Intelligent monitoring method and device for construction, operation and maintenance of super high-rise structure

Abstract

The invention provides an intelligent monitoring method and device for construction, operation and maintenance of a super high-rise structure, and solves the technical problems that the monitoring process is lack of timeliness and continuity and technical support cannot be provided for accurate operation and maintenance of urban infrastructures due to the fact that the existing monitoring means cannot fully utilize monitoring data. The method comprises the following steps: processing prior data according to Bayesian fast modal recognition to form a modal recognition model of a complex structure, and updating the modal recognition model according to observation data; establishing a structural dynamic model according to the modal characteristics, and evaluating the dynamic performance of the complex structure according to the dynamic model; and extracting surface features of the complex structure according to the dynamic model, training a regional convolutional neural network according to the surface features, mapping the surface features and dynamic performance through the regional convolutional neural network, and performing structure deformation prediction according to updating of the modal recognition model. And a theoretical method and engineering equipment support are provided for safe operation of major urban infrastructures.

Description

Technical Field

[0001] This invention relates to the field of structural monitoring technology, specifically to an intelligent monitoring method and device for the construction and operation and maintenance of super high-rise structures. Background Technology

[0002] Large-scale complex structures formed during the construction of hub projects, especially urban complexes with semi-underground or underground reinforced concrete frames, are often affected by the construction and operation of transportation facilities such as subway tunnels and railway tracks, and face potential risks from geological and load changes. For example, the excavation of foundation pits, tunnel boring, and dewatering settlement during subway construction are the main causes of geological changes, which indirectly lead to changes in load transmission paths. These changes, combined with the dynamic loads from the complex itself and transportation operations, form a chain of risks: geology-load-structure. Geological changes are mainly manifested in uneven settlement and differential deformation wind geological risks, directly impacting the load-bearing foundation of the reinforced concrete frame. The existing foundation pit and the subway foundation pit may form a "double foundation pit" superposition, or the subway foundation pit may be adjacent to the complex's foundation pit, significantly increasing the risk of foundation pit slope instability and soil slippage. Subway construction (dewatering, tunnel boring) and dewatering of the complex's foundation pit will change the regional groundwater level, inducing risks caused by groundwater disturbance. Load changes are mainly manifested in the dynamic load risks brought about by rail transit operations. The vibrations and impacts generated by operation are long-term and continuous dynamic loads, easily leading to structural fatigue damage. The static load changes used by the complex itself also pose risks due to its function. Complex structures (such as underground commercial areas, parking garages, and equipment floors) are prone to exceeding design limits due to load adjustments during use. External load disturbances, and changes in external loads during track construction and operation, can indirectly affect the complex's structure. When geological changes and load variations overlap, the risk level amplifies, creating a vicious cycle. Current technologies for geological surveys and monitoring, structural design optimization, construction coordination control, load management, and durability design primarily aim to mitigate core risks by increasing the structural margin of large complexes. However, traditional monitoring methods suffer from technical bottlenecks in identification speed, accuracy, uncertainty expression, and intelligent management. They cannot provide continuous monitoring and evaluation of complex structures given the complex operating environment, variable load conditions, and frequent disturbances. This results in persistent industry pain points throughout the entire lifecycle of large complexes, including difficulties in identification, insufficient prediction, delayed diagnosis, and low levels of intelligence. Summary of the Invention

[0003] In view of the above problems, embodiments of the present invention provide an intelligent monitoring method and device for the construction and operation and maintenance of super high-rise structures, which solves the technical problem that existing monitoring methods cannot make full use of monitoring data, resulting in a lack of timeliness and continuity in the monitoring process, and failing to provide technical support for the precise operation and maintenance of urban infrastructure.

[0004] The intelligent monitoring method for construction and operation and maintenance of super high-rise structures according to embodiments of the present invention includes:

[0005] Based on Bayesian fast modal recognition, a modal recognition model of the complex structure is formed by processing prior data, and the modal recognition model is updated based on the observation data;

[0006] A structural dynamics model is established based on modal characteristics, and the dynamic performance of the complex structure is evaluated based on the dynamics model.

[0007] The surface features of the complex structure are extracted based on the dynamic model. A regional convolutional neural network is trained based on the surface features. The surface features are mapped to the dynamic performance through the regional convolutional neural network. The structural deformation is predicted based on the modal recognition model update.

[0008] In one embodiment of the present invention, the construction of the modality recognition model includes:

[0009] -Perform signal acquisition and preprocessing;

[0010] - Construct a parameterized modality recognition model;

[0011] - Fast solution for the posterior distribution;

[0012] - Uncertainty quantification and result verification.

[0013] In one embodiment of the present invention, the evaluation of the dynamic performance of the complex structure using a dynamic model includes:

[0014] Modal characteristic analysis;

[0015] -Structural condition assessment;

[0016] - Reliability and risk assessment.

