "weak correlation" characteristics of high-rise suspended structures is solved;

[0047] Predictive risk warning is realized, and potential risks can be identified in advance; The causal explanation ability is improved, and multi-step causal chains can be automatically identified and traced; The utilization of computing resources is optimized, and the computing load is reduced; The adaptability to the "weak correlation" structural characteristics is improved, and the accuracy in identifying the weak association patterns of suspended structures is increased; The deep integration of BIM and the monitoring system is realized, and the data interpretation workload is reduced; It has the ability of self-evolution and can automatically adjust the focus of attention and the prediction model as the project progresses. Description of the Drawings

[0048] Figure 1 It is a flowchart of a BIM high-rise suspended structure monitoring method based on machine learning in the present invention. Detailed Embodiments

[0049] Now the subject matter described herein will be discussed with reference to exemplary embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein, and changes can be made to the functions and arrangements of the elements discussed without departing from the scope of protection of the content of this specification. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described in some examples can also be combined in other examples.

[0050] At least one embodiment of the present invention discloses a BIM high-rise suspended structure monitoring method based on machine learning, as Figure 1 shown, including:

[0051] Step 1: Construct a physical constraint variational causal inference framework to integrate BIM model data and sensing monitoring data;

[0052] Step 1.1: Extract BIM data;

[0053] Extract the geometric model, material properties, and component relationships of the high-rise suspended structure from the BIM system, including information such as the positions of steel tie rods, connection node parameters, floor layouts, etc., to generate a digital representation of the structure.

[0054] Step 1.2: Deploy a sensing network;

[0055] Deploy a multi-modal sensor network to collect real-time monitoring data of the structure, including physical quantities such as displacement, strain, inclination angle, temperature, etc., to form an original monitoring data stream.

[0056] Step 1.3: Construct physical constraints;

[0057] Construct physical constraints based on the structural dynamics equation, and represent the mechanical behavior of the high-rise suspended structure as a mathematical model, including:

[0058] Structural equilibrium equation:

[0059] ;

[0060] where is the stiffness matrix; is the displacement vector; is the external force vector;

[0061] Material constitutive relationship:

[0062] ; <s

[0063] where is the stress; is the material matrix; is the strain;

[0064] Geometric compatibility condition:

[0065] ;

[0066] where is the strain; is the strain-displacement matrix; is the displacement vector;

[0067] Step 1.4: Establish a PC-VCI framework;

[0068] Construct a Physically-Constrained Variational Causal Inference (PC-VCI) framework, which is uniformly represented by the following formula:

[0069] ;

[0070] where is the probability density function; is the causal graph; is the latent variable; is the model parameter; is the observed data; is the physical constraint; is the conditional probability separator;

[0071] This framework unifies structural dynamics equations, Bayesian causal inference, and deep learning in a mathematical model to achieve comprehensive inference of the state of high-rise suspended structures.

[0072] The specific implementation of the PC-VCI framework includes three core units:

[0073] The variational encoder unit, the physical constraint unit, and the probabilistic graphical inference unit.

[0074] The variational encoder unit consists of a multi-layer feedforward neural network that maps the observed data to the latent space Z;

[0075] The physical constraint unit realizes the conversion of mechanical equations into constraint terms through the Lagrange multiplier method;

[0076] The probabilistic graphical inference unit uses the Monte Carlo variational inference algorithm to estimate the posterior distribution.

[0077] The entire framework realizes the deep integration of mechanical knowledge and data-driven by alternately optimizing the parameters of these three units. In the application of high-rise suspended structures, this framework specifically introduces the temperature-stress relationship as a physical constraint for the characteristics of steel tie rods affected by temperature, improving the ability to separate temperature effects and structural problems.

[0078] Step 2: Based on the fusion result of BIM model data and sensing monitoring data, train the dynamic physical causal graph learning module to identify the causal relationship network in the high-rise suspended structure;

[0079] Step 2.1: Initialize the dynamic causal graph;

[0080] Construct an initial topological structure with the sensor monitoring points as nodes. For high-rise suspended structures, the nodes include stress sensors of tie rods on each floor, displacement sensors at connection nodes, temperature sensors, etc., and the initial topology is determined based on the component connection relationship in the BIM model.

