Real-time monitoring method and system for large steel structure hoisting accuracy

Abstract

The present application relates to the field of data processing technology, and discloses a method and system for real-time monitoring of the hoisting accuracy of large steel structures. The method includes: scanning and collecting the edge of the steel structure through a distributed sensor network to obtain structural data to obtain the original monitoring data set; extracting and analyzing the transient response characteristics of the structure based on the original monitoring data set to generate a state characteristic matrix; analyzing the coupling relationship between local deformation and overall torsion to obtain a dynamic response prediction matrix; performing spatiotemporal registration and data fusion on the monitoring data to construct a state assessment data set; optimizing and calculating the hoisting parameters to generate a hoisting control data packet; combining historical case data for intelligent reasoning analysis and outputting a control instruction sequence. The present application realizes real-time monitoring of the coupling relationship between local deformation and overall torsional deformation during the hoisting of large steel structures, overcoming the technical defects of the prior art that cannot accurately identify and predict the development trend of structural deformation.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a method and system for real-time monitoring of the hoisting accuracy of large steel structures. Background Art

[0002] Hoisting is a critical step in the construction of large steel structures. Currently, monitoring the hoisting accuracy of large steel structures primarily relies on traditional measurement methods, such as fixed-point measurements using optical measuring equipment like total stations and levels, or local deformation monitoring using strain sensors. Meanwhile, some projects are beginning to adopt computer vision-based monitoring methods, deploying cameras to track the hoisting process in real time and combining image processing technology to analyze structural displacement and deformation. These monitoring methods have accumulated extensive engineering experience and technical data in practical applications.

[0003] However, existing monitoring methods have significant shortcomings. First, traditional optical measurement equipment can only obtain data from discrete measurement points, making it difficult to continuously monitor the overall deformation of the structure. Second, while local strain monitoring can obtain continuous data, it is difficult to reflect the overall deformation characteristics of the structure. Third, computer vision-based monitoring methods are easily affected by environmental factors such as light changes and occlusion, which can affect monitoring accuracy. Most importantly, existing monitoring methods generally lack the ability to analyze the coupling relationship between local deformation and overall torsional deformation of the structure, making it difficult to accurately predict and control the deformation development trend during the lifting process. Summary of the Invention

[0004] The present application provides a real-time monitoring method and system for the hoisting accuracy of large steel structures, which is used to realize real-time monitoring of the coupling relationship between local deformation and overall torsional deformation during the hoisting process of large steel structures, overcoming the technical defects of the existing technology that cannot accurately identify and predict the development trend of structural deformation.

[0005] In a first aspect, the present application provides a method for real-time monitoring of the hoisting accuracy of large steel structures, the method comprising: scanning and collecting the edge of the steel structure through a distributed sensor network to obtain structural angular velocity data, strain data and load distribution data to obtain an original monitoring data set; extracting and analyzing the transient response characteristics of the structure based on the original monitoring data set to generate a structural state characteristic matrix containing local deformation data and overall torsion data; calculating and analyzing the coupling relationship between local deformation and overall torsion based on the structural state characteristic matrix to obtain torsional deformation evolution data and stress distribution prediction data, and integrating them to obtain a dynamic response prediction matrix; based on the dynamic response prediction matrix, performing spatiotemporal registration and data fusion on the monitoring data collected in real time to construct a state assessment data set containing torsional deformation indicators and stress distribution indicators; performing dynamic optimization calculation on the hoisting parameters based on the state assessment data set to generate a hoisting control data packet containing hoisting point position parameters, hoisting torque parameters and compensation control parameters; utilizing the hoisting control data packet and combining it with historical case data for intelligent reasoning analysis to output a real-time control instruction sequence to achieve precision monitoring and control of the hoisting process.

[0006] In a second aspect, the present application provides a large-scale steel structure hoisting accuracy real-time monitoring system, the large-scale steel structure hoisting accuracy real-time monitoring system comprising:

[0007] The acquisition module is used to scan and collect the edge of the steel structure through a distributed sensor network to obtain the structural angular velocity data, strain data and load distribution data to obtain the original monitoring data set;

[0008] An extraction module is used to extract and analyze the transient response characteristics of the structure based on the original monitoring data set, and generate a structural state characteristic matrix including local deformation data and overall torsion data;

[0009] A coupling module is used to calculate and analyze the coupling relationship between local deformation and overall torsion based on the structural state characteristic matrix, obtain torsional deformation evolution data and stress distribution prediction data, and integrate them to obtain a dynamic response prediction matrix;

[0010] A fusion module is used to perform spatiotemporal registration and data fusion on the monitoring data collected in real time based on the dynamic response prediction matrix to construct a state assessment data set including torsional deformation indicators and stress distribution indicators;

[0011] an optimization module, configured to perform dynamic optimization calculations on the hoisting parameters based on the state evaluation data set, and generate a hoisting control data packet including hoisting point position parameters, hoisting torque parameters, and compensation control parameters;

[0012] The control module is used to utilize the hoisting control data packet in combination with historical case data to perform intelligent reasoning analysis, output a real-time control instruction sequence, and realize precision monitoring and control of the hoisting process.

[0013] In the technical solution provided by the present application, the edge of the steel structure is scanned and collected through a distributed sensor network, thereby realizing multi-source heterogeneous data collection of structural angular velocity data, strain data and load distribution data, solving the problem that a traditional single sensor is difficult to comprehensively monitor the structural state; by extracting and analyzing the transient response characteristics of the original monitoring data set, a structural state characteristic matrix containing local deformation data and overall torsion data is generated, thereby realizing a comprehensive characterization of the structural deformation characteristics; a coupling relationship calculation and analysis method is used to conduct an in-depth analysis of the correlation between local deformation and overall torsion, obtain torsional deformation evolution data and stress distribution prediction data, and integrate these data into a dynamic The response prediction matrix provides a reliable basis for structural deformation prediction; based on the dynamic response prediction matrix, the monitoring data collected in real time are temporally and spatially aligned and fused, and a state evaluation data set including torsional deformation indicators and stress distribution indicators is constructed, thereby improving the temporal and spatial consistency of the monitoring data; according to the state evaluation data set, a dynamic optimization algorithm is used to calculate the lifting parameters in real time, and a lifting control data packet including lifting point position parameters, lifting torque parameters and compensation control parameters is generated, making the generation of control parameters more scientific; by combining historical case data for intelligent reasoning analysis, a real-time control instruction sequence is output, thereby realizing closed-loop optimization of precision monitoring and control of the lifting process. In terms of the application of artificial intelligence algorithms, the present invention innovatively combines deep learning with traditional mechanical models, establishes a prediction model that is more in line with engineering practice through mining and learning of historical data, and adopts intelligent reasoning algorithms to dynamically optimize the control strategy, which significantly improves the accuracy and real-time performance of the control instructions, making the entire monitoring and control scheme more adaptable and robust. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 This is a schematic diagram of an embodiment of a method for real-time monitoring of the hoisting accuracy of a large steel structure in an embodiment of the present application;

[0016] Figure 2 A timing diagram for extracting and analyzing structural transient response characteristics in an embodiment of the present application;

[0017] Figure 3This is a schematic diagram of an embodiment of a real-time monitoring system for the hoisting accuracy of a large steel structure in an embodiment of the present application. DETAILED DESCRIPTION