[0017] In one embodiment of the present invention, the mapping of surface features and dynamic performance through a region convolutional neural network includes:

[0018] First, image feature maps are extracted using backbone networks such as ResNet;

[0019] Candidate regions are generated through a region proposal network.

[0020] The ROI Align layer extracts candidate region features from the feature map;

[0021] The classification branch determines whether the region is a target feature point;

[0022] Regression branch predicts the precise location of feature points;

[0023] Displacement is calculated by tracking the positional changes of feature points in consecutive frames through time-series analysis.

[0024] The intelligent monitoring device for construction and operation and maintenance of super high-rise structures according to embodiments of the present invention includes:

[0025] The memory is used to store the program code in the process of the intelligent monitoring method for the construction and operation and maintenance of super high-rise structures.

[0026] A processor for executing the program code.

[0027] The intelligent monitoring device for construction and operation and maintenance of super high-rise structures according to embodiments of the present invention includes:

[0028] The modality recognition module is used to process prior data using Bayesian fast modality recognition to form a modality recognition model of the complex structure, and to update the modality recognition model based on observation data;

[0029] The performance evaluation module is used to establish a structural dynamics model based on modal characteristics and evaluate the dynamic performance of the complex structure based on the dynamics model.

[0030] The feature mapping module is used to extract surface features of the complex structure based on the dynamic model, train a regional convolutional neural network based on the surface features, map the surface features and dynamic performance through the regional convolutional neural network, and predict structural deformation based on the modal recognition model update.

[0031] The intelligent monitoring method and device for the construction and operation and maintenance of super high-rise structures in this invention provide theoretical and engineering equipment support for the safe operation of major urban infrastructure. It improves the efficiency and reliability of modal identification and model correction for large-scale urban rail transit hub structures, and provides support for the precise operation and maintenance of such structures. A dynamic performance evaluation technology system for large-scale urban rail transit hub structures is constructed, improving the efficiency of model correction and damage identification. It realizes intelligent monitoring and evaluation throughout the entire process, from local defect identification to underground structure response analysis and cross-space risk early warning, thus enhancing the intelligence level of the monitoring system for large-scale urban rail transit hub structures. Attached Figure Description

[0032] Figure 1 The diagram shown is a flowchart of an intelligent monitoring method for the construction and operation and maintenance of super high-rise structures according to an embodiment of the present invention.

[0033] Figure 2 The diagram shown is an architectural schematic of an intelligent monitoring device for the construction and operation and maintenance of super high-rise structures according to an embodiment of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this invention clearer and more understandable, the invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0035] An embodiment of the present invention provides an intelligent monitoring method for the construction and operation and maintenance of super high-rise structures, as follows: Figure 1 As shown. In Figure 1 In this embodiment, the following are included:

[0036] Step 100: Based on the prior data processed by Bayesian fast modal recognition, a modal recognition model of the complex structure is formed, and the modal recognition model is updated based on the observation data.

[0037] Modal identification is the process of extracting inherent characteristics (frequency, damping ratio, mode shape) from the vibration response of a structure. It is the foundation of structural health monitoring and dynamic analysis. Traditional methods can provide parameter point estimates but cannot quantify uncertainties, while Bayesian methods can provide a complete probability distribution of parameters and assess the reliability of identification. Bayesian fast modal identification treats parameter estimation as a probabilistic inference problem. By integrating prior knowledge and observational data through Bayes' theorem, it can provide parameter uncertainty quantification (posterior variance), naturally integrate multi-source information (different test conditions, sensor data), handle model incompleteness, quantify the impact of model errors, support parameter updates, and adapt to time-varying structural characteristics. It is suitable for stress analysis of complex building structures, providing complete uncertainty quantification while performing efficient calculations, thus providing more reliable technical support for structural health monitoring and safety assessment.

[0038] Step 200: Establish a structural dynamics model based on the modal characteristics, and evaluate the dynamic performance of the complex structure based on the dynamics model.

[0039] By utilizing the modal characteristics of large-scale structural systems determined by Bayesian theory for dynamic performance evaluation, a rigorous probabilistic framework has been established. This framework enables high-precision parameter estimation while quantifying uncertainties, making it particularly suitable for health monitoring and safety assessment of modern large and complex structures (such as long-span bridges, super high-rise buildings, and nuclear power plants). Combining visual monitoring, fiber optic sensing, and traditional vibration measurement, a comprehensive monitoring network is constructed to achieve long-term online monitoring of structures. Time-varying reliability models can be built to address environmental factors (temperature, humidity) and load changes, and Bayesian filtering can be used to update the structural state online and predict the remaining service life.

[0040] Step 300: Extract the surface features of the complex structure based on the dynamic model, train a regional convolutional neural network based on the surface features, map the surface features and dynamic performance through the regional convolutional neural network, and predict structural deformation based on the modal recognition model update.