[0081] Step 2.2, learning dynamic causal relationships;

[0082] Use the physically constrained Bayesian network learning algorithm to learn the dynamic causal relationships between nodes based on time series monitoring data. The formula is as follows:

[0083] ;

[0084] where is the probability function; is the causal graph structure at time is the causal graph structure at time is the observed data at time is the physical constraint; is the conditional probability separator;

[0085] This algorithm overcomes the limitations of pure data-driven methods by combining monitoring data and physical knowledge, and improves the accuracy of causal discovery.

[0086] The specific implementation of the physically constrained Bayesian network learning algorithm is based on the Structural Equation Modeling (SEM) framework, which includes a graph structure search module and a parameter learning module. The graph structure search module uses a scoring-based method to define a scoring function , representing the score of the graph structure G given the data and the physical constraint Φ. This module first generates a candidate graph set, and then uses a greedy search algorithm to select the optimal graph structure. The parameter learning module is based on the principle of maximum likelihood estimation and uses the expectation maximization algorithm with physical constraints to optimize the model parameters. In the actual monitoring of high-rise suspension structures, this algorithm maintains the sparse connectivity of the graph structure, which is particularly suitable for capturing the weak correlations between tie rods on different floors of the suspension structure and satisfies the mechanical equilibrium constraints.

[0087] Step 2.3, optimizing sparsity;

[0088] Apply structural sparsity regularization to optimize the causal graph based on the "weak correlation" characteristics of high-rise suspension structures. The optimization objective function is:

[0089] ;

[0090] where is the loss function of the graph; is the natural logarithm function; is the probability function; is the observed data; is the causal graph; is a physical constraint; is a sparsity balance parameter; is the L1 norm of the graph structure; is a conditional probability separator;

[0091] This step specifically targets the characteristics of each floor in the suspension structure being connected by independent tie rods, and can accurately capture the "weak correlation" causal relationship between floors.

[0092] Optionally, in some embodiments, the structural sparsity regularization can adopt a more complex group sparsity regularization method, expressed as:

[0093] ;

[0094] where is the loss function of the graph after adopting a more complex group sparsity regularization method; is the natural logarithm function; is the probability function; is the observed data; is the causal graph; is a physical constraint; is the L1 norm balance parameter; is the L1 norm of the graph structure; is the group norm balance parameter; is the group norm of the graph structure; is a conditional probability separator;

[0095] This method is particularly suitable for groups of components with similar functions in high-rise suspension structures (such as multiple tie rods on the same floor), and can better capture the association patterns between components with similar functions.

[0096] In another embodiment, a prior graph structure initialization method based on expert knowledge can be adopted. By combining the experience of structural engineers, a better initial state can be provided for the dynamic causal graph. Specifically, the influence matrix calculated by structural analysis software can be used as prior knowledge to reduce the search space for causal discovery and accelerate the algorithm convergence, which is applicable to scenarios with a clear understanding of the structural system.

[0097] Step 2.4, dynamically update the causal graph structure;

[0098] As the structural state and external environment change, periodically update the causal relationship network to ensure that the model can adapt to the dynamic changes of high-rise suspension structures in different construction stages and usage stages.

[0099] Step 3, use the identified causal relationship network to construct and train a hierarchical attention perception module to capture structural abnormal patterns;

[0100] Specifically, step 3 includes the following sub-steps:

[0101] Step 3.1, construct a three-level hierarchical attention network;

[0102] The three-level hierarchical attention network includes a temporal attention layer , a spatial attention layer and a component attention layer , which is formally represented as:

[0103] ;

[0104] where is the hierarchical attention function; is the input monitoring data; is the temporal attention layer function; is the spatial attention layer function; is the component attention layer function;

[0105] This network is specifically designed for the possible local-to-global anomaly propagation patterns in high-rise suspension structures and can capture structural anomalies across scales.