[0018] The embodiments of the present application provide a method and system for real-time monitoring of the hoisting accuracy of a large steel structure. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0019] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In one embodiment of the present application, a method for real-time monitoring of the hoisting accuracy of a large steel structure includes:

[0020] Step S101: Scan and collect data on the edge of the steel structure through a distributed sensor network to obtain structural angular velocity data, strain data, and load distribution data to obtain an original monitoring data set;

[0021] Step S102: extracting and analyzing the transient response characteristics of the structure based on the original monitoring data set to generate a structural state characteristic matrix containing local deformation data and overall torsion data;

[0022] Step S103: Based on the structural state characteristic matrix, the coupling relationship between local deformation and overall torsion is calculated and analyzed to obtain torsional deformation evolution data and stress distribution prediction data, and the dynamic response prediction matrix is obtained by integrating them;

[0023] Step S104: Based on the dynamic response prediction matrix, the real-time collected monitoring data is subjected to spatiotemporal registration and data fusion to construct a state assessment data set including torsional deformation indicators and stress distribution indicators;

[0024] Step S105: Perform dynamic optimization calculation on the hoisting parameters according to the state evaluation data set to generate a hoisting control data packet including hoisting point position parameters, hoisting torque parameters, and compensation control parameters;

[0025] Step S106: Utilize the hoisting control data packet and combine it with historical case data to perform intelligent reasoning analysis and output a real-time control instruction sequence to achieve precision monitoring and control of the hoisting process.

[0026] It is understandable that the execution subject of this application can be a large steel structure hoisting accuracy real-time monitoring system, or a terminal or a server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0027] Specifically, the real-time monitoring method for the hoisting accuracy of large steel structures focuses on the accuracy control of large steel components during the hoisting process. In its implementation, data is collected through a distributed sensing network consisting of fiber grating (FBG) sensors, a gyroscope array, and dynamic strain sensors. Fiber grating (FBG) sensors are arranged in a spiral pattern at 0.5-meter intervals along the edge of the steel structure, sampling strain at the structure's edge at high frequency. Initial strain signals are obtained through wavelength demodulation analysis, then decomposed using a wavelet transform for multi-scale analysis. The resulting frequency band signals are thresholded and reconstructed to generate edge strain data. The gyroscope array uses a sampling frequency of 200 Hz to acquire three-axis angular velocity at key nodes. The raw data is band-filtered to remove high-frequency noise and low-frequency drift, and then state estimated using a Kalman filter to obtain angular velocity data. Dynamic strain sensors monitor the load points of the hoisting equipment. The collected load time series data is converted to the frequency domain using a fast Fourier transform. Interference frequency components are identified and eliminated, and then inversely transformed to obtain load distribution data.

[0028] When extracting transient response features from the original monitoring data set, time domain segmentation is used to extract data segments within key time windows, generating time series feature data. Spatial correlation analysis is performed on the data at each measurement point, and the correlation coefficient matrix between measurement points is calculated to select key measurement point combinations. The data from these key measurement point combinations is then decomposed into multiple dimensions to extract displacement, velocity, and acceleration components, generating local response features. Data reconstruction is used to spatially map the local response features, establish a nodal displacement field, and obtain local deformation data. Based on the local deformation data, an overall coordination analysis is performed to calculate deformation gradients and deformation rates in each region, and construct global torsion data. The local deformation data and global torsion data are organized and arranged according to their spatiotemporal correspondence to generate a structural state feature matrix. When performing coupling relationship analysis based on the structural state feature matrix, the matrix is segmented into time series, and local deformation and global torsion data segments within each time window are extracted. Spatial decoupling analysis is performed on the local deformation data segments, and the relative displacement and deformation rate of each monitoring point are calculated to generate local deformation feature vectors. The torsion angle distribution of the global torsion data segments is calculated, and the torsion center trajectory is established to generate the global torsion feature vector. The two eigenvectors are coupled and measured using tensor operations to establish a stress-deformation mapping relationship, yielding torsional deformation evolution data. This stress-deformation mapping relationship is modified based on the local stress concentration factor to construct a stress distribution field and generate stress distribution prediction data. The torsional deformation evolution data and stress distribution prediction data are combined and correlated according to the principle of spatial and temporal correspondence to form a dynamic response prediction matrix.

[0029] When performing data fusion based on the dynamic response prediction matrix, the matrix is time-series sampled and reference time information is extracted to obtain a time reference sequence. The real-time monitoring data is timestamped and mapped to the time reference sequence to obtain time-registered data. The time-registered data is spatially mapped using the monitoring point coordinates, and the relative position deviation of each node is calculated to obtain spatially registered data. Correlation analysis is performed between the time-registered and spatially registered data and the dynamic response prediction matrix to calculate the deviation distribution and generate a torsional deformation index. The monitoring area is meshed based on the torsional deformation index, and the stress distribution characteristics at the mesh nodes are calculated to construct a stress distribution index. The torsional deformation index and stress distribution index are organized and organized according to topological relationships to obtain a multidimensional feature data table, which forms the state assessment dataset. When performing parameter optimization based on the state assessment dataset, a gradient analysis of the torsional deformation index in the dataset is performed to calculate the spatial distribution of deformation-sensitive areas of the structure and obtain deformation sensitivity distribution data. Based on the deformation sensitivity distribution data, stress analysis is performed on the structural stress points, and the stress concentration at each node is calculated to obtain stress distribution weight data. The deformation sensitivity distribution data and stress distribution weight data are numerically superimposed to evaluate and analyze the load state of the lifting points, generating the lifting point position parameters. A torque balance calculation is performed based on the lifting point position parameters to establish the force balance relationship between each lifting point and construct the lifting torque parameters. The lifting torque parameters are used to predict the deformation trend during the lifting process, calculate the displacement compensation for each key point, and obtain the compensation control parameters. The lifting point position parameters, lifting torque parameters, and compensation control parameters are integrated according to control priority to generate a lifting control data package.

[0030] Perform feature analysis on the hoisting control data packet, extract the numerical range of each parameter, and obtain control parameter feature data. Extract control parameter records under similar working conditions from historical case data, establish parameter correspondence, and generate historical parameter mapping data. Calculate the similarity between the control parameter feature data and the historical parameter mapping data, screen out historical control strategies with high matching, and build a control strategy database. Perform time series analysis on the control parameters in the control strategy database, calculate the change trend of each control parameter, and obtain the parameter evolution sequence. Based on the parameter evolution sequence, perform predictive reasoning on the control parameters, calculate the control parameter value at the next moment, and generate control parameter update data. Organize the control parameter update data in a time series according to the control execution order to generate a real-time control instruction sequence.

[0031] For example, during the hoisting of large steel structures, the initial strain signals collected by fiber Bragg grating sensors are wavelength-demodulated to obtain the strain distribution. The angular velocity data collected by the gyroscope array is filtered to reflect the structure's tilt. The load data collected by the dynamic strain sensor is analyzed in the frequency domain to reveal the stress conditions at each lifting point. Feature extraction and analysis of these data yield a state characteristic matrix reflecting local deformation and overall torsion. Coupled analysis of this characteristic matrix predicts the deformation trend and stress distribution of the structure. Based on these predictions and combined with real-time monitoring data, a state assessment is performed to generate optimized hoisting control parameters, including lifting point location, hoisting torque, and compensation control variables. Drawing on experience from successful historical cases, a control command sequence is generated to guide the precise execution of the hoisting operation.

