Wind uplift resistance monitoring method and system for large-span space steel structure roof enclosure system

Abstract

The invention relates to the field of building structure monitoring, and discloses a large-span space steel structure roof enclosure system wind uplift resistance monitoring method and system.The method comprises the steps that a sensing layer is established based on a sensor to monitor roof sensing data, and a transmission layer is established through cooperative work of LoRa and an optical fiber composite cable to transmit the sensing data; an analysis layer is established through a dynamic wind-structure coupling model, a multi-source data fusion algorithm and early warning grading, and dynamic evaluation and risk early warning are carried out; according to the method, historical data is subjected to data preprocessing and feature extraction, Bayesian parameters are updated and deep learning compensation is constructed, sensor errors are adaptively compensated through drift modeling and multi-sensor cross validation compensation, and an early warning threshold value is dynamically adjusted based on the number of samples exceeding a threshold value.

Description

Technical Field

[0001] The present invention relates to the technical field of building structure monitoring, and in particular to a wind-uplift resistance monitoring method and system for a large-span steel structure roof enclosure system. Background Art

[0002] In the existing technology, with the continuous development of construction technology, large-span steel structure roof enclosure systems have been widely used in various large buildings. Large-span steel structure roof enclosure systems have the advantages of large span and open space, but they also face severe challenges in wind-resistant uplift. Under the action of strong winds, the roof enclosure system is easily damaged by wind-resistant uplift, causing damage to the roof and endangering the safety of people in the building. Traditional wind-resistant uplift monitoring methods and systems have many shortcomings, such as incomplete coverage of monitoring points, single sensor type, lagging data processing capabilities, defects in real-time and early warning functions, and poor system stability. They cannot meet the requirements of modern large-span steel structure roof enclosure systems for wind-resistant uplift safety monitoring. It cannot deeply analyze the mechanism of wind action on roof structures, accurately predict the dynamic response of roofs, fail to consider the influence of multiple factors on the probability of damage of calculation nodes, and cannot respond to different risk conditions of roofs in a timely and accurate manner; it cannot intelligently filter out noise in data preprocessing and feature extraction. While making a preliminary health judgment on the roof enclosure system, it is necessary to timely predict future strains. Based on the prediction of future strains, the existing technology cannot timely combine synchronous data to compensate and calibrate sensors, and cannot reasonably adjust the warning threshold dynamically to improve the accuracy and timeliness of warnings; therefore, it is necessary to provide a wind-resistant monitoring method and system for large-span steel structure roof enclosure systems. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for monitoring the wind-uplift resistance of a large-span steel structure roof enclosure system. To solve the above-mentioned problems in the prior art, the present invention is achieved through the following technical solutions:

[0004] In a first aspect, an embodiment of the present invention provides a method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system, comprising the following steps:

[0005] A sensor-based sensing layer is established to monitor roof sensor data, and a transmission layer is established to transmit sensor data using LoRa and fiber optic composite cables.

[0006] Establish an analysis layer through dynamic wind-structure coupling model, multi-source data fusion algorithm and warning classification, dynamic assessment and risk warning;

[0007] Perform data preprocessing and feature extraction on historical data, update Bayesian parameters and build deep learning compensation;

[0008] Sensor errors are adaptively compensated through drift modeling and multi-sensor cross-validation, and the warning threshold is dynamically adjusted based on the number of samples exceeding the threshold.

[0009] In a second aspect, an embodiment of the present invention provides a wind-uplift resistance monitoring system for a large-span steel structure roof enclosure system, comprising the following modules:

[0010] Data perception module: establishes a sensing layer based on sensors to monitor roof sensing data;

[0011] Transmission control module: uses LoRa and optical fiber composite cable to work together to establish a transmission layer to transmit sensor data;

[0012] Modeling and analysis module: establishes an analysis layer through dynamic wind-structure coupling model, multi-source data fusion algorithm and warning classification, and provides dynamic assessment and risk warning;

[0013] Early warning decision module: Based on the dynamic wind-structure coupling model, multi-source data fusion algorithm and the analysis layer established by early warning classification, it makes decisions and issues early warnings for the roof enclosure system across the steel structure;

[0014] Dynamic Optimization Module: Adaptively compensates for sensor errors through drift modeling and multi-sensor cross-validation, and dynamically adjusts the warning threshold based on the number of samples exceeding the threshold.