[0106] The specific implementation of the three-level hierarchical attention network adopts a multi-branch parallel computing architecture, including a feature extraction backbone network and three parallel attention branches. The feature extraction backbone network uses a one-dimensional convolutional neural network to extract temporal features and combines a graph convolutional network to capture spatial topological relationships. The three attention branches respectively implement attention mechanisms in different dimensions: the temporal attention branch uses a bidirectional long short-term memory network (BiLSTM) to capture long-term dependencies; the spatial attention branch uses a graph attention network (GAT) to process irregular spatial topologies; the component attention branch processes component attribute information through a multi-layer perceptron (MLP). The outputs of each branch are integrated through an adaptive fusion module to generate the final attention weights. In the monitoring of high-rise suspension structures, this network particularly strengthens the attention to the connection nodes of steel tie rods, which are usually key areas with abnormal deformations.

[0107] Step 3.2, temporal attention layer training;

[0108] For monitoring data at different time scales, a multi-head self-attention mechanism is used to identify key time points. For the input sequence:

[0109] ;

[0110] where Denote the time - series monitoring data set, , , respectively represent the monitoring data at the 1st, 2nd, th time step, and \(L\) is the total length of the time series.

[0111] Calculate the attention weights:

[0112] ;

[0113] where is the attention weight at time \(t\); is the softmax function; is the query weight matrix; is the key weight matrix; is the input data at time \(t\); is the attention input data; is the transpose operator; is the square - root function; is the feature dimension.

[0114] This layer can identify early signals of structural behavior anomalies from different time scales (minute - level, hour - level, day - level).

[0115] Step 3.3, Spatial attention layer training;

[0116] Using the graph attention network, based on the spatial distribution of sensors and the structural topological relationship, learn the spatial correlation between different monitoring points. For nodes and its neighbor , calculate the attention coefficient:

[0117] ;

[0118] where is the attention coefficient between nodes and is the softmax function; is the leaky rectified linear unit activation function; is the learnable parameter vector; is the transpose operator; is the learnable parameter matrix; is the feature vector of node \(i\); is the feature vector of node \(j\); is the vector concatenation operator.

[0119] This layer is particularly suitable for capturing the “weakly correlated” spatial relationships between floors in a suspended structure.

[0120] Step 3.4, component attention layer training;

[0121] According to the component type, functional importance and other information in the BIM model, dynamic attention weights are assigned to different structural components. The component attention weight is calculated as:

[0122] ;

[0123] in For components The attention weight of is the activation function; is the learnable parameter matrix of component attention; For components The eigenvector of Components extracted from the BIM model The attribute vector of .

[0124] This layer ensures that higher attention is paid to the critical components of the structure (such as connection nodes and steel ties).

[0125] For example, in practical applications, different attention weight calculation methods can be used for different types of monitoring scenarios. For construction phase monitoring, a dynamic gated attention mechanism can be used, which is expressed as:

[0126] ;

[0127] in To adopt the dynamic gated attention mechanism The attention weight of For components The gating coefficient of is the activation function; is the learnable parameter matrix of component attention; For components The eigenvector of Components extracted from the BIM model The attribute vector of .

[0128] This method can automatically adjust the attention paid to different components according to the construction stage, and give higher weight when the component status is changed during construction (such as installation and prestressing).

[0129] In some implementations, an uncertainty-based attention allocation strategy can also be introduced. This strategy estimates the uncertainty of the monitoring data through a Bayesian neural network and allocates more attention to high-uncertainty areas:

[0130] ;

[0131] where is the attention weight of the component after allocating more attention to the high-uncertainty region ; is the activation function; is the learnable parameter matrix of the component attention; is the component feature vector; is the attribute vector of the component extracted from the BIM model; is the component estimated uncertainty.

[0132] This method is particularly suitable for situations where sensor noise is inconsistent or data is missing in some areas, and can improve the robustness of the monitoring system to uncertainty.

