A health monitoring method for a high-rise cantilevered steel structure observation tower

Abstract

The invention provides a health monitoring method for a tall cantilevered steel structure sightseeing tower, and belongs to the technical field of health monitoring of tall cantilevered steel structure sightseeing towers. In order to solve the problem that the existing tall cantilevered steel structure buildings are difficult to monitor the structural health in a complex working environment, a health monitoring method for a tall cantilevered steel structure sightseeing tower is provided. The method constructs a finite element model of the sightseeing tower, analyzes the structural characteristics of the tower body, understands the operating characteristics of the tower body, and determines the health monitoring content and the selection of the distribution point positions according to the finite element calculation results, combined with the characteristics of the tall cantilevered steel structure sightseeing tower with a thin diameter, a high height and a large top mass. The distribution point positions are optimized by using a Bayesian optimization algorithm, a cloud monitoring platform is established based on the collected data, and a CNN+LSTM deep learning algorithm is integrated for safety warning, so as to realize the monitoring and warning of the operation status of the sightseeing tower. The invention is applied to the health status assessment of the tall cantilevered steel structure sightseeing tower.

Description

Technical Field

[0001] The invention provides a health monitoring method for a tall cantilevered steel structure sightseeing tower, belonging to the technical field of health monitoring of tall cantilevered steel structure sightseeing towers. Background Art

[0002] With the development of construction technology and building materials, people's demand for architectural style has gradually changed from building conventional buildings to building iconic buildings. Some large-scale architectural facilities with landmark and sightseeing features have emerged, and sightseeing towers are a typical example.

[0003] Since the tall sightseeing tower has a slender appearance and a large aspect ratio, and the tower body is a multi-degree-of-freedom system, a large-mass sightseeing layer structure is often set at its end to meet the needs of sightseeing and entertainment. As a result, the entire tower structure is sensitive to external environmental stimuli. Therefore, a targeted health monitoring program is needed to evaluate whether the sightseeing tower structure is safe and to warn of possible dangers, so as to reduce the probability of safety accidents and improve the safety of the sightseeing tower steel structure during operation.

[0004] However, in the process of health monitoring of high-rise cantilevered observation towers, there are the following technical difficulties and defects:

[0005] (1) The high cantilevered steel structure is a multi-degree-of-freedom system with diverse and complex vibration modes, making it difficult to pick up vibration modes during monitoring;

[0006] (2) The top of a tall cantilevered steel structure has a large mass, and the vibration displacement under external excitation is large, which can easily cause multiple changes in the stress state of the components, and the change amplitude is large. Due to the large number of structural components, it is difficult to fully monitor the stress state of the structure and the arrangement of monitoring points;

[0007] (3) The stress state of the structure changes, and it is difficult to warn of the health status;

[0008] (4) Tall structures are easily affected by the environment, and high accuracy and flexibility are required in monitoring environmental factors. Summary of the invention

[0009] In order to solve the problem that the existing high-rise cantilevered steel structure buildings are difficult to monitor their structural health in a complex working environment, the present invention provides a health monitoring method for a high-rise cantilevered steel structure sightseeing tower.

[0010] In order to solve the above technical problems, the technical solution adopted by the present invention is: a health monitoring method for a high-rise cantilevered steel structure sightseeing tower, comprising the following monitoring steps:

[0011] Step 1: Use finite element calculation software to establish a three-dimensional finite element model of the towering cantilevered steel structure sightseeing tower, monitor the structural vibration test of the towering cantilevered steel structure sightseeing tower, feedback the monitoring parameters and modify the accuracy of the three-dimensional finite element model, and perform finite element numerical simulation of the towering cantilevered steel structure under different working conditions during the construction and use stage;

[0012] Step 2: Comprehensively consider the structural characteristics of the tall cantilevered steel structure sightseeing tower and combine engineering experience to preliminarily determine the monitoring plan;

[0013] Step 3: Comprehensively consider economic benefits and monitoring effect factors, determine the prior distribution probability of the location and number of measuring points, update the posterior distribution probability according to the finite element calculation results of step 1, select the optimal layout plan, and use the Bayesian algorithm to further optimize the monitoring plan determined in step 2;

[0014] Step 4: According to the finite element calculation results of step 1, combined with relevant industry specifications, set the monitoring threshold, arrange monitoring equipment, collect monitoring data, pre-process the collected monitoring data, obtain the natural frequency according to the finite element calculation, set the low-pass filter cutoff frequency, and remove high-frequency noise;

[0015] The data collected by the monitoring equipment is transmitted to the local server, and then uploaded to the cloud database through the network for storage and backup. The cloud database is connected to the health monitoring website to observe the data changes of each monitoring point in the structure in real time;

[0016] Step 5: The cloud database conducts a comprehensive analysis of the data collected by the sensor and uses a safety warning module with a built-in CNN+LSTM deep learning algorithm model to predict structural response changes based on changes in the current monitoring data. When the predicted value approaches the preset monitoring threshold or the sensor works abnormally, an alarm is issued to provide a safety warning.

[0017] The three-dimensional finite element model established in the step one is specifically a tall cantilevered steel structure, including a podium, a core tube structure, a multi-layered ring structure, and a spire from bottom to top, wherein the podium is arranged at the bottom of the core tube structure, the multi-layered ring structure is arranged at the top periphery of the core tube structure, and the spire is arranged at the top of the core tube structure.

[0018] The specific method of performing finite element numerical simulation on the high-rise cantilever steel structure in the construction and use stage according to the working conditions in step 1 is:

[0019] The force calculation is carried out for each construction condition of the towering cantilever steel structure during the completion and use stage, including the splicing stage, lifting stage, unloading stage and closing stage of the cantilever structure:

[0020] In the assembly stage: firstly, analyze and calculate the stress state of the temporary assembly platform built on the roof of the podium building, and then calculate the stress state of the components of the revolving restaurant after assembling them on the platform;

[0021] During the hoisting stage: after the assembly is completed, the connection between the main structure and the assembly platform and other structures is gradually released, the structure is hoisted, and the changes in the structural stress state of the revolving restaurant from leaving the assembly platform to being lifted to the designed position and spliced ​​with the main structure are calculated;

[0022] In the unloading stage: after the splicing is completed, the steel strands are unloaded and removed, and the stress changes of the structure before and after the removal of the steel strands are calculated;

[0023] During the closing stage: after removing the steel strands, gradually remove the temporary reinforcement components and calculate the impact of the removal of each component on the structural stress;

[0024] Among them, in the cantilever structure splicing stage and cantilever structure closure stage, the implicit static module of the finite element calculation software is used for calculation. During the calculation process, only the gravity load is applied to the whole structure. The two-node linear beam unit B12 is used for the steel structure main body and the components in the revolving restaurant, and the bottom of the main structure adopts full freedom constraint;

[0025] Among them, during the cantilever structure lifting stage and cantilever structure unloading stage, the transient dynamic module of the finite element calculation software is used for calculation. The two-node linear beam unit B12 is used for the steel structure main body and the components in the revolving restaurant, and the two-node linear three-dimensional truss unit T3D2 is used for the steel strand used for lifting. The load is deadweight, the bottom degree of freedom of the main structure column is fully constrained, and the lifting speed of the steel strand is set to 1mm / s.