[0015] Beneficial effects of the present invention:

[0016] 1. The sensing layer features a rational layout of multiple types of sensors, enabling comprehensive and accurate monitoring of key data such as roof strain, wind speed, and vibration frequency. In addition, it provides detailed monitoring of key locations. The transmission layer innovatively uses LoRa and fiber-optic composite cables to work together, balancing low power consumption, long distance, and high sampling rate data transmission requirements to adapt to complex roof environments. The analysis layer uses a dynamic wind-structure coupling model, integrating CFD simulation and finite element modeling, to deeply analyze the mechanism of wind's effect on the roof structure and accurately predict the dynamic response of the roof. The multi-source data fusion algorithm, combined with the Bayesian probability network, considers multiple factors affecting the calculation of node damage probability and implements intelligent graded warnings based on different thresholds, enabling timely and accurate responses to different roof risk conditions.

[0017] 2. In data preprocessing and feature extraction, EEMD is used to decompose the original vibration signal. After adding noise, the signal is averaged through multiple EMDs to obtain the IMF. The noise is filtered out, and the IMF is screened and weighted according to the Pearson correlation coefficient. The root mean square of the vibration acceleration and the extreme strain are calculated to synthesize the fusion health index. When updating Bayesian parameters and performing deep learning compensation, MCMC is used to correct the stiffness matrix of the finite element model based on historical data, and an LSTM neural network is constructed to predict future strain. In drift modeling and multi-sensor cross-validation compensation, the FBG wavelength offset is calculated based on temperature and service time. The residual function is constructed based on the piezoelectric and FBG synchronization data to calibrate the sensor. The warning threshold is dynamically adjusted based on the generalized Pareto distribution and the number of samples exceeding the threshold to improve the accuracy and timeliness of the warning and comprehensively optimize the monitoring system's ability to assess the roof status. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in 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 only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 This is a flowchart of the steps of a method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system provided by Examples 1 and 2 of the present invention;

[0020] Figure 2 This is a structural diagram of a wind-uplift resistance monitoring system for a large-span steel structure roof enclosure system provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention.

[0022] Example 1

[0023] like Figure 1 As shown, an embodiment of the present invention provides a method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system, which specifically includes the following steps:

[0024] Step 1: Establish a sensor layer based on sensors to monitor roof sensor data. Use LoRa and fiber optic composite cables to work together to establish a transmission layer to transmit sensor data. Establish an analysis layer through a dynamic wind-structure coupling model, multi-source data fusion algorithm and early warning classification to dynamically evaluate and issue risk warnings.

[0025] It should be noted that the sensor data include but are not limited to: roof strain, wind speed, temperature and humidity, and vibration acceleration;

[0026] In step one:

[0027] First, the specific process of establishing a sensor layer based on sensors to monitor roof sensor data is as follows:

[0028] In the roof lock edge bite area, a group of FBG sensors are precisely deployed every unit monitoring distance a meter, with an accuracy set to ±1με;

[0029] It should be noted that the locking edge joint area is a key connection part of the roof enclosure system. Under the action of wind, the strain changes frequently. By densely arranging the sensor array, the sudden strain changes can be captured in real time and accurately, providing core data for subsequent analysis.

[0030] Ultrasonic anemometers are installed at the ridge and eaves, and three-dimensional anemometers collect wind speed v, wind direction angle θ and turbulence intensity I in real time. u ;

[0031] It should be noted that the ridge and eaves are located at the edge of the roof and are greatly affected by wind flow. They are the key areas for wind load action. Wind speed directly determines the size of the wind pressure on the roof, wind direction angle determines the direction of wind action, and turbulence intensity reflects the pulsation characteristics of wind speed. Wind speed, wind direction angle and turbulence intensity are extremely critical for accurately assessing wind load action.

[0032] Installed at the support node, the frequency response range is set to 0.1–200 Hz to monitor the vibration acceleration a(t);

[0033] It should be noted that the support nodes support the roof structure. When the wind blows, the roof vibration is transmitted through this node. Monitoring its vibration acceleration can effectively reflect the overall vibration state of the roof and help analyze the stability of the roof structure.