[0133] Step 4: According to the captured structural anomaly patterns, develop a physical-constrained neural ordinary differential equation module to achieve accurate prediction of the dynamic behavior of high-rise structures;

[0134] Step 4.1: Establish a neural ODE model;

[0135] Establish a neural ordinary differential equation model, representing the time evolution of the structural state as:

[0136] ;

[0137] where is the derivative of the structural state with respect to time ; is the structural state vector at time ; is the neural network function with parameter ; is the physical model term function; is the time variable; is the neural network parameter.

[0138] This model ensures that the prediction results conform to the laws of structural mechanics by explicitly introducing physical constraints.

[0139] The specific implementation of the neural ordinary differential equation model adopts a hybrid architecture, including a physics-independent neural network part and the physical model term function . The neural network part adopts a multi-layer residual network architecture, including three hidden layers with skip connections, each layer having 64 neurons, and using the GELU activation function. The physical constraint part is based on the simplified structural dynamics equation, and physical quantities such as the axial stiffness of the tie rod and the influence of temperature change are specifically introduced for high-rise suspended structures. The model is trained using the Adam optimizer, with the learning rate set to 0.001, the batch size to 64, and the number of training iterations to 1000 rounds. For numerical integration, the Dormand-Prince method with adaptive step size is adopted, and the tolerance is set to 1e-5. In practical applications, the model can effectively capture the nonlinear dynamic responses of high-rise suspended structures under conditions such as temperature change and load redistribution.

[0140] Step 4.2, design the hybrid loss function;

[0141] Design the hybrid loss function to optimize both the data fitting accuracy and the physical law consistency simultaneously:

[0142] ;

[0143] where is the total loss function; is the data fitting loss function; is the physical loss trade-off parameter; is the physical constraint loss function.

[0144] The physical constraint loss is specifically defined as:

[0145] ;

[0146] where is the physical constraint loss function; is the square of the L2 norm; is the physical model term function; is the structural state vector at time is the structural state with respect to time derivative; is the time variable.

[0147] The physical constraint loss ensures that the model predictions are consistent with both the observed data and the structural mechanics laws.

[0148] Step 4.3, adaptive solution;

[0149] Adopt the numerical integration method with adaptive step size to solve the neural ordinary differential equation, dynamically adjust the calculation step size, and improve the calculation efficiency on the premise of ensuring accuracy. For the time interval , the solution process is expressed as:

[0150] ;

[0151] wherein is the structural state at time is the structural state at time is the definite integral operator from to ; is the neural network function with parameter ; is the structural state vector at time is the physical constraint balance parameter; is the physical model term function; is the time differential element; is the time variable; is the starting time; is the ending time; is the neural network parameter.

[0152] This method is particularly suitable for dealing with the multi-time scale dynamic responses that may occur in high-rise suspended structures.

[0153] Optionally, according to specific application requirements and computing resource limitations, different numerical solution strategies can be adopted. For scenarios that require high precision but have sufficient computing resources, high-order Runge-Kutta methods such as the RK45 (Dormand-Prince) algorithm can be used; for applications with high real-time requirements, simplified Euler methods or improved Euler methods can be adopted, sacrificing some precision in exchange for computing speed.

[0154] For example, when running on edge computing devices, an approximate solution method based on Taylor expansion can be adopted:

[0155] ;

[0156] wherein is the approximate structural state at time is the structural state at time is the time step; is the neural network function with parameter ; is the second-order term coefficient; is the derivative of the neural network function with respect to time; is the time variable; is the neural network parameter.

[0157] This method has low computational overhead and is suitable for monitoring terminal devices with limited resources.

[0158] In another implementation, for applications where long-term behavior is of particular interest, an inverse stochastic differential equation solution method can be used, which is more suitable for capturing rare events and long-tail distribution phenomena that may exist in high-rise suspended structures:

[0159] ;

[0160] in for The structural state at the moment; End time structural state; For arrive The definite integral operator of ; The parameters are Neural network function; is the integration variable The structural state at the moment; is the physical constraint balance parameter; is the physical model term function; is the differential element of the integral variable; is the noise factor; for The Wiener process of moments; is the end time; is the time variable; is the integral variable; are the neural network parameters.