[0032] In the embodiment of the present application, the edge of the steel structure is scanned and collected through a distributed sensor network, thereby realizing multi-source heterogeneous data collection of structural angular velocity data, strain data and load distribution data, solving the problem that a traditional single sensor is difficult to fully monitor the structural state; by extracting and analyzing the transient response characteristics of the original monitoring data set, a structural state characteristic matrix containing local deformation data and overall torsion data is generated, thereby realizing a comprehensive characterization of the structural deformation characteristics; a coupling relationship calculation and analysis method is used to conduct an in-depth analysis of the correlation between local deformation and overall torsion, obtain torsional deformation evolution data and stress distribution prediction data, and integrate these data to form a dynamic response The prediction matrix provides a reliable basis for structural deformation prediction; based on the dynamic response prediction matrix, the monitoring data collected in real time are temporally and spatially aligned and data fused, and a state evaluation data set including torsional deformation indicators and stress distribution indicators is constructed, thereby improving the temporal and spatial consistency of the monitoring data; according to the state evaluation data set, a dynamic optimization algorithm is used to calculate the lifting parameters in real time, and a lifting control data packet including lifting point position parameters, lifting torque parameters and compensation control parameters is generated, making the generation of control parameters more scientific; by combining historical case data for intelligent reasoning analysis, a real-time control instruction sequence is output, thereby realizing closed-loop optimization of precision monitoring and control of the lifting process. In terms of the application of artificial intelligence algorithms, the present invention innovatively combines deep learning with traditional mechanical models, establishes a prediction model that is more in line with engineering practice through mining and learning of historical data, and adopts intelligent reasoning algorithms to dynamically optimize the control strategy, which significantly improves the accuracy and real-time performance of the control instructions, making the entire monitoring and control scheme more adaptable and robust.

[0033] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0034] (1) Fiber Bragg grating sensors are spirally arranged along the edge of the steel structure at a spacing of 0.5 meters. High-frequency sampling is performed on the edge of the structure, and wavelength demodulation analysis is performed at each sampling point to obtain the initial strain signal. The initial strain signal is then decomposed into multiple scales using wavelet transform. The decomposed frequency band signals are threshold processed and reconstructed to generate the structural edge strain data.

[0035] (2) The gyroscope array continuously samples the key nodes of the steel structure at a sampling frequency of 200 Hz to collect the original data of the three-axis angular velocity. The raw data is filtered by a bandpass filter to remove high-frequency noise and low-frequency drift. The Kalman filter algorithm is then used to perform state estimation and prediction on the processed data to obtain the structural angular velocity data.

[0036] (3) Dynamic strain sensors are deployed at the load points of the lifting equipment to collect load time series data. The time domain data is converted to the frequency domain through fast Fourier transform, the interference frequency components are identified and eliminated, and the signal is reconstructed using inverse Fourier transform to obtain load distribution data with high signal-to-noise ratio;

[0037] (4) Temperature compensation and linear calibration are performed on the edge strain data of the structure to remove the influence of environmental factors. The strain change rate is then calculated using the sliding window method. Feature extraction is performed in combination with the strain amplitude to generate edge strain feature data.

[0038] (5) The structural angular velocity data is processed by principal component analysis to reduce the dimension, extract the main change characteristics, and then the data is resampled and synchronized by spline interpolation algorithm to obtain angular velocity characteristic data;

[0039] (6) Normalize the maximum and minimum values of the load distribution data, perform standardization based on the statistical characteristics of historical data, and then use the moving average method to eliminate short-term fluctuations to construct load characteristic data;

[0040] (7) The edge strain characteristic data, angular velocity characteristic data and load characteristic data are aligned according to the time reference, and data association is established through timestamp matching. The aligned data are then interpolated and supplemented and outlier processing is performed. The three types of characteristic data are integrated into a structural state description matrix through the data fusion algorithm to generate the original monitoring data set.

[0041] It should be noted that fiber Bragg grating sensors (FBGs) are sensors that use Bragg gratings (FBGs) inscribed within the core of an optical fiber. When subjected to external strain or temperature changes, the grating period shifts, causing the wavelength of the reflected light to change. These sensors are arranged in a spiral pattern along the edge of the steel structure at a fixed spacing of 0.5 meters, creating a continuous monitoring network. Each FBG sensing point corresponds to a specific central wavelength. A demodulator detects changes in the reflected light wavelength in real time to obtain the initial strain signal. These signals are decomposed using a wavelet transform with multi-scale decomposition, breaking them down into wavelet coefficients in different frequency bands. Appropriate thresholds are set for these coefficients to eliminate noise. Finally, the signals are reconstructed using an inverse wavelet transform to obtain strain data at the structure's edge. Gyroscope arrays are deployed at key structural nodes to monitor angular velocity. Each gyroscope simultaneously measures angular velocity in the X, Y, and Z directions, with a sampling frequency of 200 Hz to accurately capture the structure's transient response. The collected raw data is processed through a bandpass filter to remove high-frequency noise above 100 Hz and low-frequency drift signals below 0.1 Hz. The Kalman filter algorithm is then used to perform state estimation on the filtered data. The Kalman filter continuously optimizes the state estimation value through the iteration of prediction step and update step to obtain the structural angular velocity data.

[0042] Dynamic strain sensors are installed at the main load-bearing points of the lifting equipment. These sensors continuously collect load time-series data. The collected data is converted to the frequency domain using a fast Fourier transform, where various frequency components can be clearly identified. Interference frequency components are identified by analyzing the spectral characteristics, and the amplitudes corresponding to these frequencies are set to zero. The signal is then converted back to the time domain using an inverse Fourier transform to obtain the load distribution data after interference removal. The acquired edge strain data is temperature compensated and linearly calibrated. Based on the real-time temperature values measured by the temperature sensors, the raw strain values are corrected using the temperature-strain relationship. The strain values are then corrected using linear calibration coefficients determined through calibration experiments to eliminate systematic errors. The rate of strain change is calculated using a sliding window method. The window length is determined by the data sampling rate, and the rate of strain change is obtained by taking the difference between the strain values within the window. By combining the absolute strain value and rate of change information, characteristic parameters are extracted to generate edge strain feature data.

[0043] Structural angular velocity data contains a large amount of redundant information, so principal component analysis (PCA) is used to reduce its dimensionality. The data is standardized, the covariance matrix is calculated, the eigenvalues and eigenvectors are solved, and the principal components with the largest contributions are selected as key features. The reduced dimensionality data is then resampled using a cubic spline interpolation algorithm to even out the sampling intervals. The load distribution data is normalized, mapping the data range to the [0, 1] interval. The mean and standard deviation are calculated based on historical data, and the data is standardized to conform to a standard normal distribution. The standardized data is smoothed using a moving average method to remove the effects of short-term random fluctuations.

