A steel temporary bridge construction hidden danger monitoring method and system

Abstract

The application provides a steel temporary bridge construction hidden danger monitoring method and system, and belongs to the technical field of construction safety. The method comprises the following steps: collecting multi-source monitoring data of structure response data and environmental load data through a data acquisition device; preprocessing and feature processing the environmental load data to obtain a target environmental load feature vector; preprocessing the structure response data to obtain target structure response data; using a wavelet transform algorithm to extract features from the target structure response data, and extracting a multi-scale time-frequency feature vector related to wind-induced vibration and wave-induced vibration; inputting the multi-scale time-frequency feature vector and the target environmental load feature vector into a pre-trained hidden danger evaluation model to obtain a preliminary hidden danger grade; correcting the preliminary hidden danger grade based on a pre-set structure performance index system to obtain a hidden danger grade; and inputting the hidden danger grade and sensor health state data into an early warning decision model to obtain a warning instruction. The application improves the accuracy of construction hidden danger monitoring.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of construction safety, in particular to a steel temporary bridge construction hidden danger monitoring method and system. BACKGROUND

[0002] As a core facility for temporary traffic and construction in engineering construction, the steel temporary bridge has the characteristics of temporary structure, dynamic load and complex working condition. During the construction period, the structure is easily affected by the coupling of multiple sources of load such as wind load, wave impact and construction machinery vibration, which may cause stiffness attenuation, node loosening and other hidden dangers. If not timely warning, it may lead to structural instability accidents. Traditional monitoring relies on manual inspection and single sensor data. Manual inspection is difficult to capture the dynamic response of the structure under instantaneous load, and single data is easily disturbed by equipment failure, which cannot accurately associate the cause and effect relationship between load and structural hidden danger, and has problems such as response lag and one-sided hidden danger identification. With the development of Internet of Things and intelligent algorithms, multi-source monitoring technology has been gradually applied. However, how to realize accurate matching of load-response characteristics, effective extraction of time-frequency domain features and dynamic correction of hidden danger level is still the core technical bottleneck of steel temporary bridge construction safety control.

[0003] Therefore, there is an urgent need for a steel temporary bridge construction hidden danger monitoring method and system. SUMMARY

[0004] In order to solve the above technical problems, the application provides a steel temporary bridge construction hidden danger monitoring method and system.

[0005] The first aspect of the embodiment of the application provides a steel temporary bridge construction hidden danger monitoring method, which comprises the following steps:

[0006] A data acquisition device deployed on the steel temporary bridge collects multi-source monitoring data in the construction process, wherein the multi-source monitoring data comprises structural response data and environmental load data;

[0007] The environmental load data is preprocessed and feature processed to obtain a target environmental load feature vector;

[0008] The structural response data is preprocessed to obtain target structural response data;

[0009] The target structural response data is subjected to time-frequency domain decomposition by using a wavelet transform algorithm to obtain a group of sub-band signals of different frequencies;

[0010] From the group of sub-band signals of different frequencies, a multi-scale time-frequency feature vector related to wind-induced vibration and wave-induced vibration is extracted;

[0011] The multi-scale time-frequency feature vector and the target environmental load feature vector are input into a pre-trained hidden danger evaluation model to obtain a preliminary hidden danger level;

[0012] Based on the multi-source monitoring data, a structure performance index system is established, and the preliminary hidden danger grade is corrected to obtain a final hidden danger grade.

[0013] The final hidden danger grade and the obtained sensor health state data are input into a fuzzy logic-based early warning decision model to generate a corresponding early warning instruction.

[0014] In a second aspect, the embodiment of the application provides a steel temporary bridge construction hidden danger monitoring system, which comprises:

[0015] A multi-source monitoring data acquisition module is configured to acquire multi-source monitoring data in a construction process by a data acquisition device deployed on a steel temporary bridge, wherein the multi-source monitoring data comprises structure response data and environmental load data.

[0016] An environmental load data processing module is configured to pre-process and feature-process the environmental load data to obtain a target environmental load feature vector.

[0017] A structure response data preprocessing module is configured to pre-process the structure response data to obtain target structure response data.