[0161] This method can more effectively explore the possible evolution paths of the system under different initial conditions and is suitable for risk assessment and extreme event analysis.

[0162] Step 4.4, enhance robustness;

[0163] By integrating the results of multiple numerical solution methods, the model's robustness to noise and abnormal data is improved. The integrated prediction result is calculated as:

[0164] ;

[0165] in for The integrated prediction results at the moment; is the symbol for summation; For the The weight of the method; For the Numerical solution method in The prediction results at the moment; is the total number of numerical solution methods.

[0166] This effectively reduces the numerical instability that may be brought about by a single solution method and improves the prediction accuracy of the complex dynamic behavior of high-rise suspended structures.

[0167] Step 5: Based on the predicted structural dynamic behavior, implement the BIM semantic mapping and visualization module to establish a two-way real-time mapping between the monitoring data and the BIM model elements;

[0168] Step 5.1: Establish semantic mapping;

[0169] Establish a BIM semantic mapping function to map the monitoring data, causal relationship, and uncertainty information to the BIM model elements:

[0170] ;

[0171] where is the BIM semantic mapping function; is the state estimation data; is the causal relationship data; is the uncertainty measurement data; is the BIM model element; is the mapping direction symbol

[0172] The BIM semantic mapping function transforms the abstract monitoring results into a three-dimensional visual representation through semantic association.

[0173] The specific implementation of the BIM semantic mapping function is based on a multi-level semantic indexing architecture, including three levels: element-level indexing, component-level indexing, and system-level indexing. The underlying element-level indexing establishes a direct correspondence between the monitoring data points and the geometric elements in the BIM model; the middle-level component-level indexing organizes the relevant elements into meaningful structural components, such as steel tie rods, connection nodes, etc.; the top-level system-level indexing constructs a relationship network between the components, corresponding to the causal graph structure. The mapping process uses a graph-based search algorithm. First, the element-level correspondence is determined through spatial position matching, then the component-level mapping is established based on the component classification information of the BIM model, and finally the system-level mapping is constructed using the topological relationship. In the application of high-rise suspended structures, this function particularly strengthens the semantic understanding of the key components (such as tie rods, connection nodes) in the suspension system, enabling the monitoring results to accurately correspond to the structural entities.

[0174] Step 5.2: Develop a visualization interface;

[0175] Develop a hierarchical visualization interface to support the display of structural states at multiple scales and dimensions. The interface includes:

[0176] Global view: Visualize the overall structural health status, and use color coding to represent the degree of abnormality;

[0177] Local view: detailed status information of key nodes and components, including deformation, stress, and warning levels;

[0178] Time view: historical evolution and future prediction trends of the structural status, supporting time-axis interaction.

[0179] The visualization interface enables engineers to intuitively understand the complex status information of high-rise suspended structures.

[0180] In some embodiments, different visualization strategies can be adopted according to the usage scenario and user requirements. For example, for mobile devices used at the construction site, simplified 2D representations and symbolic coding can be used to reduce the rendering burden while retaining key information; for the large-screen display in the control center, immersive 3D visualization can be adopted, combined with ambient lighting and material rendering, to provide a more rich information expression.

[0181] Optionally, a on-site visualization solution can also be implemented based on augmented reality (AR) technology. By identifying the structural location through the mobile device camera, the monitoring data and warning information are directly superimposed on the actual structural image, achieving a "perspective" effect, enabling on-site personnel to intuitively understand the status of invisible components. This method is particularly suitable for maintenance and repair scenarios.

[0182] For team collaboration scenarios, a cloud-based shared visualization platform can also be adopted, allowing different professionals (such as structural engineers, construction teams, and owners) to view the same monitoring information simultaneously, but display different levels and types of data according to their respective needs. This method supports remote collaboration and multi-level decision-making, improving the emergency response efficiency.