[0044] The three types of characteristic data (edge strain, angular velocity, and load) are time-synchronized. A time base is determined, and the data are aligned based on their timestamps. For data points with inconsistent sampling times, missing values are filled using interpolation. Outliers are detected and processed to ensure data continuity and reliability. Finally, a data fusion algorithm is used to integrate the three types of characteristic data into a matrix based on their physical meaning and spatial relationships, resulting in the original monitoring dataset.

[0045] For example, during the hoisting of a 30-meter-long steel beam, fiber Bragg grating sensors were deployed every 0.5 meters along the beam's edge, gyroscopes were placed at the beam's four endpoints, and dynamic strain sensors were installed at the four lifting points. As the beam began to be hoisted, the fiber optic sensor network detected the strain distribution along the edge. After temperature compensation and wavelet noise reduction, the real-time deformation state of the structure was revealed. The angular velocity data detected by the gyroscopes was filtered and state estimated to reveal the spatial posture changes of the structure. The load data at the lifting points was analyzed and reconstructed in the frequency domain to reveal the force balance at each lifting point. Feature extraction and fusion of these data were then performed to generate the monitoring dataset.

[0046] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0047] (1) The original monitoring data set is intercepted by time domain segmentation, and the data segments within the key time window are selected to generate time series feature data;

[0048] (2) Perform spatial correlation analysis on the data of each measuring point in the time series feature data, calculate the correlation coefficient matrix between the measuring points, and screen out the key measuring point combination;

[0049] (3) Decompose the data of the key measurement point combination into multiple dimensions, extract the displacement component, velocity component and acceleration component, and obtain the local response characteristics;

[0050] (4) Through data reconstruction, the local response characteristics are spatially mapped, the node displacement field is established, and the local deformation data is obtained;

[0051] (5) Perform overall coordination analysis based on local deformation data, calculate the deformation gradient and deformation rate of each region, and construct overall torsion data;

[0052] (6) The local deformation data and the overall torsion data are organized and arranged according to the time-space correspondence to generate a structural state characteristic matrix.

[0053] Specifically, if Figure 2 As shown in the figure, in an embodiment of the present application, a time series diagram for extracting and analyzing the transient response characteristics of the structure is shown. The original monitoring data set is segmented in the time domain to obtain time series feature data; then, the time series feature data is subjected to spatial correlation analysis to screen out key measurement point combinations; the key measurement point combinations are subjected to multi-dimensional decomposition to extract local response features; the local response features are converted into local deformation data through spatial mapping reconstruction; the local deformation data is subjected to overall coordination analysis to construct overall torsion data; finally, the local deformation data and the overall torsion data are organized and arranged according to the time-space correspondence to generate a structural state feature matrix.

[0054] Time domain segmentation is the process of dividing continuous time series data into several data segments of specific time lengths. In practice, fixed-length time windows are selected for data capture. The time window length is determined based on the dynamic characteristics of the structure, generally covering a complete vibration cycle. A sliding window technique is used during data capture, with some overlap between adjacent windows to ensure data continuity. Data within each time window is extracted to produce independent data segments. These segments contain the response characteristics of the structure over different time periods, forming the time series feature data.

[0055] Spatial correlation analysis is performed on the obtained time series characteristic data. The purpose of spatial correlation analysis is to identify the degree of correlation between measurement points and to identify key measurement points that are representative of the structural deformation characteristics. Specifically, a correlation coefficient matrix is calculated between different measurement points. The correlation coefficient matrix reflects the degree of linear correlation between the data at each measurement point. The Pearson correlation coefficient method is used to calculate the correlation coefficient. For any two measurement point data series, their covariance is calculated and divided by the product of their respective standard deviations to obtain the correlation coefficient. The correlation coefficient ranges from -1 to 1, with a larger absolute value indicating a stronger correlation. By setting a correlation coefficient threshold, measurement point combinations with significant correlations are screened. These measurement point combinations can better reflect the overall deformation characteristics of the structure. The data from the selected key measurement point combinations are subjected to multidimensional decomposition to extract component information of different physical quantities. Multidimensional decomposition refers to the process of converting the displacement time history data of the measurement points into velocity and acceleration information through numerical differencing. The displacement data is numerically differentiated, and the velocity component is calculated using the central difference method. The acceleration component is then differentiated from the velocity component. The displacement component reflects the deformation state of the structure, the velocity component reflects the development trend of the deformation, and the acceleration component reflects the dynamic response characteristics of the structure. These three types of component data complement each other and together constitute complete information describing the local response characteristics of the structure.

[0056] The local response characteristic data is reconstructed into a continuous nodal displacement field through a spatial mapping method. This spatial mapping process employs shape function interpolation, using data from a limited number of key measurement points to estimate the continuous displacement distribution across the entire structural domain. Shape functions are a set of interpolation functions defined based on spatial coordinates that ensure the continuity and smoothness of the interpolation results. By combining the position coordinates and response data of key measurement points with shape function interpolation to calculate displacement values at any location, a nodal displacement field is constructed, resulting in local deformation data.

[0057] Based on local deformation data, global compatibility analysis is performed. The core of this analysis is the calculation of deformation gradients and deformation rates. The deformation gradient refers to the rate of change of the displacement field in space and is obtained by calculating the displacement differentials between adjacent nodes. The deformation rate is the rate of change of displacement over time and is obtained by calculating the displacement differentials at consecutive moments. These two parameters together reflect the torsional deformation characteristics of the structure. By establishing the trajectory of the torsion center and the distribution of torsion angles, global torsion data is constructed.

[0058] Local deformation data and global torsion data are organized and arranged according to a spatiotemporal correspondence. This refers to the mapping of data between the temporal and spatial dimensions, ensuring that different types of data are synchronized in time and spatially aligned. All data is organized according to the format to create a multidimensional array structure, where different dimensions correspond to time, spatial location, and physical quantity type, ultimately generating a structural state feature matrix.

[0059] Take the hoisting process of a large steel structure as an example: the original monitoring data is divided into 10-second time windows with a 5-second overlap between the windows to obtain a series of time series data fragments. The correlation coefficients between the measuring points are calculated for these data fragments, and it is found that the measuring points located at the four corners of the structure have a strong correlation. These measuring points are combined as key measuring points. The data of the key measuring points are differentially calculated to obtain the response characteristics in three dimensions: displacement, velocity, and acceleration. The discrete measuring point data are reconstructed into a continuous displacement field through shape function interpolation, and the deformation distribution of each area is calculated. The spatial gradient and time change rate of the deformation field are analyzed to identify the torsional center position and torsional angle change law of the structure. Finally, all the data are integrated into a matrix. The rows of the matrix represent different moments, the columns represent different spatial positions, and the depth direction represents different physical quantities to obtain a description of the state characteristics.

[0060] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0061] (1) Segment the structural state characteristic matrix according to the time series, divide the continuous data stream into multiple time windows, and extract the local deformation and overall torsion data segments in each window;

[0062] (2) Perform spatial decoupling analysis on the local deformation data segments, calculate the relative displacement and deformation rate of each monitoring point, and generate the local deformation feature vector;

[0063] (3) Calculate the torsion angle distribution for the overall torsion data segment, establish the torsion center trajectory, and obtain the overall torsion eigenvector;

[0064] (4) The local deformation eigenvector and the global torsion eigenvector are coupled and measured by tensor operations to establish a stress-deformation mapping relationship and obtain the torsional deformation evolution data;

[0065] (5) Based on the local stress concentration coefficient, the stress-deformation mapping relationship is corrected and calculated, the stress distribution field is constructed, and stress distribution prediction data is generated;

[0066] (6) The torsional deformation evolution data and stress distribution prediction data are combined and correlated according to the principle of time-space correspondence to establish a time-varying characteristic mapping matrix to form a dynamic response prediction matrix.