[0026] The monitoring scheme preliminarily determined in step 2 is as follows: data collection is performed by installing sensors with corresponding functions at different locations of the towering cantilevered steel structure sightseeing tower, specifically including:

[0027] Set vibration measurement points and arrange vibration sensors to pick up the vibration state of the tower body or the sensor setting location. The arrangement positions are as follows: install acceleration sensors at every H / 10 height along the tower body to pick up the vertical vibration state of the main structure; install acceleration sensors at the multi-layer ring structure and the top of the tower to pick up the plane vibration state of the structure at the corresponding position, where H is the tower body height;

[0028] Stress and strain measuring points are set up, and stress and strain sensors are arranged to collect structural stress and strain change data. The arrangement positions are as follows: stress and strain measuring points are arranged at intervals of H / 10 along the tower body; stress and strain measuring points are arranged at the rods with larger forces in the conversion truss; stress and strain measuring points are arranged at the splicing position of each layer of multi-layer annular structure and core tube structure, and at the splicing node position of the ring beam;

[0029] Set up displacement measurement points and arrange displacement sensors to collect structural displacement data;

[0030] Set up temperature and humidity measurement points, and arrange temperature and humidity sensors to collect temperature and humidity data of the structure itself and the surrounding environment. The arrangement positions are as follows: temperature and humidity sensors are arranged at the top of the tower and at half of the tower;

[0031] Wind speed and wind direction measuring points are set up, and wind speed and wind direction sensors are arranged to collect wind speed and wind direction data around the structure. The wind speed and wind direction sensors are arranged at the top of the tower, the top of the multi-layer ring structure, and half of the tower body.

[0032] The specific method for further optimizing the monitoring scheme in step 3 is:

[0033] Specifically, the Bayesian algorithm is used to select the optimal layout plan based on the posterior distribution probability:

[0034] Step 3.1: Define the evaluation index of the model, the expression is:

[0035] ;

[0036] Where N is the number of model substructures, i.e., the number of members. a i is the member damage coefficient, which is 0 under standard conditions, and the state index is also 0; c i represents the average decrease in structural performance when the i-th substructure is damaged; the effective stiffness value s i represents the force required to cause damage to the i-th substructure; c is a normalization constant, and T is the matrix transpose symbol;

[0037] Step 3.2: Using the damage coefficient α and the model prediction error η as variables, calculate the model prediction error. The calculation formula is:

[0038] ;

[0039] Among them, G M represents the modal properties predicted by the model;

[0040] Based on Bayes' theorem, given an observation output And the selected structural model and probability model, calculate the joint posterior probability density function of the unknown parameter α, the calculation formula is:

[0041] p( | )=cp( | )p( );

[0042] Among them, p(α| ) is given The probability density function under the state, c is the normalization constant, p( |α) is the probability density function of the given parameter α, and p(α) is the initial probability;

[0043] Step 3.3: Choose an uncorrelated zero-mean Gaussian model represented by the covariance matrix ∑η to represent the prediction error η, and simplify the likelihood function to a multivariate normal PDF, expressed as:

[0044] ;

[0045] in, represents the probability density function under a given parameter α; T is the matrix transpose symbol;

[0046] The joint posterior probability density function is asymptotically approximated by a joint Gaussian probability density function centered on the maximum a posteriori probability MAP estimate of α, and the covariance matrix used is , the maximum a posteriori probability MAP estimate is obtained by minimizing the negative logarithm of the posterior PDF, and the calculation formula is:

[0047] ;

[0048] Among them, p(α) is the initial probability, MAP refers to the maximum a posteriori probability, and F MAP is the objective function of MAP;

[0049] Evaluate at the MAP point and the calculation formula is:

[0050] ;

[0051] Among them, F MAP represents the objective function, express The maximum a posteriori probability estimate under F MAP is the second-order derivative of the objective function FMAP with respect to parameter α;

[0052] The formula is derived to obtain:

[0053] ;

[0054] in, is the diagonal covariance matrix, is the standard deviation coefficient of λ;

[0055] N m Indicates the number of natural frequencies;

[0056] λ r represents the eigenvalue;

[0057] represents the mode vector;

[0058] represents the estimated value of α;

[0059] represents the standard deviation coefficient;

[0060] diag(δ) represents a matrix with diagonal elements δ, where δ is the sensor configuration vector;

[0061] Use c T α is used to evaluate the state of the structure, and the uncertainty associated with the state indicator estimate is SD ( ) quantization, where c is the normalization constant, c T is the transposed matrix of c, is the covariance matrix of α, and the optimal degree of freedom is selected from different model degrees of freedom to minimize SD and obtain accurate index estimation in structural state assessment;

[0062] Step 3.4: Quantify the uncertainty in the updated damage coefficients by using the modal characteristics of the first 10 vibration modes in the Bayesian update process and calculating the standard deviation coefficients of all modes based on engineering experience. and Take 0.1, and let C The solution with the smallest estimated SD value is the optimal sensor layout solution.

[0063] The specific method of using the CNN+LSTM deep learning algorithm model for state monitoring in step 5 is:

[0064] Construct the input layer, CNN convolution layer, pooling layer, LSTM layer, and output layer of the CNN+LSTM model respectively;

[0065] The input signal is standardized and input into the CNN convolution layer. The wide convolution kernel is used to adaptively extract features. The extracted features are pooled by the maximum pooling layer to reduce the data dimension and retain the main feature information.

[0066] The reduced feature data is then used as the feature input of the LSTM layer to train the neural network and automatically learn the sequence features.

[0067] Back propagate the training error and gradually update the model parameters layer by layer;

[0068] Use activation functions to complete regression tasks for multi-feature input sequences.