[0018] A time-frequency feature vector extraction module is configured to perform time-frequency domain decomposition on the target structure response data by using a wavelet transform algorithm to obtain a group of sub-band signals of different frequencies, and extract multi-scale time-frequency feature vectors related to wind-induced vibration and wave-induced vibration from the group of sub-band signals of different frequencies.

[0019] A preliminary hidden danger grade evaluation module is configured to input the multi-scale time-frequency feature vectors and the target environmental load feature vector into a pre-trained hidden danger evaluation model to obtain a preliminary hidden danger grade.

[0020] Based on the multi-source monitoring data, a structure performance index system is established, and the preliminary hidden danger grade is corrected to obtain a final hidden danger grade.

[0021] An early warning decision instruction generation module is configured to input the final hidden danger grade and obtained sensor health state data into a fuzzy logic-based early warning decision model to generate a corresponding early warning instruction.

[0022] In a third aspect, the embodiment of the application provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the steel temporary bridge construction hidden danger monitoring method when executing the computer program.

[0023] In a fourth aspect, the application provides a computer readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the steel temporary bridge construction hazard monitoring method.

[0024] The steel temporary bridge construction hazard monitoring method and system provided by the application has the beneficial effects that the application takes into account the structural response and the environmental load through multi-source data acquisition, avoiding the monitoring blind area of a single data dimension. The targeted preprocessing of the environmental and structural data solves the problem of difficult capture of vibration characteristics under complex load. The output of the hazard evaluation model is corrected based on the performance index system, improving the reliability of hazard grade determination. At the same time, the early warning model based on fuzzy logic integrates the health status of the data acquisition device, which can avoid misjudgment caused by abnormal data acquisition device, improving the accuracy and practicality of the early warning instruction. The application improves the accuracy of steel temporary bridge construction hazard monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 A flowchart of the steel temporary bridge construction hazard monitoring method provided by an embodiment of the application is shown in the figure.

[0026] Figure 2 A structural block diagram of the steel temporary bridge construction hazard monitoring system provided by an embodiment of the application is shown in the figure.

[0027] Figure 3 A schematic block diagram of the electronic device provided by an embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0028] In the following description, specific details are set forth such as particular system configurations, techniques, etc., in order to provide a thorough understanding of the embodiments of the application. However, it will be apparent to one skilled in the art that the application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the application with unnecessary detail.

[0029] In order to make the objectives, technical solutions and advantages of the application clearer, the following will be combined with the accompanying drawings to make a detailed description. Figures 1-3 The application is described in detail through specific embodiments.

[0030] Reference should be made to Figure 1 , Figure 1 A flowchart of the steel temporary bridge construction hazard monitoring method provided by an embodiment of the application is shown in the figure. The method comprises:

[0031] S101: Collecting multi-source monitoring data in the construction process through the data acquisition device deployed on the steel temporary bridge, the multi-source monitoring data comprising structural response data and environmental load data.

[0032] In this embodiment, according to the structural form, span distribution and key stress nodes of the steel temporary bridge, strain sensors and acceleration sensors are arranged at stress concentration positions such as the midspan of the main beam, the support and the top of the pier, for collecting structural response data such as structural strain, vibration frequency and displacement; at the same time, wind speed and direction instruments, wave height monitors and environmental temperature and humidity sensors are arranged in the surrounding area of the steel temporary bridge to realize the collection of environmental load data such as wind load, wave load change and construction vehicle load. The deployment of the data acquisition equipment needs to meet the protection level requirements and adapt to the complex working conditions of the construction site to avoid affecting the data collection accuracy due to dust, vibration and water vapor erosion.

[0033] Secondly, the structural response data is adjusted in frequency according to the load of the construction stage, specifically, a sampling frequency of 10 Hz is used in the conventional construction stage, and the sampling frequency is increased to 50 Hz under special working conditions such as extreme weather, so that the sudden change of structural response under instantaneous load can be captured; the environmental load data is continuously collected according to the minute level, and through the time stamp synchronization technology, a unified time reference of multi-source monitoring data is obtained.

[0034] S102: Preprocessing and feature processing are performed on the environmental load data to obtain a target environmental load feature vector.