[0041] While Bayesian modal feature recognition can quantify parameter uncertainty, it cannot guarantee the efficiency of feature extraction and computation. The high-dimensional complexity of sensor signals makes it difficult to capture global characteristics through manually designed features and is susceptible to noise interference. By combining data-driven Bayesian inference with neural networks, a new monitoring method can simultaneously meet the requirements in terms of real-time performance, accuracy, and reliability.

[0042] This invention provides an intelligent monitoring method for the construction and operation of super high-rise structures. This method utilizes Bayesian rapid modal recognition to form a progressively improving yet quantifiable modal recognition model. The structural parameters provided by the modal recognition model determine the structural dynamics model of the current complex structure, and the structural performance is quantified based on observation data. Simultaneously, based on the parameter characteristics of the dynamic model, a regional convolutional neural network is trained by extracting features from the observation data. Leveraging the powerful automatic feature extraction and nonlinear mapping capabilities of the convolutional neural network, high-dimensional modes are extracted from the observation data. This allows for detailed and minute deformation prediction capabilities during the construction and operation monitoring process, improving the early prediction ability of potential hazards.

[0043] In one embodiment of the present invention, step 100, the construction of the modality recognition model includes:

[0044] - Perform signal acquisition and preprocessing

[0045] Sensors are deployed at key locations on the structure to record environmental parameter responses; the sensors include, but are not limited to, various types of signal sensors for pressure, temperature, humidity, vibration and displacement.

[0046] The acquired signal is filtered and denoised to retain the effective mode frequency band. The signal is segmented to eliminate the trend term, the power spectral density matrix is ​​calculated, and the mode frequency is initially located.

[0047] - Construct a parameterized modality recognition model

[0048] Choose the modal order (usually the first 5-10 dominant modes), and define the parameter vector: Determine the prior distribution (e.g., f ~ U(0.1, 50Hz), ζ ~ U(0, 0.1)).

[0049] - Construct the likelihood function

[0050] L(θ|D)=∏p(yi|θ)=∏N(yi|H(θ)ui,σ 2 I)

[0051] H(θ) is the frequency response function matrix, which is obtained through structural dynamics theory.

[0052] - Fast solution for posterior distribution

[0053] Using Laplace optimization or HMC sampling, where Laplace optimization includes:

[0054] Calculate the negative log-likelihood function: -lnL(θ|D);

[0055] Find the minimum point (posterior mode) using Newton's method / coordinate descent method;

[0056] Calculate the Hessian matrix at the mode and inverse it to obtain the covariance matrix.

[0057] HMC sampling includes:

[0058] Construct the Hamiltonian function: H(θ,p)=U(θ)+K(p), (U is the potential energy function=-lnL(θ|D), K is the kinetic energy function);

[0059] Update parameters and momentum using leapfrog integrals, and sample rapidly along the gradient direction;

[0060] The acceptance-rejection criterion ensures that the sample conforms to the target distribution.

[0061] - Uncertainty Quantification and Result Verification

[0062] Calculate the 95% confidence interval of the parameters (mean ± 1.96 × standard deviation), evaluate the rationality of the mode shape using the modal confidence criterion (MAC) (MAC > 0.9 indicates reliability), and screen effective modes using the modal participation factor (MPF) (MPF > 0.1).

[0063] In one embodiment of the present invention, in step 200, kinematic parameters are provided based on modal characteristics, and a dynamic model is constructed using these kinematic parameters. Evaluating the dynamic performance of the complex structure using the dynamic model includes:

[0064] Modal characteristics analysis

[0065] Determine the probability distribution of modal parameters and quantify the variability of parameters, including but not limited to the mean and variance of frequency, damping ratio, and mode shape;

[0066] Analyze the time-varying characteristics of modal parameters under extreme loads, such as typhoons / earthquakes;

[0067] Modal contribution analysis is performed to calculate the contribution ratio of each mode to the structural response, identify the control mode, and assess the structural stiffness distribution and dynamic weak points.

[0068] -Structural status assessment

[0069] Model updates and validations are performed by substituting the identified parameters back into the finite element model to verify the model's accuracy. The model error is quantified by comparing the predicted and measured responses.

[0070] Damage detection was performed, modal parameters at different times were compared, regions of parameter abrupt change were identified, and nonparametric Bayesian clustering was used to quantitatively analyze the changes in modal parameters.

[0071] -Reliability and Risk Assessment

[0072] Probabilistic earthquake demand analysis is conducted, and by combining the Bayesian update model and signal randomness, the response distribution of the structure under different influences is predicted, the probability of structural failure is calculated, and the safety is assessed.

[0073] Time-varying reliability assessment is conducted, taking into account environmental factors (temperature, humidity) and load changes. A time-varying reliability model is constructed, and Bayesian filtering is used to update the structural state online to predict the remaining service life.