[0067] Specifically, the structural state characteristic matrix is segmented in the time domain, using a sliding window method to divide the continuous data stream into multiple time windows. The length of each time window is determined by the structural characteristic frequency, typically 3-5 times the structure's primary vibration period. Adjacent windows maintain a 50% overlap to ensure data continuity. Within each window, data segments for local deformation and global torsion are extracted. These data segments contain information about the structural deformation characteristics within a specific time period.

[0068] The spatial decoupling analysis of local deformation data segments involves calculating relative displacement and deformation rate. The formula for calculating relative displacement is:

[0069] ;

[0070] in, Represents the relative displacement of the monitoring point (i, j), is the weight coefficient of the kth measurement point, is the measured value of the current position, is the measured value at the reference position, is the spatial attenuation coefficient, and n is the number of reference points.

[0071] For the calculation of the distribution of the overall torsion angle and the establishment of the torsion center trajectory, the following formula is used:

[0072] ;

[0073] in, represents the torsion angle in the xy plane, is the measurement point weight matrix, and is the coordinate of the measuring point, is the initial phase angle of the measuring point, M and N are the number of measuring points in the x and y directions respectively.

[0074] The coupling metric between the local deformation eigenvector and the global torsion eigenvector uses tensor operations, and its mathematical expression is:

[0075] ;

[0076] in, is the coupled metric tensor, is the weight tensor, is the local deformation eigenvector, is the overall torsion eigenvector, is the coupling coefficient tensor, and I, J, and K represent the sizes of the three dimensions respectively.

[0077] The results obtained from the above calculations are combined and the stress-deformation mapping relationship is corrected based on the local stress concentration factor. The local stress concentration factor is a dimensionless parameter that reflects the uneven stress distribution caused by changes in the local geometry of the structure. This factor is used to correct the stress distribution and obtain stress distribution prediction data. The torsional deformation evolution data and the stress distribution prediction data are combined and correlated according to the principle of time-space correspondence. The time-space correspondence principle requires that all data must strictly correspond in time and space. By establishing a time-space coordinate system, different types of data are integrated into the same matrix framework.

[0078] Taking the hoisting process of a 30-meter-span steel structure truss as an example, the collected data is segmented into 5-second time windows, with an overlap of 2.5 seconds between windows. In each time window, the local deformation data and overall torsion data at the key nodes are extracted. For the local deformation data, the relative displacement is obtained by calculating the displacement difference of each monitoring point relative to the reference point; the deformation rate is obtained by calculating the rate of change of displacement over time. For the overall torsion data, the torsion angle distribution is calculated by analyzing the displacement vector direction of each measuring point, and the motion trajectory of the torsion center is obtained by fitting the least squares method. The two sets of data are coupled and analyzed through tensor operations to establish a mapping relationship between local deformation and overall torsion. According to the stress concentration coefficient determined by the geometric characteristics of the structure, the stress distribution is corrected, and finally the dynamic response prediction matrix is obtained.

[0079] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0080] (1) Align the dynamic response prediction matrix through time series sampling, extract the reference time information of each monitoring point, and generate a time reference sequence;

[0081] (2) Compare the timestamps of the real-time collected monitoring data, establish a mapping relationship with the time reference sequence, and obtain time-registered data;

[0082] (3) Use the monitoring point coordinate information to perform spatial mapping on the temporal registration data, calculate the relative position deviation of each node, and obtain the spatial registration data;

[0083] (4) Perform correlation analysis on the temporal registration data and spatial registration data with the dynamic response prediction matrix, calculate the deviation distribution, and generate the torsional deformation index;

[0084] (5) Divide the monitoring area into grids according to the torsional deformation index, calculate the stress distribution characteristics at the grid nodes, and construct the stress distribution index;

[0085] (6) The torsional deformation index and stress distribution index are organized and sorted according to the topological relationship to obtain a multidimensional feature data table, which constitutes the state assessment data set.

[0086] Specifically, in the real-time monitoring of large steel structure hoisting accuracy, time series sampling is a method of processing data in the time domain. When aligning the data of the dynamic response prediction matrix, a reference time point is determined. Starting from this time point, data from each monitoring point is extracted according to a fixed sampling interval. The following formula is used in the time series sampling process:

[0087] ;

[0088] in, is the data matrix after sampling, is the sampling weight coefficient, and are upsampling and downsampling operators respectively, is the time synchronization factor, M and N are the number of sampling levels. This process generates a time reference sequence.

[0089] When comparing timestamps of real-time monitoring data, a time correspondence needs to be established. Real-time monitoring data usually contains data streams with different sampling frequencies, which need to be unified to the same time base through difference calculation. The specific calculation formula is:

[0090] ;

[0091] in, is the data matrix after time registration, is the registration weight coefficient, and are the forward and backward temporal registration operators, respectively, is the temporal alignment factor, A and B are the number of registration levels.

[0092] When using the monitoring point coordinate information to perform spatial mapping on the time registration data, the following calculation formula is used:

[0093] ;

[0094] in, is the spatial registration data matrix, is the spatial mapping weight, and are the mapping operators for the xy plane and yz plane respectively, is the spatial correction factor, and I and J are mapping scale parameters.

[0095] During the correlation analysis, the deviation distribution between the temporally and spatially registered data and the dynamic response prediction matrix is calculated. This process involves comparing and matching multiple data sets. By calculating correlation coefficients and deviation values, torsional deformation indicators are generated. Based on these indicators, the monitoring area is meshed, with the mesh size and density determined based on the structural characteristics. At each mesh node, stress distribution characteristics, including principal stress, shear stress, and equivalent stress, are calculated to construct a stress distribution index.

[0096] Torsional deformation and stress distribution indicators are organized according to topological relationships. Topological relationships reflect the spatial connections between various parts of the structure. By establishing a data structure, different types of indicator data are integrated into a multidimensional feature data table to form a condition assessment data set.

[0097] For example: During the hoisting process, the dynamic response prediction matrix is sampled at a sampling frequency of 100Hz, and the response data of each monitoring point at each moment is extracted. Among the real-time monitoring data, the strain data sampling frequency is 200Hz, and the angular velocity data sampling frequency is 50Hz. These data are unified to a time base of 100Hz through time alignment. These unified data are mapped according to the spatial coordinate information of the monitoring point, and the deviation value relative to the design position is calculated. These deviation data are compared with the prediction matrix to calculate the torsional deformation index. Based on the torsional deformation index, the monitoring area is divided into several grid units, and the stress distribution characteristics are calculated at each grid node. These index data are finally integrated into a multidimensional data table to obtain a state assessment data set that reflects the real-time state information of the structure during the hoisting process.