[0183] Step 5.3, adaptive resource allocation;

[0184] Implement adaptive computing resource allocation. Dynamically adjust the computing resource allocation according to the output of the dynamic causal graph and hierarchical attention to optimize the computing efficiency:

[0185] ;

[0186] where is the computing resource allocated to the monitoring point ; is the total computing resource; is the exponential function; is the attention score of the monitoring point ; is the summation symbol; is the attention score of the monitoring point ; is the total number of monitoring points.

[0187] This mechanism ensures that limited computing resources are preferentially allocated to high-risk areas, reducing the computational overhead while maintaining monitoring accuracy.

[0188] Step 5.4, construct decision support;

[0189] Construct a predictive maintenance decision support system, and based on the model prediction results, automatically generate maintenance recommendations, including:

[0190] Priority ranking: Sort components according to risk assessment to identify parts that need to be inspected first;

[0191] Intervention recommendation: Provide solutions to potential problems based on causal analysis;

[0192] Operation plan: Generate a specific inspection and maintenance schedule.

[0193] This system converts complex monitoring data into directly executable decision-making recommendations, improving the operation and maintenance efficiency of high-rise suspended structures.

[0194] A BIM high-rise suspended structure monitoring system based on machine learning, which is used to execute the above-mentioned BIM high-rise suspended structure monitoring method based on machine learning, includes:

[0195] A physical constraint variational causal inference module, which is used to fuse BIM model data and sensing monitoring data to form a unified data processing basis;

[0196] A dynamic physical causal graph learning module, which is used to identify the causal relationship network in high-rise suspended structures;

[0197] A hierarchical attention perception module, which is used to capture structural anomaly patterns from multiple time and space scales;

[0198] A physical constraint neural ordinary differential equation module, which is used to accurately predict the dynamic behavior of high-rise suspended structures;

[0199] A BIM semantic mapping and visualization module, which is used to establish a two-way real-time mapping between monitoring data and BIM model elements.

[0200] Here, the present invention provides an implementation example:

[0201] This application example is for a 28-story commercial and office complex. The 11th - 28th floors adopt a suspended structure design and support the lower floors through the top truss and vertical steel tie rods. This building has typical "weak correlation" structural characteristics. The loads are mainly transmitted between floors through steel tie rods. The diameter of the steel tie rods is 80 - 120 mm, and the length is 3.6 - 4.2 m. The monitoring requirements cover the construction process and the use stage, with a focus on the stress changes of the tie rods, the deformation of the connection nodes, and the overall stability of the structure.

[0202] The main challenges faced by the project include:

[0203] The structure is sensitive to temperature changes, and the thermal expansion and contraction of steel tie rods may lead to stress redistribution;

[0204] During the construction process, the load state is constantly changing, and the structural state needs to be evaluated in real time;

[0205] The irregular design makes it difficult to accurately evaluate the structural safety status using traditional monitoring methods.

[0206] In this application example, 158 sensors are deployed, including 48 strain sensors (installed on steel tie rods), 36 displacement sensors (installed at key connection nodes), 32 inclination sensors, and 42 temperature sensors. All sensors are bound to the BIM model to form a digital twin system with semantic associations.

[0207] First, extract the structural geometric model and material property information from the project's BIM system to generate a digital representation of the structure. Some key component information extracted from the BIM model is shown in Table 1:

[0208] Table 1: Example of Structural Component Information Extracted from the BIM Model

[0209]

[0210] Among them, Φ120, Φ100, and Φ80 represent circular cross-section diameters of 120mm, 100mm, and 80mm respectively;

[0211] Subsequently, combined with the deployed sensor network, collect the real-time monitoring data of the structure, and form a monitoring data stream available for analysis after preprocessing. Some examples of preprocessed sensor data are shown in Table 2:

[0212] Table 2: Example of Sensor Monitoring Data (Sampling Time: 2022-07-15 14:00-15:00)