[0034] The second specific process of using LoRa and optical fiber composite cable to work together to establish a transmission layer to transmit sensor data is as follows:

[0035] Adopt LoRa wireless networking and optical fiber composite cable collaborative working mode;

[0036] It should be noted that LoRa wireless networking, with its low power consumption and long-distance transmission characteristics, meets the data transmission requirements of various sensors in the sensing layer; optical fiber composite cables achieve high sampling rate data transmission, ensuring that the acquired data accurately reflects the dynamic changes in roof wind shedding, providing sufficient and timely data support for subsequent analysis. The combined wired and wireless transmission method ensures data transmission stability and adapts to complex roof installation environments.

[0037] For example, the data transmission access mode is determined according to different data transmitted by the sensor, as shown in Table 1:

[0038] Table 1 Statistics of sensor data access methods

[0039] Sensor Type Access method Data characteristics Transmission requirements FBG strain sensor Fiber direct connection High sampling, low latency Fiber switch port aggregation Vibration accelerometer Fiber optic RS485 bus Burst data Fiber Optic Modbus TCP Protocol Temperature and humidity sensors LoRa Wireless Low frequency, small packet LoRa Class C mode Ultrasonic anemometer LoRa+Fiber Redundancy IF, key data Dual-channel hot standby transmission

[0040] Third, the specific process of establishing a dynamic wind-structure coupling model is as follows:

[0041] Integrate dynamic wind loads with roof structure responses, use wind pressure data obtained from the spatiotemporal distribution model of wind pressure, and combine it with the mechanical properties of roof structures to deeply analyze the mechanism of wind action on roof structures, laying a theoretical foundation for accurately predicting roof dynamic responses;

[0042] Specifically, the roof surface wind pressure coefficient C is constructed based on the fusion of CFD simulation and field measurement data. p (x, y, t) distribution function, through the wind pressure coefficient distribution function formula Construct a spatiotemporal distribution model of wind pressure, where ρ represents the roof geometry correction coefficient, v(x,y,t) represents the wind speed at position (x,y) at time t, v0 represents the preset reference wind speed, and I u (x, y, t) represents the turbulence intensity at position (x, y) at time t;

[0043] It should be noted that CFD simulation uses numerical methods to solve the fluid flow control equations and simulate the movement of air around buildings and bridge structures. The roof geometry correction coefficient K is calibrated by CFD simulation to reflect the influence of the specific roof geometry on the wind pressure distribution. Different roof shapes have different airflow distributions under the action of wind. The K value reflects the differences in airflow distribution under the action of wind on different roof shapes. The turbulence intensity I u (x, y, t) reflects the characteristics of wind speed fluctuation. Unstable wind speed fluctuations produce additional dynamic effects on the roof.

[0044] Obtain the mass, damping and stiffness of the roof span steel structure support node based on the finite element model through the formula The roof displacement response u(t) is obtained by solving the equation, where M represents the mass matrix, C represents the damping matrix, and K represents the stiffness matrix.

[0045] It should be noted that the finite element model discretizes the continuous large-span steel structure roof enclosure system into a combination of finite finite elements. The finite elements are connected to each other through nodes. By establishing physical equations for each finite element and assembling them, a set of equations for the entire system is formed.

[0046] Fourth, the specific process of multi-source data fusion algorithm and warning classification is as follows:

[0047] The Bayesian probability network BPN is used to calculate the damage probability P of each node f , based on the temperature sensor to obtain the real-time temperature across the steel structure, through the formula Get the node destruction probability P f , where w i It represents the weight of the i-th sensor. According to the position sensitivity analysis, different position sensors have different effects on the overall wind resistance of the roof. The position sensitivity analysis quantitatively determines the differences and reasonably allocates the weights. crit It represents the critical strain of the edge locking failure, T represents the temperature correction factor, and a represents the vibration acceleration;

[0048] It should be noted that the critical strain for edge-locking failure, calibrated in the laboratory, is a key indicator for determining whether the edge-locking area has failed. The temperature correction factor eliminates the interference of thermal stress on monitoring data. Temperature changes cause thermal expansion and contraction of the roof span steel structure, generating thermal stress, which affects the sensor monitoring results. This factor is introduced to improve data accuracy. The Bayesian Probabilistic Network (BPN) represents a graphical model based on probabilistic reasoning, effectively representing the dependencies between variables and performing probabilistic reasoning.