[0069] The specific construction method of using the CNN+LSTM deep learning algorithm model in step 5 is:

[0070] Construct an original database, collect field measured data to construct a data set, and use finite element calculation software to analyze the changes in structural response under different states and environmental factors. Use the calculation results to supplement the data set. In the subsequent continuous monitoring process, continuously collect data and supplement the data set. Then divide the data set into training set, verification set, and test set in a ratio of 60%, 20%, and 20%;

[0071] The normalization method is used to preprocess the original data, the PSO particle swarm optimization algorithm is used to select the optimal input length and output length of the data, and the sliding window method is used to complete the construction of the data set. The expression is:

[0072] ;

[0073] in, x norm represents the normalized x value, x Indicates the actual value, x min represents the minimum value of x, x max Indicates the maximum value of x;

[0074] Construct a CNN+LSTM neural network, using temperature, wind speed, and wind direction environmental variables as input variables to predict changes in structural response over a period of time in the future;

[0075] Then, the model parameters are adjusted continuously, the model prediction results are compared with the actual results, and the mean absolute error MAE is used as the evaluation standard to select the prediction model with the best prediction effect, and this model is used as the model for subsequent data prediction. The expression is:

[0076] ;

[0077] Among them, MAE means mean absolute error;

[0078] i represents the data number;

[0079] n represents the total number of data;

[0080] y i represents the i-th actual value;

[0081] represents the i-th predicted value.

[0082] The beneficial effects of the present invention compared with the prior art are as follows: the present invention provides a health monitoring method for a tall cantilevered steel structure sightseeing tower, which comprehensively considers the structural characteristics of a tall building with a large height, a small cross-sectional area and a large end mass, and presents an overall single-degree-of-freedom system. The structural vibration effect is significant, and the structural vibration needs to be strictly monitored. In view of the above structural characteristics, the present invention installs various types of sensors on the tower body, structural splicing positions and important structural positions to form a data acquisition system, formulates a reasonable health monitoring plan for the tall cantilevered steel structure, provides scientific guidance for the arrangement of measuring points, data transmission, data processing, and structural safety performance evaluation, etc., performs health monitoring of the tall structure sightseeing tower throughout its life cycle, provides a guarantee for the normal use of the structure, and improves the safety of the steel structure during construction and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] The present invention will be further described below in conjunction with the accompanying drawings:

[0084] Figure 1 It is a structural schematic diagram of a health monitoring system for a high-rise cantilevered steel structure sightseeing tower according to the present invention;

[0085] Figure 2 It is a schematic diagram of the force analysis of the cantilever structure of the present invention at the splicing stage;

[0086] Figure 3 It is a schematic diagram of the force analysis of the cantilever structure of the present invention during the lifting stage;

[0087] Figure 4 It is a schematic diagram of force analysis of the cantilever structure of the present invention in the unloading stage;

[0088] Figure 5 It is a schematic diagram of the force analysis of the cantilever structure of the present invention at the closing stage;

[0089] Figure 6 A schematic diagram of the positions of the vibration measuring points arranged on the high-rise annular steel structure of the present invention;

[0090] Figure 7 A schematic diagram of the arrangement of strain measuring points on the multi-layer annular structure of the present invention;

[0091] Figure 8 A schematic diagram of the arrangement of strain measuring points on the conversion truss of the present invention;

[0092] Fig. 9 It is a schematic diagram of the arrangement of strain measuring points on the podium, tower body and tower spire of the present invention;

[0093] Fig.10 It is a schematic diagram of the arrangement of temperature and humidity measuring points on the high cantilevered steel structure of the present invention;

[0094] Fig.11A schematic diagram of the arrangement of wind speed measuring points on a high-rise cantilevered steel structure of the present invention;

[0095] Fig.12 This is a working principle diagram of the CNN+LSTM neural network of the present invention;

[0096] The numbers that appear in the figure are: 1 for the podium, 2 for the tower body, 3 for the multi-layered ring structure, and 4 for the spire. DETAILED DESCRIPTION

[0097] The present invention provides a health monitoring method for a high-rise cantilevered steel structure sightseeing tower. The method constructs a finite element model of the sightseeing tower, analyzes the structural characteristics of the tower body, understands the operating characteristics of the tower body, and determines the health monitoring content and the selection of the point locations based on the finite element calculation results, combined with the characteristics of the high-rise cantilevered steel structure sightseeing tower with thin diameter, high height and large top mass. The point locations are further optimized using a Bayesian optimization algorithm, a cloud monitoring platform is established based on the collected data, and a CNN+LSTM deep learning algorithm is integrated for safety warning, so as to achieve the purpose of monitoring and warning the operating status of the sightseeing tower. The monitoring method can be used for health status assessment of high-rise cantilevered steel structures to ensure their operational safety and stability.

[0098] like Figure 1 As shown, the present invention proposes a health monitoring method for a high-rise cantilevered steel structure, which mainly includes the following monitoring steps:

[0099] Step 1: Use finite element calculation software to establish a three-dimensional finite element model of the towering cantilevered steel structure observation tower, and conduct structural vibration test monitoring to feedback and modify the model accuracy, and conduct finite element numerical simulation of the working conditions of the structure during the construction and use stage considering earthquakes, wind loads, etc.;

[0100] Step 2: Based on experience, taking into account the characteristics of the tall cantilevered steel structure sightseeing tower, which has a thin diameter, high height, and large top mass, and combined with engineering experience, a preliminary monitoring plan is determined;

[0101] Step 3: Based on the finite element calculation results of step 1, taking into account economic benefits, monitoring effects and other factors, the Bayesian algorithm is used to further optimize the monitoring plan determined in step 2;

[0102] Step 4: According to the finite element calculation results of step 1 and in combination with relevant industry specifications, set the monitoring threshold, arrange monitoring equipment, collect monitoring data, and observe the data changes of each monitoring point of the structure in real time;

[0103] Step 5: Establish a cloud database, transfer the collected data to the cloud database, conduct comprehensive analysis, use deep learning algorithms for safety warnings, and establish a health monitoring website to display monitoring data in real time.

[0104] Furthermore, in step 1, the finite element model is mainly a high-rise cantilevered steel structure, including a podium 1, a core tube structure 2, a multi-layered ring structure 3 and a spire 4 from bottom to top, the podium 1 and the multi-layered ring structure 3 are located at the ends of the core tube structure 2, and the bottom of the spire 4 structure is connected to the top of the core tube structure 2;

[0105] Furthermore, the finite element model parameters described in step 1 are obtained by conducting on-site vibration tests after the core tube is built, and the finite element calculation accuracy is continuously corrected in combination with the monitoring test data. Finally, the finite element model parameters after repeated corrections are obtained, which also provide accurate model parameters for subsequent structural vibration control and safety monitoring, and provide a strong basis for the arrangement of measurement points.