[0035] In this embodiment, first, data outliers are identified and corrected, for example, through statistical methods such as 3σ criterion and Grubbs test, abnormal data points caused by sudden fluctuations such as electromagnetic interference are screened out; linear interpolation method is used to complete the occasional outliers, and for continuous abnormal sections, they are repaired based on the data of adjacent sensors at the same period, and missing data segments caused by equipment power failure and transmission interruption during data collection are also removed.

[0036] Secondly, data standardization and smoothing processing are carried out on the environmental load data, specifically, through Z-score standardization or minimum-maximum standardization method, all kinds of data are mapped to a unified numerical interval to eliminate the interference of dimension difference on subsequent feature extraction; at the same time, moving average filtering, Gaussian filtering and other algorithms are used to smooth the standardized data, filter high-frequency random noise and retain the change trend of environmental load data.

[0037] Subsequently, based on the stress characteristics of the steel temporary bridge structure, targeted feature processing is carried out. For example, for wind load data, the average wind speed, maximum wind speed, wind speed standard deviation, gust factor in the time domain, and the peak frequency of wind speed power spectrum, energy proportion in the frequency domain, and other features are extracted; for wave load data, the effective wave height, wave period, wave surge intensity, wave energy flux, and other core features are extracted; for construction vehicle load data, the load peak value, load duration, load frequency, load distribution mean and variance, and other features are extracted, and these features are classified to obtain time domain features, frequency domain features, and coupling features. Among them, the time domain features include the average wind speed, maximum wind speed, wind speed standard deviation, gust factor of wind load data in the time domain, the effective wave height, wave period, wave surge intensity, wave energy flux of wave load data, and the load peak value, load duration, load frequency, load distribution mean and variance of construction vehicle load data; the frequency domain features include the peak frequency of wind speed power spectrum, energy proportion of wind load data in the frequency domain; the coupling features are based on the correlation coefficient, time sequence synchronization index of wind-wave, wind-construction vehicle, wave-construction vehicle, etc. different load data types, which are used to quantify the synergy strength between different loads.

[0038] Finally, the correlation coefficient and time sequence synchronization index between different environmental loads are calculated to quantify the synergy strength of wind-wave, wind-vehicle, wave-vehicle, etc. load combination. The time domain features, frequency domain features and coupling features are integrated, and the vector normalization processing is carried out based on the preset dimension to obtain the target environmental load feature vector.

[0039] S103: Preprocess the structure response data to obtain target structure response data.

[0040] In this embodiment, first, the data outliers of the structure response data are identified and processed. Specifically, the abnormal mutation points in the structure response data which are prone to be caused by factors such as sensor loosening, electromagnetic interference, and strong vibration of construction machinery are determined by the box plot method and the interquartile range criterion, and the abnormal points are obtained according to the normal fluctuation range. If the abnormal point is an isolated point, the adjacent data interpolation method is used for correction, and if the abnormal point is a continuous abnormal segment, whether it is removed is determined based on the sensor health state data. Secondly, the missing values of the structure response data are completed and synchronized. For example, the complex working conditions of the steel temporary bridge construction site are prone to cause data transmission interruption, so for short-time missing, linear interpolation or spline interpolation is used for completion; for long-time missing, the mean model of the historical same period data under the same working condition is called to fit and complete, and the data segment that cannot be repaired is removed. Among them, the data transmission interruption within 5 sampling points leads to short-time missing, and vice versa.

[0041] In addition, due to the difference in the sampling frequency of different types of sensors, the synchronization calibration of the structural response data is completed based on the time stamp, and all the structural response data is unified to the same time resolution, so that the consistency of the structural response data in the time sequence dimension is ensured.

[0042] Finally, the structural response data is denoised. The high-frequency interference signals such as construction machinery vibration and environmental noise in the structural response data are denoised. Specifically, for the acceleration vibration data, the wavelet threshold denoising method is used to separate the effective signal and the noise by selecting the appropriate wavelet basis function and the decomposition level. For the strain and displacement data, the low-pass filtering algorithm is used to filter the high-frequency interference and retain the true trend of the structural deformation. After the completion of the denoising, the target structural response data is obtained.