[0049] Based on the comparison between the obtained node damage probability and the preset damage probability threshold, the warning level is divided;

[0050] Specifically, if the node failure probability is greater than the preset failure probability threshold P0, it indicates that the local lock edge strain exceeds the limit, and a level 1 warning signal is generated. The system prompts manual inspection, and the operation and maintenance personnel go to the corresponding area of the roof to carefully check the abnormal strain and structural damage in the lock edge bite area, providing a basis for subsequent measures.

[0051] If the node damage probability is greater than the preset damage probability threshold P1, it indicates that the multi-node linkage has failed, and a secondary warning signal is generated. The system initiates emergency reinforcement measures for wind-resistant clips. Wind-resistant clips strengthen the roof structure connection strength, improve wind-resistant uplift capabilities, and prevent further damage to the roof structure.

[0052] If the node damage probability is greater than the preset damage probability threshold P2, it indicates that the roof enclosure system is facing the risk of being completely lifted. A level 3 warning signal is generated, and the system triggers an evacuation order to ensure the safety of people in the building and avoid major casualties.

[0053] It should be noted that the preset destruction probability threshold P2>P1>P0, which is set by technicians in the professional field of the present invention based on historical experience and can be adjusted according to specific circumstances;

[0054] The technical solution of the embodiment of the present invention is as follows: the sensing layer rationally arranges multiple types of sensors to comprehensively and accurately monitor key data such as roof strain, wind speed, and vibration frequency, and conducts fine monitoring of key locations. The transmission layer innovatively adopts LoRa and fiber optic composite cables to work together, taking into account the requirements of low power consumption, long distance, and high sampling rate data transmission, and adapting to complex roof environments. The analysis layer uses a dynamic wind-structure coupling model, integrating CFD simulation and finite element modeling, to deeply analyze the mechanism of wind action on the roof structure and accurately predict the dynamic response of the roof. The multi-source data fusion algorithm is combined with the Bayesian probability network to consider the influence of multiple factors on the calculation of node damage probability, and implement intelligent graded warnings based on different thresholds, which can respond to different roof risk conditions in a timely and accurate manner.

[0055] Example 2

[0056] like Figure 1 As shown, the embodiment of the present invention provides a method and system for monitoring the wind-uplift resistance of a large-span steel structure roof enclosure system, which specifically includes the following steps:

[0057] Step 2: Perform data preprocessing and feature extraction on historical data, update Bayesian parameters and build deep learning compensation, adaptively compensate for sensor errors through drift modeling and multi-sensor cross-validation, and dynamically adjust the warning threshold based on the number of samples exceeding the threshold;

[0058] In step 2:

[0059] First, the specific process of data preprocessing and feature extraction for historical data is as follows:

[0060] Perform EEMD on the original vibration signal. By adding N groups of Gaussian white noise with a standard deviation of α times the signal amplitude, M intrinsic mode functions (IMFs) are decomposed to cover the preset frequency band.

[0061] It should be noted that the ensemble empirical mode decomposition (EEMD) is developed on the basis of the empirical mode decomposition (EMD). It adds multiple white noise sequences to the original signal, then performs EMD decomposition on the noise-added signal, and finally averages the decomposition results to eliminate the influence of noise and obtain a more accurate intrinsic mode function (IMF). The empirical mode decomposition (EMD) represents a method of decomposing a complex signal into several intrinsic mode functions, which is mainly used for the analysis of nonlinear and non-stationary signals. The intrinsic mode function (IMF) means that in the entire data sequence, the number of extreme points and the number of zero-crossing points must be equal or differ by at most one. At any time, the upper envelope and lower envelope formed by the local maximum points and local minimum points respectively have a mean of zero.

[0062] When the residual component becomes a monotonic function or the standard deviation is less than the preset standard deviation threshold, the iteration is stopped;

[0063] Use ensemble empirical mode decomposition (EEMD) to filter data noise and eliminate environmental vibration interference;

[0064] Calculate the Pearson correlation coefficient between each intrinsic mode function IMF component and the original signal, The weight of It represents the empirical threshold;

[0065] By formula The intrinsic mode function IMF component after weighted superposition is obtained, where ε clea m represents the intrinsic mode function IMF component, K represents the total number of intrinsic mode function IMF components, and H represents the correlation coefficient, which is