[0098] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0099] (1) Perform gradient analysis on the torsional deformation index in the state assessment data set, calculate the spatial distribution of the structural deformation sensitive area, and obtain the deformation sensitivity distribution data;

[0100] (2) Based on the deformation sensitivity distribution data, the stress analysis of the structural stress points is performed, the stress concentration degree of each node is calculated, and the stress distribution weight data is obtained;

[0101] (3) Numerically superimpose the deformation sensitivity distribution data and the stress distribution weight data, evaluate and analyze the stress state of the hanging point, and generate the hanging point position parameters;

[0102] (4) Perform moment balance calculation based on the lifting point position parameters, establish the force balance relationship between each lifting point, and construct the lifting moment parameters;

[0103] (5) Predict the deformation trend during the lifting process based on the lifting torque parameters, calculate the displacement compensation of each key point, and obtain the compensation control parameters;

[0104] (6) The lifting point position parameters, lifting torque parameters and compensation control parameters are integrated according to the control priority to generate a lifting control data package.

[0105] Specifically, gradient analysis is achieved by calculating the deformation difference between adjacent monitoring points in space. The specific calculation formula is:

[0106] ;

[0107] in, represents the deformation gradient tensor, is the gradient weight coefficient, and are the gradient operators of the jk plane and the kl plane respectively, is the spatial sensitivity factor, and D and E are the dimensional parameters for gradient calculation. This calculation process performs a scanning analysis on the entire structural area, determines the sensitive areas with large deformation gradient values, and obtains deformation sensitivity distribution data.

[0108] Based on the deformation sensitivity distribution data, the stress state analysis of the structural stress point is carried out, and the calculation formula is:

[0109] ;

[0110] in, is the stress distribution weight tensor, is the stress weight coefficient, and are the stress operators of the uv plane and vw plane respectively, is the stress concentration factor, and F and G are scale parameters for stress calculation. This calculation takes into account the geometric characteristics and material properties of the structure, and reflects the effect of local geometric shape changes on stress distribution through the stress concentration factor.

[0111] When superimposing the deformation sensitivity and stress distribution weight data, the following formula is used:

[0112] ;

[0113] in, is the hanging point position parameter tensor, is the position weight coefficient, and are the position operators of the ab plane and the bc plane respectively, is the position correction factor, H and I are the spatial dimensions of the position parameters. This calculation process comprehensively evaluates the suitability of each candidate hanging point location through a weighted average method.

[0114] For the calculation of the moment balance at the lifting point, the following formula is used:

[0115] ;

[0116] in, is the lifting moment parameter tensor, is the moment weight coefficient, and are the moment operators of the rs plane and st plane respectively, is the moment balance factor, M and N are the direction parameters of the moment calculation. This calculation ensures that all suspension points are in a state of force equilibrium.

[0117] The displacement compensation amount is calculated using:

[0118] ;

[0119] in, To compensate the control parameter tensor, is the compensation weight coefficient, and are the compensation operators for the xy plane and yz plane respectively, is the compensation correction factor, P and Q are the level parameters of the compensation calculation.

[0120] The actual application process of these formulas is as follows: For a large steel structure crane beam, through the gradient analysis of the torsional deformation index, it was found that there were large deformation gradients at the ends and mid-span positions of the beam. These positions are marked as deformation-sensitive areas, specifically, the gradient value at the end position is 2-3 times that of the middle position. In the stress analysis at the node, the stress concentration factor at the connection is significantly higher than that at other positions. After superimposing these two types of data, four optimal lifting point positions are selected. These positions avoid high stress areas and ensure overall force balance. Through moment balance calculation, the specific lifting torque value required for each lifting point is obtained, and the structural deformation trend under these torques is predicted. Based on the prediction results, the displacement compensation required for each key point is calculated. These parameters are integrated in the order of priority of stability influence, accuracy influence and auxiliary control to obtain the lifting control data package.

[0121] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0122] (1) Perform feature analysis on the lifting control data packet, extract the numerical range of the lifting point position parameters, lifting torque parameters and compensation control parameters, and obtain the control parameter feature data;

[0123] (2) Extract control parameter records under similar working conditions from historical case data, establish parameter correspondence, and generate historical parameter mapping data;

[0124] (3) Calculate the similarity between the control parameter feature data and the historical parameter mapping data, select the historical control strategies with high matching degree, and build a control strategy database;

[0125] (4) Perform time series analysis on the control parameters in the control strategy database, calculate the change trend of each control parameter, and obtain the parameter evolution sequence;

[0126] (5) Predict and infer the control parameters based on the parameter evolution sequence, calculate the control parameter values at the next moment, and generate control parameter update data;

[0127] (6) The control parameter update data is organized in time sequence according to the control execution order to generate a real-time control instruction sequence.

[0128] Specifically, the lifting point position parameters are extracted, including the spatial coordinates and allowable deviation range of each lifting point. Next, the lifting torque parameters are analyzed, including the magnitude and direction of the force applied to each lifting point. Finally, the compensation control parameters are extracted, including displacement compensation and angle correction values. Numerical interval statistics are performed on these parameters to determine the value range and distribution characteristics of each parameter type, thereby obtaining control parameter characteristic data. This characteristic data reflects the specific control requirements of the current lifting condition. When extracting parameter records for similar conditions from the historical case database, similarity criteria must be defined. These similarity criteria include key characteristics such as structural type, dimensional ratio, weight distribution, and environmental conditions. These criteria are used to select historical cases with high similarity, and the control parameter records from these cases are then extracted. The extracted parameter records are standardized to unify the data format and dimensions across different cases, and correspondences between the parameters are established. These correspondences reflect the variation patterns of the parameters under different conditions and constitute the historical parameter mapping data.

[0129] A multi-dimensional similarity measurement method is used to calculate the similarity between control parameter feature data and historical parameter mapping data. For each control parameter, the degree of similarity with the corresponding parameter in the historical data is calculated, comprehensively considering the parameter's numerical differences, change trends, and control effects. Historical control strategies with high similarity are screened out. These strategies include parameter configuration plans and control process records. These screened control strategies are systematically organized to form a control strategy database. Time series analysis is performed on the data in the control strategy database, focusing on the temporal variation characteristics of the control parameters. Time series analysis methods are used to identify the periodicity, trend, and mutation characteristics of parameter changes. The variation patterns of each parameter type are modeled to generate mathematical models describing the parameter evolution process. These models reflect the parameter variation characteristics at different stages, forming a parameter evolution sequence. A progressive prediction method is used for predictive reasoning based on the parameter evolution sequence. Based on the parameter's historical change trend and current state, the parameter value at the next moment is predicted. The prediction process considers the mutual influence and constraints between parameters to ensure the rationality of the prediction results. The predicted parameter values constitute the control parameter update data, which reflects the latest change trends of the control parameters.

[0130] Organize control parameter update data based on execution priority and timing. Determine the order in which parameters are executed, typically prioritizing those that affect structural stability, followed by precision control parameters, and finally auxiliary control parameters. Following this order, organize the parameters into instruction sequences, each containing a specific execution time, target parameter value, and execution conditions.