[0213]

[0214] Based on structural mechanics theory, physical constraints for the suspended structure are constructed, including the relationship between the axial force and deformation of steel tie rods, the influence of temperature changes on the stress of tie rods, and the node equilibrium equation, etc. These physical constraints are transformed into mathematical expressions and embedded into the PC-VCI framework. Some key physical constraints and their mathematical expressions are shown in Table 3:

[0215] Table 3: Example of Physical Constraints and Their Mathematical Expressions

[0216]

[0217] Based on the above data and constraints, the PC-VCI framework realizes the overall inference of the structural state and establishes the foundation of the causal relationship network among the monitoring variables.

[0218] According to the results processed by the PC-VCI framework, an initial dynamic causal graph is constructed. The nodes of the graph correspond to the sensor monitoring points, and the initial topology is determined based on the component connection relationships in the BIM model. Some key nodes and their initial connection relationships are shown in Table 4:

[0219] Table 4: Examples of Causal Graph Nodes and Their Physical Meanings

[0220]

[0221] Subsequently, based on the historical monitoring data for 3 months (from April 15, 2022 to July 15, 2022), the system applies the physical constraint Bayesian network learning algorithm to identify the true causal relationships between the nodes. The system identifies the influence of temperature changes on the strain of steel tie rods and the weak correlations existing between the tie rods on different floors. Some important causal relationships discovered by the system and their strengths are shown in Table 5:

[0222] Table 5: Key Causal Relationships and Strengths Discovered by the System

[0223]

[0224] Based on the established causal graph, the system constructs and trains a three-level hierarchical attention network. The network includes a temporal attention layer, a spatial attention layer, and a component attention layer, which together constitute the attention mechanism. After the model training is completed, examples of the attention weights assigned to different structural components are shown in Table 6:

[0225] Table 6: Component Attention Weights Assigned by the Hierarchical Attention Network

[0226]

[0227] To achieve accurate prediction of the structural dynamic behavior, the system establishes a physical constraint neural ordinary differential equation model. The performance comparison of the model under different prediction targets and durations is shown in Table 7:

[0228] Table 7: Prediction Accuracy of the Physical Constraint Neural Ordinary Differential Equation Model

[0229]

[0230] The system establishes a two-way mapping between the monitoring data and the BIM model elements through the BIM semantic mapping function. The mapping success rate statistics and related performance indicators are shown in Table 8:

[0231] Table 8: BIM Semantic Mapping Performance Indicators

[0232]

[0233] The embodiments of the present invention have been described above. However, these embodiments are not limited to the specific implementation manners described above. The above specific implementation manners are merely illustrative rather than restrictive. Under the inspiration of these embodiments, those of ordinary skill in the art can also make more equivalent embodiments in various forms, all of which fall within the protection scope of these embodiments.

Claims

1. A BIM high-rise suspended structure monitoring method based on machine learning, characterized in that Including: Construct a physical constraint variational causal inference framework, integrating BIM model data and sensing monitoring data; Based on the fusion result of BIM model data and sensing monitoring data, train a dynamic physical causal graph learning module to identify the causal relationship network in the high-rise suspended structure; Utilize the identified causal relationship network to construct and train a hierarchical attention perception module to capture the structural anomaly pattern; According to the captured structural anomaly pattern, develop a physical constraint neural ordinary differential equation module to achieve accurate prediction of the dynamic behavior of the high-rise structure; Based on the predicted structural dynamic behavior, implement a BIM semantic mapping and visualization module to establish a two-way real-time mapping between the monitoring data and the BIM model elements.

2. The BIM high-rise suspension structure monitoring method based on machine learning according to claim 1, characterized in that, The steps of constructing the physical constraint variational causal inference framework include: Extract the geometric model, material properties and component relationships of the high-rise suspended structure from the BIM system; Deploy a multi-modal sensor network to collect the real-time monitoring data of the structure; Construct physical constraints based on the structural dynamics equation; Construct a physical constraint variational causal inference framework, combining the structural dynamics equation, Bayesian causal inference and deep learning through a unified model.