[0043] S104: The wavelet transform algorithm is used to perform time-frequency domain decomposition on the target structural response data to obtain a group of sub-band signals with different frequencies.

[0044] From the group of sub-band signals with different frequencies, the multi-scale time-frequency feature vectors related to wind-induced vibration and wave-induced vibration are extracted.

[0045] In this embodiment, the wavelet basis function is selected according to the spectral characteristics of the target structural response data and the monitoring requirements, and the target wavelet basis function is obtained. Then, based on the Nyquist sampling theorem and the natural frequency of the steel bridge, the wavelet decomposition level is determined, and the target structural response data is decomposed based on the target wavelet basis function and the wavelet decomposition level to obtain the approximation coefficients and the detail coefficients. The group of sub-band signals with different frequencies includes the approximation coefficients and the detail coefficients. The specific steps of selecting the wavelet basis function include: selecting the candidate wavelet basis function (such as Db4, Morlet, etc.); calculating the sub-band energy entropy after decomposition of each wavelet basis function, and selecting the basis function with the smallest entropy; calculating the root mean square error of the reconstructed signal and the original signal, and selecting the basis function with the smallest error; and comprehensively determining the optimal wavelet basis function.

[0046] Secondly, the multi-scale time-frequency features strongly related to wind-induced and wave-induced vibration are selected and extracted from the group of sub-band signals with different frequencies. Specifically, based on the structural dynamics characteristics of the steel bridge and the engineering measured data, it is found that the wind-induced vibration is mainly concentrated in the low frequency band of 0.1-1Hz and the secondary low frequency band of 2-5Hz, and the wave-induced vibration is mainly distributed in the extremely low frequency band of 0.05-0.5Hz due to the difference in wave period. Accordingly, the frequency interval matching of the sub-band signals is performed, and the high-frequency noise sub-band unrelated to wind and wave load is eliminated, such as the construction machinery vibration sub-band above 10Hz, to obtain the multi-scale time-frequency feature vectors related to wind-induced vibration and wave-induced vibration.

[0047] S105: The multi-scale time-frequency feature vectors and the target environmental load feature vectors are input into the pre-trained hidden danger assessment model to obtain the preliminary hidden danger grade.

[0048] Based on multi-source monitoring data, a structure performance index system is established, and the preliminary hidden danger grade is corrected to obtain the final hidden danger grade.

[0049] In this embodiment, the pre-trained hidden danger evaluation model is trained and verified based on the historical data of steel bridge structure health monitoring. The hidden danger evaluation model architecture adopts a hybrid model of gradient boosting decision tree and deep neural network: the gradient boosting decision tree module is responsible for capturing the nonlinear correlation and hierarchical logic between features, and the deep neural network module realizes the deep fitting of the complex load-vibration coupling relationship. In the training stage of the hidden danger evaluation model, a labeled data set including different hidden danger grades needs to be constructed. The labeled data set includes the structure vibration features and the corresponding hidden danger state under different wind and wave load conditions. At the same time, the model hyperparameters are optimized through k-fold cross-validation.

[0050] For example, first determine the reasonable selection of k value, for example, based on the sample size and working condition diversity of the steel bridge monitoring data set, select the cross-validation scheme of k=5 or k=10, randomly and uniformly divide the labeled data set into k mutually exclusive subsets, of which k-1 mutually exclusive subsets are used as the training set for model fitting, and the remaining 1 mutually exclusive subset is used as the validation set to evaluate the performance of the hidden danger evaluation model.

[0051] For different modules of the hybrid model architecture, the specific range and target of hyperparameter optimization need to be determined: for the gradient boosting decision tree module, the core hyperparameters such as decision tree depth, learning rate, and decision tree number are optimized, and the goal is to balance the model fitting accuracy and overfitting risk, where the decision tree depth is set to a adjustable interval of 2-10; the learning rate is set to an adjustable interval of 0.01-0.3, and the decision tree number is set to an adjustable interval of 50-200; for the deep neural network module, the number of hidden layers, the number of neurons in each layer, the learning rate, and the batch size are optimized, where the number of hidden layers is adjustable in the range of 2-5 layers; the number of neurons in each layer is adjustable in the range of 32-128; the learning rate is adjustable in the range of 0.001-0.01; and the batch size is adjustable in the range of 16-64.