[0106] Further, such as Figures 2 to 5 As shown, the calculation process using finite element software analysis and calculation is divided into four stages, namely, cantilever structure splicing stage, cantilever structure hoisting stage, unloading stage, and closing stage. The implicit static module of finite element calculation software is used for calculation in the structural assembly stage and closing stage, while there are sudden changes in loads in the hoisting stage and steel strand unloading stage. Considering its dynamic effect, the transient dynamic module of finite element calculation software is used for calculation.

[0107] Furthermore, the structural assembly stage refers to first building a temporary assembly platform on the roof of the podium, and then assembling the various components of the revolving restaurant on the platform; the hoisting stage refers to releasing the connection between the main structure and the assembly platform and other structures after the assembly is completed, hoisting the structure, and splicing the revolving restaurant with the main structure when it is lifted to the designed position; the steel strand unloading stage refers to unloading the steel strand after the splicing is completed; the closing stage refers to removing the temporary reinforcement components after the steel strand is removed.

[0108] Furthermore, the implicit static module of the finite element calculation software is used for calculation in the structural assembly and connection stages. During the calculation process, only gravity load is applied to the entire structure. The components in the steel structure body and the revolving restaurant all use two-node linear beam units (B12), and the bottom of the main structure adopts full freedom constraints.

[0109] Furthermore, the lifting stage and the steel strand unloading stage are calculated using the transient dynamic module of the finite element calculation software, the steel structure main body and the components in the revolving restaurant are both calculated using two-node linear beam units (B12), and the lifting cables are calculated using two-node linear three-dimensional truss units (T3D2); the load is deadweight, and the bottom degrees of freedom of the main structure column are fully constrained; the steel strand lifting speed is 1 mm / s.

[0110] Further, such as Figures 6 to 11 As shown, the types of sensors selected for the preliminary monitoring scheme described in step 2 are as follows:

[0111] Set up vibration measurement points and arrange vibration sensors to pick up the vibration state of the tower body or where the sensors are set;

[0112] Set up stress and strain measuring points and arrange stress and strain sensors to collect structural stress and strain change data;

[0113] Set up displacement measurement points and arrange displacement sensors to collect structural displacement data;

[0114] Set up temperature and humidity measurement points, and arrange temperature and humidity sensors to collect temperature and humidity data of the structure itself and the surrounding environment;

[0115] Set up wind speed and direction measurement points, and arrange wind speed and direction sensors to collect wind speed and direction data around the structure;

[0116] Furthermore, the arrangement of the vibration measuring point positions comprehensively considers the special form of the structure, that is, the structure is a single degree of freedom system with a huge top mass. According to the finite element calculation results, the vibration mode of the structure is wavy under the influence of wind-induced effects, and the vibration mode contains torsion. Therefore, in order to effectively pick up the vibration state of the main structure of the tower body, the specific arrangement is as follows:

[0117] Acceleration sensors are installed at intervals of H / 10 along the tower body to pick up the vertical vibration state of the main structure; acceleration sensors are installed on the multi-layer ring structure and the top of the tower to pick up the plane vibration state of the structure at the corresponding position, where H is the height of the tower body;

[0118] Furthermore, the arrangement of the stress and strain monitoring points comprehensively considers the finite element calculation results and the actual engineering conditions, and stress and strain sensors are installed along the length direction of the tower body, and stress and strain measuring points are installed at the splicing nodes, important structural components, and locations where the structural stress and strain are large. The specific arrangement is as follows:

[0119] Stress and strain measuring points are arranged at intervals of H / 10 along the tower body; stress and strain measuring points are arranged at the members with larger forces in the conversion truss; stress and strain measuring points are arranged at the joints of each layer of multi-layer annular structure and core tube structure and at the joints of the ring beam;

[0120] Furthermore, the environmental monitoring points are specifically arranged as follows:

[0121] Temperature and humidity sensors are arranged at the top of the tower and at half of the tower;

[0122] Wind speed and direction sensors are arranged at the top of the tower, the top of the multi-layer ring structure and half of the tower body.

[0123] Furthermore, the steps of optimizing the scheme described in step 3 are as follows: after the monitoring scheme is preliminarily determined in step 2, it is further optimized by the Bayesian algorithm, and the economic benefits, i.e., the monitoring results and other factors are comprehensively considered to determine the prior distribution probability of the location and number of measuring points, and then the posterior distribution probability is updated according to the finite element calculation results of step 1, and the optimal layout scheme is selected. The specific method is as follows:

[0124] (1) Define model evaluation indicators:

[0125] ;

[0126] Where N is the number of model substructures, i.e., the number of members. a i is the member damage coefficient, which is 0 under standard conditions, and the state index is also 0; c i represents the average decrease in structural performance when the i-th substructure is damaged; the effective stiffness value s i represents the force required to cause damage to the i-th substructure; c is a normalization constant, and T is the matrix transpose symbol;

[0127] With the damage coefficient α and the model prediction error η as variables, the model prediction error is shown as follows:

[0128] ;

[0129] Among them, G M represents the modal properties predicted by the model;

[0130] Using Bayes' theorem, given an observation output And under the selected structural model and probability model, the joint posterior probability density function of the unknown parameter α can be obtained:

[0131] ;

[0132] Among them, p(α| ) is given The probability density function under the state, c is the normalization constant, p( |α) is the probability density function of a given parameter α, and p(α) is the initial probability.

[0133] In structural dynamics applications, an uncorrelated zero-mean Gaussian model characterized by a covariance matrix ∑η is often chosen to represent the prediction error η, which allows the likelihood function to be simplified to a multivariate normal PDF:

[0134] ;

[0135] in, represents the probability density function under a given parameter α; T is the matrix transpose symbol;

[0136] When there is a large amount of data, the joint posterior probability density function can be asymptotically approximated by a joint Gaussian probability density function centered around the maximum a posteriori probability (MAP) estimate of α, characterized by the covariance matrix The MAP estimate can be obtained by minimizing the negative logarithm of the posterior PDF:

[0137] ;

[0138] Among them, p(α) is the initial probability, MAP refers to the maximum a posteriori probability, and F MAP is the objective function of MAP;

[0139] Evaluate at the MAP point:

[0140] ;

[0141] Among them, F MAP represents the objective function, express The maximum a posteriori probability estimate under F MAP is the second-order derivative of the objective function FMAP with respect to parameter α;

[0142] The following formula can be obtained by deducing the formula:

[0143] ;

[0144] in, is the diagonal covariance matrix, is the standard deviation coefficient of λ;

[0145] N m Indicates the number of natural frequencies;

[0146] λ r represents the eigenvalue;

[0147] represents the mode vector;

[0148] represents the estimated value of α;

[0149] represents the standard deviation coefficient;

[0150] diag(δ) represents a matrix with diagonal elements δ, where δ is the sensor configuration vector;

[0151] Use c T α is used to evaluate the state of the structure, and the uncertainty associated with the state indicator estimate is SD ( ) quantification. The optimal degree of freedom is selected from different model degrees of freedom to minimize the SD, thereby obtaining the most accurate indicator estimation in structural state assessment.