[0074] In one embodiment of the present invention, R-CNN is a deep learning object detection framework that has now been extended to the field of structural deformation prediction. Step 300, mapping surface features and dynamic properties through a region convolutional neural network, includes:

[0075] First, image feature maps are extracted using backbone networks such as ResNet;

[0076] Candidate regions (anchor boxes) are generated using a Region Proposal Network (RPN);

[0077] The ROI Align layer extracts candidate region features from the feature map;

[0078] The classification branch determines whether the region is a target feature point;

[0079] The regression branch predicts the precise location (pixel level) of feature points;

[0080] Displacement is calculated by tracking the positional changes of feature points in consecutive frames through time-series analysis.

[0081] An embodiment of the present invention provides an intelligent monitoring device for the construction and operation and maintenance of super high-rise structures, comprising:

[0082] The memory is used to store the program code in the process of the intelligent monitoring method for construction and operation and maintenance of super high-rise structures in the above embodiments.

[0083] The processor is used to execute program code in the intelligent monitoring method for construction and operation and maintenance of super high-rise structures described in the above embodiments.

[0084] The processor can be a DSP (Digital Signal Processor), an FPGA (Field-Programmable Gate Array), an MCU (Microcontroller Unit) system board, a SoC (System on a Chip) system board, or a PLC (Programmable Logic Controller) minimum system including I / O.

[0085] An embodiment of the present invention provides an intelligent monitoring device for the construction and operation and maintenance of super high-rise structures, such as... Figure 2 As shown. In Figure 2 In this embodiment, the following are included:

[0086] The modality recognition module 10 is used to form a modality recognition model of the complex structure based on prior data processed by Bayesian fast modality recognition, and to update the modality recognition model based on observation data;

[0087] Performance evaluation module 20 is used to establish a structural dynamics model based on modal characteristics and evaluate the dynamic performance of the complex structure based on the dynamics model;

[0088] The feature mapping module 30 is used to extract the surface features of the complex structure based on the dynamic model, train a regional convolutional neural network based on the surface features, map the surface features and dynamic performance through the regional convolutional neural network, and predict structural deformation based on the modal recognition model update.

[0089] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An intelligent monitoring method for the construction and operation and maintenance of super high-rise structures, characterized in that, include: Based on Bayesian fast modal recognition, a modal recognition model of the complex structure is formed by processing prior data, and the modal recognition model is updated based on the observation data; A structural dynamics model is established based on modal characteristics, and the dynamic performance of the complex structure is evaluated based on the dynamics model. The surface features of the complex structure are extracted based on the dynamic model. A regional convolutional neural network is trained based on the surface features. The surface features are mapped to the dynamic performance through the regional convolutional neural network. The structural deformation is predicted based on the modal recognition model update.

2. The intelligent monitoring method for construction and operation and maintenance of super high-rise structures as described in claim 1, characterized in that, The construction of the modality recognition model includes: -Perform signal acquisition and preprocessing; - Construct a parameterized modality recognition model; - Fast solution for the posterior distribution; - Uncertainty quantification and result verification.

3. The intelligent monitoring method for construction and operation and maintenance of super high-rise structures as described in claim 1, characterized in that, The evaluation of the dynamic performance of the complex structure using a dynamic model includes: Modal characteristic analysis; -Structural condition assessment; - Reliability and risk assessment.

4. The intelligent monitoring method for construction and operation and maintenance of super high-rise structures as described in claim 1, characterized in that, The mapping of surface features and dynamic properties through a region convolutional neural network includes: First, image feature maps are extracted using backbone networks such as ResNet; Candidate regions are generated through a region proposal network. The ROI Align layer extracts candidate region features from the feature map; The classification branch determines whether the region is a target feature point; Regression branch predicts the precise location of feature points; Displacement is calculated by tracking the positional changes of feature points in consecutive frames through time-series analysis.

5. An intelligent monitoring device for the construction and operation and maintenance of super high-rise structures, characterized in that, include: The memory is used to store the program code in the process of the intelligent monitoring method for construction and operation and maintenance of super high-rise structures as described in any one of claims 1 to 4; A processor for executing the program code.

6. An intelligent monitoring device for the construction and operation and maintenance of super high-rise structures, characterized in that, include: The modality recognition module is used to process prior data using Bayesian fast modality recognition to form a modality recognition model of the complex structure, and to update the modality recognition model based on observation data; The performance evaluation module is used to establish a structural dynamics model based on modal characteristics, and to evaluate the dynamic performance of the complex structure based on the dynamics model. The feature mapping module is used to extract surface features of the complex structure based on the dynamic model, train a regional convolutional neural network based on the surface features, map the surface features and dynamic performance through the regional convolutional neural network, and predict structural deformation based on the modal recognition model update.

Images