[0052] In the k-fold cross-validation process, the hidden danger grade classification accuracy, recall rate, and F1 score of the hidden danger evaluation model are used as evaluation indicators for hyperparameter optimization. Through grid search or Bayesian optimization algorithm, the preset hyperparameter combinations are traversed, and the parameter combination with the best comprehensive performance on the validation set is selected. At the same time, to avoid overfitting of the validation set, after optimization, the model performance is verified based on an independent test set, so that the optimized hyperparameters can maintain stable evaluation accuracy of the hidden danger evaluation model in new working condition data not involved in training, thereby improving the reliability of hidden danger judgment of steel bridges under different wind and wave load conditions.

[0053] After the feature vector is input into the hidden danger assessment model, a preliminary hidden danger level is output through multi-layer feature mapping and probability calculation. Specifically, the hidden danger assessment model first performs layer-by-layer feature extraction on the fused features, identifies abnormal association patterns of wind-induced or wave-induced vibration features and environmental load features, such as wind speed exceeding the limit corresponding to sudden changes in vibration frequency, and sudden increase in wave height corresponding to abnormal structure strain; then, based on the pre-set hidden danger judgment threshold, the feature abnormality degree is quantitatively scored, and the scoring results are mapped to the corresponding hidden danger level interval through a probability weighting algorithm; finally, the preliminary hidden danger level is obtained.

[0054] After obtaining the preliminary hidden danger level, the design specifications and safety control standards during the construction stage of the steel temporary bridge are used to determine the benchmark control limit of the structure performance, and a finite element calculation model of the steel temporary bridge during the construction stage is established, and its theoretical response value is calculated. Then, the benchmark control limit and the theoretical response value of the finite element calculation model are fused to obtain a structure performance index system. Then, according to the statistical characteristics of the multi-source monitoring data and the structure performance index system, the preliminary hidden danger level is quantitatively corrected to obtain the final hidden danger level.

[0055] S106: input the final hidden danger level and the obtained sensor health state data into the early warning decision model based on fuzzy logic to generate corresponding early warning instructions.

[0056] In this embodiment, first, the final hidden danger level is divided according to engineering specifications and historical accident data to obtain four fuzzy subsets of no hidden danger, mild hidden danger, moderate hidden danger, and severe hidden danger, and the quantified hidden danger level value is mapped to the membership degree of the corresponding subset through a triangular membership function. According to the running stability of the data acquisition equipment, the sensor health state data is divided into four fuzzy subsets of normal, mild abnormality, moderate abnormality, and severe abnormality, and based on the signal transmission success rate, data deviation rate and other indicators of the sensor, a trapezoidal membership function is used to complete the fuzzification conversion to eliminate the dimensional and semantic differences between the two types of data.

[0057] The construction of the early warning decision model based on fuzzy logic is based on a fuzzy reasoning rule base of the steel temporary bridge project. The fuzzy reasoning rule base is formulated based on the safety control requirements of the steel temporary bridge structure and the reliability constraints of the sensor data. For example, when the final hidden danger level is severe hidden danger and the sensor health state is normal, a first-level emergency warning is triggered; when the hidden danger level is moderate hidden danger but the sensor health state is moderate abnormality, a second-level pending verification warning is triggered, with additional sensor data review instructions; when the hidden danger level is mild hidden danger and the sensor state is mild abnormality, a third-level attention warning is triggered. The fuzzy reasoning rule base includes all combination scenarios of the input fuzzy subsets, and the rules are verified and supplemented through expert experience method and historical case backstepping method to ensure the rigor and practicality of the reasoning logic.

[0058] After completing the fuzzy inference, the fuzzy membership is converted into specific and executable early warning instructions. Specifically, the barycenter method is used as the defuzzification algorithm to calculate the barycenter value of the fuzzy output set, and based on the barycenter value interval, it is mapped to the corresponding early warning level and operation instruction: the first level emergency early warning corresponds to immediately suspending construction, evacuating on-site personnel, and starting the emergency repair plan; the second level verification early warning corresponds to suspending large equipment access, arranging technical personnel to review sensor data and structure state; the third level attention early warning corresponds to strengthening monitoring frequency and closely tracking the development trend of hidden dangers; and no hidden danger corresponds to maintaining regular construction monitoring and no special disposal is needed.