[0152] (2) In order to quantify the uncertainty in the updated damage coefficient, the modal characteristics of the first 10 vibration modes are used in the Bayesian correction process. Based on engineering experience, the standard deviation coefficient of all modes is and Take 0.1. Let I C The solution with the smallest estimated SD value is the optimal sensor layout solution.

[0153] Furthermore, the monitoring scheme described in step 3 is optimized, and the optimization objects include the selection of sensor types, the determination of the number of measuring points, and the selection of measuring point locations;

[0154] Furthermore, the collected data is preprocessed, and the low-pass filter cutoff frequency is set in combination with the natural frequency obtained by finite element calculation to remove high-frequency noise, and then transmitted to the cloud database via the local server for storage and backup;

[0155] Furthermore, in order to avoid the influence of harsh data environment and unsatisfactory data transmission effect caused by long-distance data transmission, step 4 adopts wired transmission to transmit the data collected by the sensor to the local server, and then upload it to the cloud database through the network. The database is connected to the health monitoring website, and users can monitor the data changes of each measuring point of the structure in real time by visiting the website;

[0156] Furthermore, step five includes a safety warning module, which uses a CNN+LSTM deep learning algorithm model and can predict the upcoming structural response changes of the structure based on the current monitoring data changes, and issue an alarm when the predicted value approaches the monitoring threshold set in step three or the sensor works abnormally.

[0157] Furthermore, the CNN+LSTM deep learning algorithm model uses the field measured data as the data set to train the model. In actual measurement, not all working conditions may be monitored. However, in order to improve the accuracy of the model prediction results, the finite element calculation and analysis method is used to supplement the monitoring data with the finite element calculation results to make the data set more complete. In actual work, the current measured data is used as the model input data, and the structural response in the future period is used as the output to predict the changes in the structural response in real time.

[0158] Further, such as Fig.12 As shown in the figure, the CNN+LSTM model mainly consists of a data input layer, a CNN convolution layer, a pooling layer, a LSTM layer, and an output layer. The specific workflow is as follows:

[0159] 1) The input signal is standardized and input into the CNN convolution layer, and features are adaptively extracted using a wide convolution kernel. The extracted features are pooled in the maximum pooling layer to reduce the data dimension and retain the main feature information.

[0160] 2) The reduced feature data is then used as the feature input of the LSTM layer to train the neural network and automatically learn the sequence features;

[0161] 3) Back propagate the training error and gradually update the model parameters layer by layer;

[0162] 4) Use activation functions to complete the regression task of multi-feature input sequences.

[0163] Furthermore, the steps for building the CNN+LSTM model are as follows:

[0164] 1) Construct the original database, collect field measured data to construct the data set, and use finite element calculation software to analyze the changes in structural response under different states and environmental factors. Use the calculation results to supplement the data set. In the subsequent continuous monitoring process, continuously collect data and supplement the data set. Then divide the data set into training set, verification set and test set according to the ratio of 60%, 20% and 20%;

[0165] 2) Preprocess the data set. Since the data acquisition devices are arranged at different rod positions, the normalization method is used to preprocess the original data to eliminate the impact of the differences caused by different dimensions on the model prediction effect. The PSO particle swarm optimization algorithm is used to select the optimal input length and output length of the data, and the sliding window method is used to complete the construction of the data set.

[0166] ;

[0167] in, x norm represents the normalized x value, x Indicates the actual value, x min represents the minimum value of x, x max Indicates the maximum value of x;

[0168] 3) Construct a CNN+LSTM neural network and use environmental variables such as temperature, wind speed, and wind direction as input variables to predict changes in structural response over a period of time in the future;

[0169] 4) Continuously adjust the model parameters, compare the model prediction results with the actual results, use the mean absolute error (MAE) as the evaluation standard, select the prediction model with the best prediction effect, and use this model as the model for subsequent data prediction;

[0170] ;

[0171] Among them, MAE means mean absolute error;

[0172] i represents the data number;

[0173] n represents the total number of data;

[0174] y i represents the i-th actual value;

[0175] represents the i-th predicted value.

[0176] Furthermore, the embodiment of the present application discloses a health monitoring method for a high-rise sightseeing tower, the specific steps of which are:

[0177] Firstly, a three-dimensional finite element model of the towering cantilevered steel structure observation tower was established using finite element calculation software, and structural vibration monitoring was carried out to provide feedback to modify the model accuracy. Finite element numerical simulation was carried out for different working conditions in the completed and operational stages of the structure, taking into account earthquakes, wind loads, etc.

[0178] Based on engineering experience and the structural characteristics of the structure with thin diameter, high height and large top mass, the detection plan was initially determined; then, based on the finite element calculation results, the monitoring plan was further optimized by comprehensively considering economic benefits and monitoring effects. The specific optimization steps are as follows:

[0179] After the monitoring plan is initially determined, the Bayesian algorithm is used for further optimization. The optimization objects include the selection of sensors, the determination of the number of measuring points, and the selection of the locations of measuring points. The economic benefits, i.e., the monitoring results, and other factors are comprehensively considered to determine the prior distribution probabilities of the locations and the number of measuring points. Then, the posterior distribution probabilities are updated according to the finite element calculation results of step one, and the optimal layout plan is selected.

[0180] The finite element model is mainly a high-rise cantilevered steel structure, including a bottom-up podium 1, a tower body 2 (core tube structure), a multi-layered ring structure 3 and a spire 4. The podium and the multi-layered replacement structure are located at the end of the core tube structure, and the bottom of the spire structure is connected to the top of the core tube structure. Each bar is modeled using a two-node linear BEAM (B12) unit. The implicit static module of the finite element calculation software is used for calculation during the structural assembly and closure stages. There are sudden changes in loads during the hoisting and strand unloading stages. Considering their dynamic effects, the transient dynamic module of the finite element calculation software is used for calculation. The bottom degree of freedom of the main structure column is fully constrained. The finite element model is a field vibration test after the core tube is completed. The finite element calculation accuracy is continuously corrected in combination with monitoring data. The repeatedly corrected finite element model also provides accurate model parameters for subsequent structural vibration control and safety monitoring.