3. A BIM high-rise suspension structure monitoring method based on machine learning according to claim 1, characterized in that, The steps of training the dynamic physical causal graph learning module include: Initialize the dynamic causal graph, with the sensor monitoring points as nodes, to construct the initial topological structure; Utilize the physical constraint Bayesian network learning algorithm to learn the dynamic causal relationships between nodes based on the time-series monitoring data; Apply structural sparsity regularization to optimize the causal graph based on the "weak correlation" characteristics of the high-rise suspended structure; Dynamically update the causal graph structure, and periodically update the causal relationship network as the structural state and external environment change.

4. A BIM high-rise suspension structure monitoring method based on machine learning according to claim 1, characterized in that, The steps of constructing and training the hierarchical attention perception module include: Construct a three-level hierarchical attention network, including a temporal attention layer, a spatial attention layer and a component attention layer; Temporal attention layer training, for the monitoring data at different time scales, adopt the multi-head self-attention mechanism to identify the key time points; Spatial attention layer training, utilize the graph attention network to learn the spatial correlation between different monitoring points based on the spatial distribution of sensors and the structural topological relationship; Component attention layer training, according to the component type and functional importance information in the BIM model, assign dynamic attention weights to different structural components.

5. A BIM high-rise suspended structure monitoring method based on machine learning according to claim 1, characterized in that, The steps of developing the physical constraint neural ordinary differential equation module include: Establish a neural ordinary differential equation model, representing the time evolution of the structural state as a combined form of a neural network and a physical model; Design a hybrid loss function to optimize the data fitting accuracy and the consistency of physical laws; Adopt an adaptive step-size numerical integration method to solve the neural ordinary differential equation, dynamically adjusting the calculation step size; By integrating the results of multiple numerical solution methods, improve the robustness of the model to noise and abnormal data.

6. A BIM high-rise suspension structure monitoring method based on machine learning according to claim 1, characterized in that The steps of implementing the BIM semantic mapping and visualization module include: Establish a BIM semantic mapping function to map the monitoring data, causal relationships and uncertainty information to the BIM model elements; Develop a hierarchical visualization interface to support the display of the structural state at multiple scales and dimensions. Implement adaptive computing resource allocation, and dynamically adjust the computing resource allocation according to the output of the dynamic causal graph and hierarchical attention. Construct a predictive maintenance decision support system, and automatically generate maintenance suggestions based on the model prediction results.

7. A BIM high-rise suspension structure monitoring method based on machine learning according to claim 3, characterized in that, The structural sparsity regularization is realized through a graph structure loss function, which includes a probability logarithm term and a graph structure regularization term to promote the sparsity of the graph.

8. A BIM high-rise suspension structure monitoring method based on machine learning according to claim 4, characterized in that, The three-level hierarchical attention network combines temporal attention, spatial attention, and component attention in a cascaded manner, and sequentially performs attention calculations on the input monitoring data.

9. A BIM high-rise suspended structure monitoring method based on machine learning according to claim 5, characterized in that, The neural ordinary differential equation model combines neural network representation and physical constraint terms to realize the dynamic evolution prediction of the structural state over time.

10. A BIM high-rise suspension structure monitoring system based on machine learning, characterized in that, A method for monitoring a BIM high-rise suspended structure based on machine learning according to any one of claims 1-9, comprising: A physical constraint variational causal inference module for fusing BIM model data and sensing monitoring data to form a unified data processing basis. A dynamic physical causal graph learning module for identifying the causal relationship network in the high-rise suspended structure. A hierarchical attention perception module for capturing structural anomaly patterns from multiple time and space scales. A physical constraint neural ordinary differential equation module for accurately predicting the dynamic behavior of the high-rise suspended structure. A BIM semantic mapping and visualization module for establishing a two-way real-time mapping between the monitoring data and the BIM model elements.

Images