[0066] Collect vibration data in a healthy state and calibrate the reference threshold a ref =μ±3σ, where μ is the mean and σ is the root mean square fluctuation;

[0067] Exemplarily, the total number of intrinsic mode function (IMF) components K=4 is obtained, including: high-frequency component intrinsic mode function (IMF) components, fundamental frequency component intrinsic mode function (IMF) components, low-frequency component intrinsic mode function (IMF) components and residual trend term intrinsic mode function (IMF) components; the Pearson correlation coefficients of the intrinsic mode function (IMF) components are calculated to be: ρ1=0.15, ρ=0.82, ρ3=0.35, ρ4=0.08, and the correlation coefficients of the intrinsic mode function (IMF) components are analyzed and judged to be: H1=0, H2=1, H3=0.5, H4=0; the fundamental frequency component intrinsic mode function (IMF) components and the low-frequency component intrinsic mode function (IMF) components are retained, and the formula The intrinsic mode function IMF component ε after weighted superposition is obtained cleam =IMF2(t)+IMF3(t); Collect 100 sets of vibration data in a healthy state, calculate the root mean square fluctuation σ=0.12 and the mean μ=0.02, and calculate the calibration reference threshold a ref =μ±3σ to obtain the calibration reference threshold a ref Range: [-0.34, 0.38];

[0068] Based on the obtained vibration acceleration a i , through the formula Calculate the root mean square value of vibration acceleration a rms , where N represents the sampling time period;

[0069] Extract strain extreme value ∈ i , count the peak values in the m-second time window, where m represents the preset time window size;

[0070] Combining strain and vibration characteristics, the formula The fusion health index HI is calculated, where a ref It represents the vibration energy benchmark in a healthy state, ∈ crit It represents the mechanical property threshold of the material, and ω1 represents the weight coefficient preset by the system;

[0071] Based on the comparison between the obtained fusion health index and the preset fusion health threshold HI0, the preliminary health status of the roof enclosure system across the steel structure is judged;

[0072] Specifically, based on the calculated fusion health index According to the generalized Pareto distribution GPD and the number of samples exceeding the threshold, HI0 is periodically optimized;

[0073] It should be noted that HI0 represents a preset fusion health threshold, which is calibrated by the professional and technical personnel of the present invention based on historical health data;

[0074] The second specific process is based on updating Bayesian parameters and constructing deep learning compensation:

[0075] Based on the Markov chain Monte Carlo MCMC method, the stiffness matrix K of the finite element model is modified: P(K|D h )∝P(D h |K)*P(K), where D h It represents N sets of historical data sets containing strain, displacement and wind speed, and P(K) represents the prior distribution, which obeys the Weibull distribution;

[0076] It should be noted that the Markov Chain Monte Carlo (MCMC) method uses the properties of the Markov chain to construct a Markov chain whose stationary distribution is the target distribution. By performing random walks on this Markov chain, when the chain reaches a stationary state, the samples obtained obey the target distribution.

[0077] Construct an LSTM neural network, input the historical strain sequence, output the strain forecast value for the next z hours, and use the error backpropagation to correct the model weights;

[0078] For example, based on the calculated historical strain sequence {∈ t-T ,∈ t-T+1 ,...,∈ t}, time window T = 24 hours, sampling interval 10 minutes, construct the input layer; set 2 layers of LSTM units, each with 128 neurons, set the dropout rate to 0.2 to prevent overfitting, and construct the hidden layer; calculate the strain prediction value for the next z = 6 hours: Construct the output layer; use Adam optimizer and set the initial learning rate to 10 -3 , batch size 32, and terminate training when the validation set loss does not decrease for 5 consecutive rounds;

[0079] Third, the specific process of adaptively compensating sensor errors through drift modeling and multi-sensor cross-validation is as follows:

[0080] Obtain the temperature T and service time t of the steel structure roof enclosure through the formula Obtain the fiber Bragg grating (FBG) wavelength offset Δλ, where γ1, γ2, and γ3 are preset correlation coefficients determined by historical data regression;

[0081] It should be noted that FBG refers to a periodic refractive index modulation structure formed in the core of an optical fiber. The wavelength shift of the fiber Bragg grating (FBG) sensor directly reflects the strain and temperature changes of the measured object. It is used to separate the true strain signal from interference factors, ensuring that the monitoring data only reflects the actual stress state of the structure.