[0181] The optimal layout after Bayesian algorithm optimization is as follows:

[0182] 1) Set up vibration measuring points and arrange vibration sensors to pick up the vibration state of the tower body or where the sensors are set; the arrangement of the vibration measuring points takes into account the special form of the structure, that is, the structure is a single-degree-of-freedom system with a huge top mass. According to the finite element calculation results, the vibration mode of the structure is wavy under the influence of wind-induced effects, and the vibration mode contains torsion. Therefore, in order to effectively pick up the vibration state of the main structure of the tower body, the specific arrangement is as follows:

[0183] Acceleration sensors are installed at intervals of H / 10 along the tower body to pick up the vertical vibration state of the main structure; acceleration sensors are installed on the multi-layer ring structure and the top of the tower to pick up the plane vibration state of the structure at the corresponding position, where H is the height of the tower body;

[0184] 2) Set up stress and strain measuring points and arrange stress and strain sensors to collect structural stress and strain change data; The arrangement of stress and strain monitoring points comprehensively considers the finite element calculation results and the actual engineering conditions, installs stress and strain sensors along the length of the tower, and installs stress and strain measuring points at splicing nodes, important structural components, and locations where the structural stress and strain are large. The specific arrangement is as follows:

[0185] Stress and strain measuring points are arranged at intervals of H / 10 along the tower body; stress and strain measuring points are arranged at the members with larger forces in the conversion truss; stress and strain measuring points are arranged at the joints of each layer of multi-layer annular structure and core tube structure and at the joints of the ring beam;

[0186] 3) Set up displacement measurement points and arrange displacement sensors to collect structural displacement data;

[0187] 4) Set up temperature and humidity measurement points, and arrange temperature and humidity sensors to collect temperature and humidity data of the structure itself and the surrounding environment; the temperature and humidity sensors are arranged at the top of the tower and at half of the tower;

[0188] 5) Set up wind speed and wind direction measuring points, and arrange wind speed and wind direction sensors to collect wind speed and wind direction data around the structure; wind speed and wind direction sensors are arranged at the top of the tower, the top of the multi-layer ring structure, and half of the tower body.

[0189] The local server provided in the present invention is used for data transfer, real-time acquisition of monitoring data, and establishment of a cloud database, and the massive data obtained by the sensor is transmitted to the cloud database via Ethernet for data storage and backup.

[0190] Due to complex environmental factors, the collected data needs to be preprocessed, and the low-pass filter cutoff frequency is set based on the natural frequency obtained by finite element calculation to remove high-frequency noise. The data is then transmitted to the cloud database for storage and backup via the local server.

[0191] In order to avoid the impact of harsh data environments and unsatisfactory data transmission effects caused by long-distance data transmission, data transmission adopts wired transmission to transmit various data collected by sensors to the local server, and then upload them to the cloud database through the network;

[0192] The present invention also proposes a safety warning scheme, which integrates the CNN+LSTM deep learning algorithm, uses the field measured data as a data set training model, and uses finite element calculation and analysis data to supplement the monitoring data, making the data set more complete. It can predict the upcoming structural response changes of the structure according to the changes in the current monitoring data, and issue an alarm when the predicted value approaches the monitoring threshold set in step three or when the sensor works abnormally.

[0193] The present invention will be further described below by means of specific embodiments:

[0194] In an embodiment, the towering annular steel structure used in the present invention is as follows Figure 6 As shown, it mainly includes a podium 1, a tower body 2 of the steel structure, a multi-layered ring structure 3 and a spire 4. The main body and tower body above ground are steel structure projects, the basement is a frame shear wall concrete structure, and the steel structure is a cylindrical structure: it is composed of steel tube concrete (steel tube) columns, steel platform beams and supports. The towering ring steel structure includes 1 underground floor and 25 above-ground floors, with a building elevation of 196m. The first underground floor is 5.20m high, the main podium is 2 floors with a floor height of 6.20m, and the tower body is 6.20m per floor. The spire has 16 floors with a total height of 54m, which is composed of columns, ring beams and diagonal braces. The 8 columns are located from the core column positioning point along a straight line to the corresponding column positioning point at the top of the tower body, and all use steel tube columns.

[0195] Finite element analysis software is used to perform modal calculations on the entire structure. The tower body 2, multi-layered ring structure 3 and spire 4 of the steel structure are all constructed using two-node linear BEAM units (B12). The constraints are: the bottom degree of freedom of the main structure column is fully constrained. The original structural steel grade of the towering ring steel structure is Q235 and Q345, with a bulk density of 78.5kN / m 3 , elastic modulus 206GPa, shear modulus 79GPa, linear expansion coefficient 12×10 -6 (1 / ℃), Poisson's ratio n s =0.3, detailed parameters see Table 1.

[0196]

[0197] Table 1 Design strength values ​​of steel (strength unit: MPa)

[0198] The first twenty modes are extracted by finite element calculation, and the first seven modes are mainly analyzed. The first-order vibration mode is the vibration of bending in the X direction, the second-order vibration mode is the vibration of bending in the Y direction, the third-order vibration mode is the vibration of torsion around the Z axis, the fourth-order vibration mode is the vibration of bending in the X direction of the spire, the fifth-order vibration mode is the vibration of bending in the Y direction of the spire, the sixth-order vibration mode is the vibration of bending in the XZ plane, and the seventh-order vibration mode is the vibration of bending in the YZ plane. The calculation results of the first seven modal frequencies are shown in Table 2.

[0199]

[0200] Table 2. Frequencies of the first seven modes.

[0201] According to the finite element calculation results, taking into account the economic benefits and monitoring effects, the Bayesian optimization algorithm is used to optimize the initial monitoring plan and determine the best monitoring plan. The specific monitoring plan is as follows:

[0202] After the multi-layer ring structure 3 is assembled, the corresponding sensors are installed and the monitoring state is entered. The monitoring points are arranged as follows:

[0203] Arrangement of vibration monitoring points: Combined with the results of finite element analysis, sensors are arranged at key layers to pick up the vibration state of the plane particles at the corresponding positions. Vibration monitoring points are installed on the main beams of the monitoring location floor. Each monitoring point is arranged in the horizontal, longitudinal and vertical directions (perpendicular to each other) of the vibration response test direction. During installation, sufficient coupling agent should be used to couple the sensor with the structure to ensure that the connection between the sensor and the structure is reliable, firm and not easy to slide.