[0131] For example, the position parameters (coordinate values and allowable deviations), torque parameters (magnitude and direction), and compensation parameters (displacement and angle compensation values) of the four lifting points are extracted from the lifting control data package. Hoisting records of the same type of roof truss are retrieved from the historical database, and five similar cases are found. By comparing the structural characteristics and construction conditions of the current working conditions with those of the historical cases, a parameter mapping relationship is established. Parameter similarity is calculated, and the two most matching historical cases are selected as references. The parameter variations in these two cases are analyzed to derive a mathematical model describing the parameter evolution process. The model predicts the next control parameters, such as the fine-tuning amount for the lifting point position and the torque compensation value. These parameters are organized into an execution instruction sequence in the order of "stability control-precision control-auxiliary control" to guide the precise implementation of the lifting process.

[0132] The above describes the real-time monitoring method for the hoisting accuracy of a large steel structure in the embodiment of the present application. The following describes the real-time monitoring system for the hoisting accuracy of a large steel structure in the embodiment of the present application. Figure 3 In one embodiment of the present application, a system for real-time monitoring of the hoisting accuracy of a large steel structure includes:

[0133] The acquisition module 201 is used to scan and collect the edge of the steel structure through a distributed sensor network to obtain the structural angular velocity data, strain data and load distribution data to obtain the original monitoring data set;

[0134] An extraction module 202 is configured to extract and analyze structural transient response characteristics based on the original monitoring data set to generate a structural state characteristic matrix including local deformation data and overall torsion data;

[0135] The coupling module 203 is used to calculate and analyze the coupling relationship between local deformation and overall torsion based on the structural state characteristic matrix, obtain torsional deformation evolution data and stress distribution prediction data, and integrate them to obtain a dynamic response prediction matrix;

[0136] A fusion module 204 is configured to perform spatiotemporal registration and data fusion on the real-time collected monitoring data based on the dynamic response prediction matrix to construct a state assessment data set including torsional deformation indicators and stress distribution indicators;

[0137] The optimization module 205 is used to perform dynamic optimization calculation on the hoisting parameters according to the state evaluation data set, and generate a hoisting control data packet including the hoisting point position parameters, hoisting torque parameters and compensation control parameters;

[0138] The control module 206 is used to use the hoisting control data packet in combination with historical case data to perform intelligent reasoning analysis and output a real-time control instruction sequence to achieve precision monitoring and control of the hoisting process.

[0139] Through the coordinated cooperation of the above components, the edges of the steel structure are scanned and collected through a distributed sensor network, realizing the multi-source heterogeneous data collection of structural angular velocity data, strain data and load distribution data, solving the problem that traditional single sensors are difficult to fully monitor the structural state; by extracting and analyzing the transient response characteristics of the original monitoring data set, a structural state characteristic matrix containing local deformation data and overall torsion data is generated, realizing a comprehensive characterization of the structural deformation characteristics; using the coupling relationship calculation and analysis method, the correlation between local deformation and overall torsion is deeply analyzed, and the torsional deformation evolution data and stress distribution prediction data are obtained, and these data are integrated to form The dynamic response prediction matrix provides a reliable basis for structural deformation prediction; based on the dynamic response prediction matrix, the monitoring data collected in real time are temporally and spatially aligned and fused, and a state evaluation data set containing torsional deformation indicators and stress distribution indicators is constructed, thereby improving the temporal and spatial consistency of the monitoring data; according to the state evaluation data set, a dynamic optimization algorithm is used to calculate the lifting parameters in real time, and a lifting control data packet containing lifting point position parameters, lifting torque parameters and compensation control parameters is generated, making the generation of control parameters more scientific; by combining historical case data for intelligent reasoning analysis, a real-time control instruction sequence is output, thereby realizing closed-loop optimization of precision monitoring and control of the lifting process. In terms of the application of artificial intelligence algorithms, the present invention innovatively combines deep learning with traditional mechanical models, establishes a prediction model that is more in line with engineering practice through mining and learning of historical data, and adopts intelligent reasoning algorithms to dynamically optimize the control strategy, which significantly improves the accuracy and real-time performance of the control instructions, making the entire monitoring and control scheme more adaptable and robust.

[0140] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application is described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A real-time monitoring method for the hoisting accuracy of a large steel structure, characterized in that: The large steel structure hoisting accuracy real-time monitoring method comprises: The distributed sensor network is used to scan and collect data on the edge of the steel structure to obtain the structural angular velocity data, strain data, and load distribution data, and obtain the original monitoring data set; Extracting and analyzing the transient response characteristics of the structure based on the original monitoring data set to generate a structural state characteristic matrix including local deformation data and overall torsion data; Based on the structural state characteristic matrix, the coupling relationship between local deformation and overall torsion is calculated and analyzed to obtain torsional deformation evolution data and stress distribution prediction data, and the dynamic response prediction matrix is obtained by integration; Based on the dynamic response prediction matrix, the real-time collected monitoring data is subjected to spatiotemporal registration and data fusion to construct a state assessment data set including torsional deformation indicators and stress distribution indicators; Performing dynamic optimization calculations on the hoisting parameters based on the state assessment data set to generate a hoisting control data packet including hoisting point position parameters, hoisting torque parameters, and compensation control parameters; The hoisting control data package is used in combination with historical case data to perform intelligent reasoning analysis and output a real-time control instruction sequence to achieve precision monitoring and control of the hoisting process.

2. The method for real-time monitoring of large steel structure hoisting accuracy according to claim 1 is characterized in that: The distributed sensor network is used to scan and collect the edge of the steel structure to obtain structural angular velocity data, strain data, and load distribution data, and obtain the original monitoring data set, including: Fiber Bragg grating sensors were placed in a spiral pattern along the edge of the steel structure at 0.5-meter intervals. High-frequency sampling was performed on the edge of the structure, and wavelength demodulation analysis was performed at each sampling point to obtain the initial strain signal. The initial strain signal was then decomposed into multiple scales using wavelet transform. The decomposed frequency band signals were thresholded and reconstructed to generate the structural edge strain data. The gyroscope array continuously samples the key nodes of the steel structure at a sampling frequency of 200Hz to collect the original data of the three-axis angular velocity. The raw data is filtered through a bandpass filter to remove high-frequency noise and low-frequency drift. The Kalman filter algorithm is then used to perform state estimation and prediction on the processed data to obtain the structural angular velocity data. Dynamic strain sensors are deployed at the load points of the lifting equipment to collect load time series data. The time domain data is converted to the frequency domain through fast Fourier transform, interfering frequency components are identified and eliminated, and the signal is reconstructed using inverse Fourier transform to obtain load distribution data with a high signal-to-noise ratio. Temperature compensation and linear calibration are performed on the edge strain data of the structure to remove the influence of environmental factors, and then the strain change rate is calculated using a sliding window method, and feature extraction is performed in combination with the strain amplitude to generate edge strain feature data; The structural angular velocity data is subjected to dimensionality reduction processing by principal component analysis to extract the main change features, and then the data is resampled and synchronized by a spline interpolation algorithm to obtain angular velocity feature data; Normalizing the load distribution data by maximum and minimum values, performing standardization based on the statistical characteristics of historical data, and then eliminating short-term fluctuations through a moving average method to construct load characteristic data; The edge strain characteristic data, angular velocity characteristic data and load characteristic data are aligned according to the time reference, and data association is established through timestamp matching. The aligned data are then interpolated and supplemented and outlier processing is performed. The three types of characteristic data are integrated into a structural state description matrix through a data fusion algorithm to generate the original monitoring data set.