[0082] Using piezoelectric sensors to measure strain ε meas Synchronize the time and space data with FBG to construct the residual function R. When |R|>3σ R When the calibration instruction is triggered, σ R It represents the standard deviation of the data;

[0083] Fourthly, the specific process of dynamically adjusting the warning threshold based on the number of samples exceeding the threshold is as follows:

[0084] Get the number of samples exceeding the threshold N u, based on the generalized Pareto distribution GPD through the formula Get the strain extreme value ε in the future time period yr , where μ, σ and ξ are the location, scale and shape parameters of the generalized Pareto distribution GPD respectively;

[0085] Based on the obtained strain extreme value ε yr Perform difference processing with the preset critical strain of edge locking failure to obtain the critical strain difference Δε;

[0086] The critical strain difference is processed by ratio with the critical strain of the edge locking failure to obtain the critical strain deviation value Pc. The warning threshold is dynamically adjusted based on the size of the critical strain deviation value.

[0087] For example, a dynamic adjustment rule for the warning threshold is formulated based on the magnitude of the critical strain deviation value, as shown in Table 2:

[0088] Table 2 Rules for dynamic adjustment of warning thresholds

[0089]

[0090] The technical solution of the embodiment of the present invention is as follows: in data preprocessing and feature extraction, the original vibration signal is decomposed using EEMD, and after adding noise, the IMF is averaged through multiple EMDs to obtain the IMF, the noise is filtered, the IMF is screened and weighted according to the Pearson correlation coefficient, the vibration acceleration root mean square and the strain extreme value are calculated, and a synthetic fusion health index is synthesized. During Bayesian parameter update and deep learning compensation, MCMC is used to correct the finite element model stiffness matrix based on historical data, and an LSTM neural network is constructed to predict future strain. In drift modeling and multi-sensor cross-validation compensation, the FBG wavelength offset is calculated based on temperature and service time, and a residual function is constructed based on piezoelectric and FBG synchronization data to calibrate the sensor. The warning threshold is dynamically adjusted based on the generalized Pareto distribution and the number of samples exceeding the threshold, thereby improving the accuracy and timeliness of the warning and comprehensively optimizing the monitoring system's ability to assess the roof status.

[0091] Example 3

[0092] like Figure 2 As shown, the embodiment of the present invention provides a method and system for monitoring the wind-uplift resistance of a large-span steel structure roof enclosure system, which specifically includes the following modules:

[0093] Data perception module: establishes a sensing layer based on sensors to monitor roof sensing data;

[0094] Transmission control module: uses LoRa and optical fiber composite cable to work together to establish a transmission layer to transmit sensor data;

[0095] Modeling and analysis module: establishes an analysis layer through dynamic wind-structure coupling model, multi-source data fusion algorithm and warning classification, and provides dynamic assessment and risk warning;

[0096] Early warning decision module: Based on the dynamic wind-structure coupling model, multi-source data fusion algorithm and the analysis layer established by early warning classification, it makes decisions and issues early warnings for the roof enclosure system across the steel structure;

[0097] Dynamic Optimization Module: Adaptively compensates for sensor errors through drift modeling and multi-sensor cross-validation, and dynamically adjusts the warning threshold based on the number of samples exceeding the threshold.

[0098] An embodiment of the present invention is described in detail above, but the content described is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention; the above formulas are all dimensionless and numerical calculations, and the formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field based on actual conditions and historical experience, and can be adjusted according to actual conditions; the above description is only a preferred embodiment of the present invention and is not used to limit the present invention. All equal changes and improvements made according to the scope of application of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system, characterized in that: The following steps are involved: A sensor-based sensing layer is established to monitor roof sensor data, and a transmission layer is established to transmit sensor data using LoRa and fiber optic composite cables. Establish an analysis layer through dynamic wind-structure coupling model, multi-source data fusion algorithm and warning classification, dynamic assessment and risk warning; Perform data preprocessing and feature extraction on historical data, update Bayesian parameters and build deep learning compensation; Sensor errors are adaptively compensated through drift modeling and multi-sensor cross-validation, and the warning threshold is dynamically adjusted based on the number of samples exceeding the threshold.