[0204] Stress and strain monitoring point arrangement: According to the finite element calculation results, stress and strain measuring points are arranged at the bottom of the transition layer, the ring structure layout layer and the place where the stiffness changes greatly. Fewer acceleration sensors can be arranged in the lower part of the tower body 2, and more strain monitoring points can be arranged; horizontal acceleration sensors are arranged on the 6th floor side beams; horizontal acceleration sensors are arranged on the 12th floor side beams; horizontal acceleration sensors are arranged on the 18th floor side beams; three-axis acceleration sensors are arranged on the center point and side beams of the 21st floor, strain gauges are arranged on the connecting beams and inclined supports, and a control room is set up; strain gauges are arranged on the 25th floor connecting beams. Strain gauges are arranged in the transition layer to test the stress of the transition structure, and temperature, humidity and anemometers are arranged. Three-axis acceleration sensors and anemometers are arranged on the 16th floor side beams of the spire.

[0205] Arrangement of environmental monitoring points: During this monitoring process, in order to grasp the temperature, humidity and wind speed of the environment in which the structure is located in real time, the temperature, humidity and wind speed should be monitored simultaneously.

[0206] Temperature and humidity measurement points: top of the tower and half of the tower;

[0207] Wind speed measuring points: arranged at the top of the tower, the top of the multi-layer ring structure and half of the tower body.

[0208] Corresponding sensors are installed at each measuring point. When the monitored control parameters change, the change in the sensor forms a proportional relationship with the change in the control parameter. The change in the measuring point is analyzed through the data acquisition and analysis system.

[0209] For the high-rise annular steel structure used in this embodiment, the structural deformation is not large in use, and the lowest frequency is about 0.3Hz, which is also the basis for selecting the vibration pickup. The structural force is simple and clear, and there will be no excessive stress changes and structural disturbances. The effects of wind, temperature and earthquake conditions on the steel structure are mainly considered. On the other hand, the influence of temperature and thermal conductivity on the local components needs to be considered.

[0210] The accelerometer sampling frequency was 5 times of its minimum vibration occupancy rate, and the sampling was continuous and recorded continuously. The strain was recorded 5 times per minute and adjusted according to the calculation and storage situation after 7 days of recording.

[0211] After preprocessing, the collected data are transmitted to the cloud database via the local server. A cloud data website is established to achieve data visualization and real-time remote access to monitoring data.

[0212] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A health monitoring method for a high-rise cantilevered steel structure sightseeing tower, characterized in that: The monitoring steps include the following: Step 1: Use finite element calculation software to establish a three-dimensional finite element model of the towering cantilevered steel structure sightseeing tower, monitor the structural vibration test of the towering cantilevered steel structure sightseeing tower, feedback the monitoring parameters and modify the accuracy of the three-dimensional finite element model, and perform finite element numerical simulation of the towering cantilevered steel structure under different working conditions during the construction and use stage; Step 2: Comprehensively consider the structural characteristics of the tall cantilevered steel structure sightseeing tower and combine engineering experience to preliminarily determine the monitoring plan; Step 3: Comprehensively consider economic benefits and monitoring effect factors, determine the prior distribution probability of the location and number of measuring points, update the posterior distribution probability according to the finite element calculation results of step 1, select the optimal layout plan, and use the Bayesian algorithm to further optimize the monitoring plan determined in step 2; Step 4: According to the finite element calculation results of step 1, combined with relevant industry specifications, set the monitoring threshold, arrange monitoring equipment, collect monitoring data, pre-process the collected monitoring data, obtain the natural frequency according to the finite element calculation, set the low-pass filter cutoff frequency, and remove high-frequency noise; The data collected by the monitoring equipment is transmitted to the local server, and then uploaded to the cloud database through the network for storage and backup. The cloud database is connected to the health monitoring website to observe the data changes of each monitoring point in the structure in real time; Step 5: The cloud database conducts a comprehensive analysis of the data collected by the sensor, and uses a safety warning module with a built-in CNN+LSTM deep learning algorithm model to predict changes in structural response based on changes in current monitoring data. When the predicted value approaches the preset monitoring threshold or the sensor works abnormally, an alarm is issued to provide a safety warning; The specific construction method of using the CNN+LSTM deep learning algorithm model is: Construct an original database, collect field measured data to construct a data set, and use finite element calculation software to analyze the changes in structural response under different states and environmental factors. Use the calculation results to supplement the data set. In the subsequent continuous monitoring process, continuously collect data and supplement the data set. Then divide the data set into training set, verification set, and test set in a ratio of 60%, 20%, and 20%; The normalization method is used to preprocess the original data, the PSO particle swarm optimization algorithm is used to select the optimal input length and output length of the data, and the sliding window method is used to complete the construction of the data set. The expression is: ； in, x norm represents the normalized x value, x Indicates the actual value, x min represents the minimum value of x, x max Indicates the maximum value of x; Construct a CNN+LSTM neural network, using temperature, wind speed, and wind direction environmental variables as input variables to predict changes in structural response over a period of time in the future; Then, the model parameters are adjusted continuously, the model prediction results are compared with the actual results, and the mean absolute error MAE is used as the evaluation standard to select the prediction model with the best prediction effect, and this model is used as the model for subsequent data prediction. The expression is: ； Among them, MAE means mean absolute error; i represents the data number; n represents the total number of data; y i represents the i-th actual value; represents the i-th predicted value.

2. The health monitoring method for a high-rise cantilevered steel structure sightseeing tower according to claim 1 is characterized in that: The three-dimensional finite element model established in the step one is specifically a tall cantilevered steel structure, including a podium, a core tube structure, a multi-layered ring structure, and a spire from bottom to top, wherein the podium is arranged at the bottom of the core tube structure, the multi-layered ring structure is arranged at the top periphery of the core tube structure, and the spire is arranged at the top of the core tube structure.