3. The method for real-time monitoring of large steel structure hoisting accuracy according to claim 1, characterized in that: The method extracts and analyzes the transient response characteristics of the structure based on the original monitoring data set to generate a structural state characteristic matrix containing local deformation data and overall torsion data, including: Data is intercepted from the original monitoring data set by time domain segmentation, and data segments within a key time window are selected to generate time series feature data; Performing spatial correlation analysis on the data of each measuring point in the time series feature data, calculating the correlation coefficient matrix between the measuring points, and screening to obtain a key measuring point combination; Decomposing the data of the key measurement point combination in multiple dimensions, extracting displacement components, velocity components and acceleration components, and obtaining local response characteristics; Performing spatial mapping on the local response characteristics through data reconstruction, establishing a node displacement field, and obtaining local deformation data; Performing overall coordination analysis based on the local deformation data, calculating deformation gradients and deformation rates in each region, and constructing overall torsion data; The local deformation data and the overall torsion data are organized and arranged according to a time-space correspondence to generate the structural state characteristic matrix.

4. The method for real-time monitoring of large steel structure hoisting accuracy according to claim 1, characterized in that: The coupling relationship between local deformation and overall torsion is calculated and analyzed based on the structural state characteristic matrix to obtain torsional deformation evolution data and stress distribution prediction data, and integrate them to obtain a dynamic response prediction matrix, including: Segmenting the structural state characteristic matrix according to time series, dividing the continuous data stream into multiple time windows, and extracting local deformation and overall torsion data segments within each window; Performing spatial decoupling analysis on the local deformation data segments, calculating the relative displacement and deformation rate of each monitoring point, and generating a local deformation feature vector; Calculating the torsion angle distribution for the overall torsion data segment, establishing a torsion center trajectory, and obtaining an overall torsion feature vector; Performing coupling measurement on the local deformation eigenvector and the overall torsion eigenvector through tensor operation, establishing a stress-deformation mapping relationship, and obtaining the torsion deformation evolution data; Performing correction calculation on the stress-deformation mapping relationship based on the local stress concentration coefficient, constructing a stress distribution field, and generating the stress distribution prediction data; The torsional deformation evolution data and the stress distribution prediction data are combined and associated according to the time-space correspondence principle to establish a time-varying characteristic mapping matrix to form the dynamic response prediction matrix.

5. The method for real-time monitoring of large steel structure hoisting accuracy according to claim 1, characterized in that: Based on the dynamic response prediction matrix, the real-time collected monitoring data is subjected to spatiotemporal registration and data fusion to construct a state assessment data set including torsional deformation indicators and stress distribution indicators, including: Performing data alignment on the dynamic response prediction matrix through time series sampling, extracting reference time information of each monitoring point, and generating a time reference sequence; Performing time stamp comparison on the monitoring data collected in real time, establishing a mapping relationship with the time reference sequence, and obtaining time registration data; Using the monitoring point coordinate information to perform spatial mapping on the temporal registration data, calculate the relative position deviation of each node, and obtain spatial registration data; Performing correlation analysis on the temporal registration data and the spatial registration data with the dynamic response prediction matrix, calculating deviation distribution, and generating a torsional deformation index; Dividing the monitoring area into grids according to the torsional deformation index, calculating stress distribution characteristics at grid nodes, and constructing stress distribution indexes; The torsional deformation index and the stress distribution index are organized and sorted according to the topological relationship to obtain a multidimensional feature data table, which constitutes the state assessment data set.

6. The method for real-time monitoring of large steel structure hoisting accuracy according to claim 1, characterized in that: The method comprises: performing dynamic optimization calculation on the hoisting parameters according to the state evaluation data set to generate a hoisting control data packet including hoisting point position parameters, hoisting torque parameters and compensation control parameters, including: Performing gradient analysis on the torsional deformation index in the state assessment data set, calculating the spatial distribution of the structural deformation sensitive area, and obtaining deformation sensitivity distribution data; Based on the deformation sensitivity distribution data, a stress analysis is performed on the stress points of the structure, the stress concentration degree of each node is calculated, and stress distribution weight data is obtained; Numerical superposition of the deformation sensitivity distribution data and the stress distribution weight data is performed to evaluate and analyze the stress state of the hanging point and generate the hanging point position parameters; Perform moment balance calculation based on the lifting point position parameters, establish a force balance relationship between the lifting points, and construct the lifting moment parameters; Predicting the deformation trend during the lifting process according to the lifting torque parameters, calculating the displacement compensation amount of each key point, and obtaining the compensation control parameters; The lifting point position parameters, lifting torque parameters and compensation control parameters are integrated according to control priorities to generate the lifting control data packet.

7. The method for real-time monitoring of large steel structure hoisting accuracy according to claim 1, characterized in that: The method utilizes the hoisting control data packet, combines historical case data to perform intelligent reasoning analysis, and outputs a real-time control instruction sequence to achieve precision monitoring and control of the hoisting process, including: Performing feature analysis on the hoisting control data packet, extracting the numerical ranges of the hoisting point position parameters, the hoisting torque parameters, and the compensation control parameters, and obtaining control parameter feature data; Extract control parameter records under similar working conditions from historical case data, establish parameter correspondence, and generate historical parameter mapping data; Calculate the similarity between the control parameter feature data and the historical parameter mapping data, select historical control strategies with high matching degree, and build a control strategy database; Performing time series analysis on the control parameters in the control strategy database, calculating the change trend of each control parameter, and obtaining a parameter evolution sequence; Predicting and reasoning the control parameters based on the parameter evolution sequence, calculating the control parameter values at the next moment, and generating control parameter update data; The control parameter update data is organized in time sequence according to the control execution order to generate the real-time control instruction sequence.

8. A large-scale steel structure hoisting accuracy real-time monitoring system, used to implement the large-scale steel structure hoisting accuracy real-time monitoring method according to any one of claims 1 to 7, characterized in that: The large steel structure hoisting accuracy real-time monitoring system includes: The acquisition module is used to scan and collect the edge of the steel structure through a distributed sensor network to obtain the structural angular velocity data, strain data and load distribution data to obtain the original monitoring data set; An extraction module is used to extract and analyze the transient response characteristics of the structure based on the original monitoring data set, and generate a structural state characteristic matrix including local deformation data and overall torsion data; A coupling module is used to calculate and analyze the coupling relationship between local deformation and overall torsion based on the structural state characteristic matrix, obtain torsional deformation evolution data and stress distribution prediction data, and integrate them to obtain a dynamic response prediction matrix; A fusion module is used to perform spatiotemporal registration and data fusion on the monitoring data collected in real time based on the dynamic response prediction matrix to construct a state assessment data set including torsional deformation indicators and stress distribution indicators; an optimization module, configured to perform dynamic optimization calculations on the hoisting parameters based on the state evaluation data set, and generate a hoisting control data packet including hoisting point position parameters, hoisting torque parameters, and compensation control parameters; The control module is used to utilize the hoisting control data packet and combine it with historical case data to perform intelligent reasoning analysis, output a real-time control instruction sequence, and realize precision monitoring and control of the hoisting process.

Images