2. The method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system according to claim 1, characterized in that: The specific process of monitoring roof sensor data is as follows: In the roof lock edge bite area, a group of FBG sensors is deployed every unit monitoring distance; Ultrasonic anemometers are installed at the roof ridge and eaves, and three-dimensional anemometers collect wind speed, wind direction angle, and turbulence intensity in real time; Install it at the support node, set the frequency response range, and monitor the vibration acceleration.

3. The method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system according to claim 1, characterized in that: The specific process of establishing a transport layer to transmit sensor data is as follows: It adopts LoRa wireless networking and optical fiber composite cable collaborative working mode.

4. The method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system according to claim 1, characterized in that: The specific process of establishing the dynamic wind-structure coupling model is as follows: The wind pressure coefficient distribution function of the roof surface is constructed based on the fusion of CFD simulation and field measurement data, and the wind pressure spatiotemporal distribution model is constructed through the wind pressure coefficient distribution function formula. The mass, damping, and stiffness of the roof spanning steel structure support nodes are obtained, and the roof displacement response is obtained by solving the formula based on the finite element model.

5. The method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system according to claim 1, characterized in that: The specific process of the multi-source data fusion algorithm and early warning classification is as follows: The Bayesian probability network (BPN) is used to calculate the failure probability of each node. The real-time temperature across the steel structure is obtained based on the temperature sensor, and the failure probability of the node is obtained through the formula. The warning levels are divided based on the comparison between the obtained node damage probability and the preset damage probability threshold.

6. The method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system according to claim 1, characterized in that: The specific process of data preprocessing and feature extraction is as follows: When the residual component becomes a monotonic function or the standard deviation is less than the preset standard deviation threshold, the iteration is stopped; Use ensemble empirical mode decomposition (EEMD) to filter data noise and eliminate environmental vibration interference; The intrinsic mode function IMF after weighted superposition is obtained through the formula; Collect vibration data in a healthy state and calibrate the reference threshold; The root mean square value of the vibration acceleration is calculated based on the obtained vibration acceleration through the formula; The strain and vibration characteristics are comprehensively considered and the fusion health index is calculated through a formula to judge the preliminary health status of the roof enclosure system spanning the steel structure.

7. The method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system according to claim 1, characterized in that: The specific process of updating the Bayesian parameters and constructing deep learning compensation is as follows: Modify the finite element model stiffness matrix K based on the Markov chain Monte Carlo MCMC method; Construct an LSTM neural network, input the historical strain sequence, output the strain forecast value for the next z hours, and use error backpropagation to correct the model weights.

8. The method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system according to claim 1, characterized in that: The specific process of adaptively compensating sensor errors is as follows: Obtain the temperature T and service time t across the steel structure roof enclosure, and use the formula to obtain the FBG wavelength offset; The residual function R is constructed using the spatiotemporal synchronization data of the piezoelectric sensor and the FBG. The calibration command is triggered when |R|>3σ, where σ represents the standard deviation of the data.

9. The method for monitoring wind-uplift resistance of a large-span steel structure roof enclosure system according to claim 1, characterized in that: The specific process of dynamically adjusting the warning threshold is as follows: Get the number of samples exceeding the threshold N u , based on the generalized Pareto distribution GPD, the strain extreme value of the future time period is obtained through the formula; The critical strain difference is obtained by performing difference processing on the obtained strain extreme value and the preset critical strain of the lock edge bite failure; The critical strain difference is ratioed with the critical strain of the edge locking failure to obtain the critical strain deviation value, and the warning threshold is dynamically adjusted based on the size of the critical strain deviation value.

10. A wind-uplift monitoring system for a large-span steel structure roof enclosure system, characterized in that: Includes the following modules: Data perception module: establishes a sensing layer based on sensors to monitor roof sensing data; Transmission control module: uses LoRa and optical fiber composite cable to work together to establish a transmission layer to transmit sensor data; Modeling and analysis module: establishes an analysis layer through dynamic wind-structure coupling model, multi-source data fusion algorithm and warning classification, and provides dynamic assessment and risk warning; Early warning decision module: Based on the dynamic wind-structure coupling model, multi-source data fusion algorithm and the analysis layer established by early warning classification, it makes decisions and issues early warnings for the roof enclosure system across the steel structure; Dynamic Optimization Module: Adaptively compensates for sensor errors through drift modeling and multi-sensor cross-validation, and dynamically adjusts the warning threshold based on the number of samples exceeding the threshold.

Images