3. The health monitoring method for a high-rise cantilevered steel structure sightseeing tower according to claim 2 is characterized in that: The specific method of performing finite element numerical simulation on the high-rise cantilever steel structure in the construction and use stage according to the working conditions in step 1 is: The force calculation is carried out for each construction condition of the towering cantilever steel structure during the completion and use stage, including the splicing stage, lifting stage, unloading stage and closing stage of the cantilever structure: In the assembly stage: firstly, analyze and calculate the stress state of the temporary assembly platform built on the roof of the podium building, and then calculate the stress state of the components of the revolving restaurant after assembling them on the platform; During the hoisting stage: after the assembly is completed, the connection between the main structure and the assembly platform and other structures is gradually released, the structure is hoisted, and the changes in the structural stress state of the revolving restaurant from leaving the assembly platform to being lifted to the designed position and spliced ​​with the main structure are calculated; In the unloading stage: after the splicing is completed, the steel strands are unloaded and removed, and the stress changes of the structure before and after the removal of the steel strands are calculated; During the closing stage: after removing the steel strands, gradually remove the temporary reinforcement components and calculate the impact of the removal of each component on the structural stress; Among them, in the cantilever structure splicing stage and cantilever structure closure stage, the implicit static module of the finite element calculation software is used for calculation. During the calculation process, only the gravity load is applied to the whole structure. The two-node linear beam unit B12 is used for the steel structure main body and the components in the revolving restaurant, and the bottom of the main structure adopts full freedom constraint; Among them, during the cantilever structure lifting stage and cantilever structure unloading stage, the transient dynamic module of the finite element calculation software is used for calculation. The two-node linear beam unit B12 is used for the steel structure main body and the components in the revolving restaurant, and the two-node linear three-dimensional truss unit T3D2 is used for the steel strand used for lifting. The load is deadweight, the bottom degree of freedom of the main structure column is fully constrained, and the lifting speed of the steel strand is set to 1mm / s.

4. The health monitoring method for a high-rise cantilevered steel structure sightseeing tower according to claim 3 is characterized in that: The monitoring scheme preliminarily determined in step 2 is as follows: data collection is performed by installing sensors with corresponding functions at different locations of the towering cantilevered steel structure sightseeing tower, specifically including: Set vibration measurement points and arrange vibration sensors to pick up the vibration state of the tower body or the sensor setting location. The arrangement positions are as follows: install acceleration sensors at every H / 10 height along the tower body to pick up the vertical vibration state of the main structure; install acceleration sensors at the multi-layer ring structure and the top of the tower to pick up the plane vibration state of the structure at the corresponding position, where H is the tower body height; Stress and strain measuring points are set up, and stress and strain sensors are arranged to collect structural stress and strain change data. The arrangement positions are as follows: stress and strain measuring points are arranged at intervals of H / 10 along the tower body; stress and strain measuring points are arranged at the rods with larger forces in the conversion truss; stress and strain measuring points are arranged at the splicing position of each layer of multi-layer annular structure and core tube structure, and at the splicing node position of the ring beam; Set up displacement measurement points and arrange displacement sensors to collect structural displacement data; Set up temperature and humidity measurement points, and arrange temperature and humidity sensors to collect temperature and humidity data of the structure itself and the surrounding environment. The arrangement positions are as follows: temperature and humidity sensors are arranged at the top of the tower and at half of the tower; Wind speed and wind direction measuring points are set up, and wind speed and wind direction sensors are arranged to collect wind speed and wind direction data around the structure. The wind speed and wind direction sensors are arranged at the top of the tower, the top of the multi-layer ring structure, and half of the tower body.

5. The health monitoring method for a high-rise cantilevered steel structure sightseeing tower according to claim 4 is characterized in that: The specific method for further optimizing the monitoring scheme in step 3 is: Specifically, the Bayesian algorithm is used to select the optimal layout plan based on the posterior distribution probability: Step 3.1: Define the evaluation index of the model, the expression is: ； Where N is the number of model substructures, i.e., the number of members. a i is the member damage coefficient, which is 0 under standard conditions, and the state index is also 0; c i represents the average decrease in structural performance when the i-th substructure is damaged; the effective stiffness value s i represents the force required to cause damage to the i-th substructure; c is a normalization constant, and T is the matrix transpose symbol; Step 3.2: Using the damage coefficient α and the model prediction error η as variables, calculate the model prediction error. The calculation formula is: ； Among them, G M represents the modal properties predicted by the model; Based on Bayes' theorem, given an observation output And the selected structural model and probability model, calculate the joint posterior probability density function of the unknown parameter α, the calculation formula is: p( | )=cp( | )p( ); Among them, p(α| ) is given The probability density function under the state, c is the normalization constant, p( |α) is the probability density function of the given parameter α, and p(α) is the initial probability; Step 3.3: Choose an uncorrelated zero-mean Gaussian model represented by the covariance matrix ∑η to represent the prediction error η, and simplify the likelihood function to a multivariate normal PDF, expressed as: ； in, represents the probability density function under a given parameter α; T is the matrix transpose symbol; The joint posterior probability density function is asymptotically approximated by a joint Gaussian probability density function centered on the maximum a posteriori probability MAP estimate of α, and the covariance matrix used is , the maximum a posteriori probability MAP estimate is obtained by minimizing the negative logarithm of the posterior PDF, and the calculation formula is: ； Among them, p(α) is the initial probability, MAP refers to the maximum a posteriori probability, and F MAP is the objective function of MAP; Evaluate at the MAP point and the calculation formula is: ； Among them, F MAP represents the objective function, express The maximum a posteriori probability estimate under F MAP is the second-order derivative of the objective function FMAP with respect to parameter α; The formula is derived to obtain: ； in, is the diagonal covariance matrix, is the standard deviation coefficient of λ; N m Indicates the number of natural frequencies; λ r represents the eigenvalue; represents the mode vector; represents the estimated value of α; represents the standard deviation coefficient; diag(δ) represents a matrix with diagonal elements δ, where δ is the sensor configuration vector; Use c T α is used to evaluate the state of the structure, and the uncertainty associated with the state indicator estimate is SD ( ) quantization, where c is the normalization constant, c T is the transposed matrix of c, is the covariance matrix of α, and the optimal degree of freedom is selected from different model degrees of freedom to minimize SD and obtain accurate index estimation in structural state assessment; Step 3.4: Quantify the uncertainty in the updated damage coefficients by using the modal characteristics of the first 10 vibration modes in the Bayesian update process and calculating the standard deviation coefficients of all modes based on engineering experience. and Take 0.1, and let C The solution with the smallest estimated SD value is the optimal sensor layout solution.

6. The health monitoring method for a high-rise cantilevered steel structure sightseeing tower according to claim 5, characterized in that: The specific method of using the CNN+LSTM deep learning algorithm model for state monitoring in step 5 is: Construct the input layer, CNN convolution layer, pooling layer, LSTM layer, and output layer of the CNN+LSTM model respectively; The input signal is standardized and input into the CNN convolution layer. The wide convolution kernel is used to adaptively extract features. The extracted features are pooled by the maximum pooling layer to reduce the data dimension and retain the main feature information. The reduced feature data is then used as the feature input of the LSTM layer to train the neural network and automatically learn the sequence features. Back propagate the training error and gradually update the model parameters layer by layer; Use activation functions to complete regression tasks for multi-feature input sequences.

Images