[0059] wherein,

[0060] From the above, it can be concluded that the present application takes into account the structure response and environmental load through multi-source data acquisition, avoiding the monitoring blind area of a single data dimension. The targeted preprocessing of environmental and structural data solves the problem of difficult capture of vibration characteristics under complex load; and the output of the hidden danger evaluation model is corrected based on the performance index system, improving the reliability of hidden danger level determination. At the same time, the early warning model based on fuzzy logic integrates the health status of the data acquisition device, which can avoid misjudgment caused by abnormal data acquisition device, and improve the accuracy and practicality of the early warning instruction. The present application improves the accuracy of steel bridge construction hidden danger monitoring.

[0061] In an embodiment of the present application, wavelet transform algorithm is used to perform time-frequency domain decomposition on target structure response data to obtain a group of sub-band signals of different frequencies, including:

[0062] According to the frequency spectrum characteristics of the target structure response data and the monitoring requirements, a wavelet base function is selected to obtain a target wavelet base function;

[0063] Based on the Nyquist sampling theorem and the natural frequency of the steel bridge, the number of wavelet decomposition layers is determined;

[0064] Based on the target wavelet base function and the number of wavelet decomposition layers, the target structure response data is decomposed at multiple scales to obtain approximation coefficients and detail coefficients;

[0065] Among them, the group of sub-band signals of different frequencies includes approximation coefficients and detail coefficients.

[0066] In the embodiment, before the multi-scale analysis of the steel bridge target structure response data, the selection of the target wavelet basis function is completed based on the spectral characteristics of the target structure response data and the monitoring requirements. Specifically, the selection of the wavelet basis function takes into account the time domain localization ability and the frequency domain resolution: for strain signals, the Daubechies series wavelet is preferred, and its orthogonality can ensure the independence of the decomposed data; for vibration and displacement signals, the Morlet wavelet or the Mexican Hat wavelet is selected, and the compactness and oscillation characteristics of such wavelets can capture instantaneous response mutations; at the same time, according to the monitoring requirements of the steel bridge construction, if it is necessary to preserve the physical meaning of the data, it is also necessary to verify the matching degree of the wavelet basis function and the structure response signal, and then the optimal wavelet basis function is determined by calculating the wavelet entropy, reconstruction error and other indicators, and finally the target wavelet basis function suitable for the steel bridge monitoring scene is obtained. For example, the candidate wavelet basis function is used to perform j-layer multi-scale decomposition on the target structure response data x(t), and 1 group of approximation coefficients Aj and j groups of detail coefficients D1, D2, …, Dj are obtained. The energy of each sub-band coefficient is calculated respectively:

[0067] wherein k is the sub-band number (k=0 corresponds to the approximation coefficient Aj, and k=1, 2, …, j corresponds to the detail coefficient D1-Dj), is the i-th coefficient of the k-th sub-band, is the length of the sub-band coefficient.

[0068] The energy probability distribution is obtained by normalizing the energy of each sub-band The wavelet entropy H is calculated using the Shannon entropy definition:

[0069] If =0, then =0. The smaller the wavelet entropy H, the more concentrated the time-frequency energy distribution of the signal under the wavelet basis function, that is, the higher the recognition degree of the signal characteristics in a specific time-frequency region, and the stronger the adaptability of the wavelet basis function to the target structure response data. The target wavelet basis function suitable for the steel bridge monitoring scene is obtained.

[0070] Based on the approximation coefficients and the detail coefficients of each layer obtained by wavelet basis function decomposition, the inverse wavelet transform is performed to complete the reconstruction of the original signal, and the reconstructed signal is generated; then the root mean square error or the normalized mean square error is used to quantify the deviation degree of the reconstructed signal and the original signal, wherein the root mean square error is the square sum of the difference between the original signal and the reconstructed signal at each sampling point, and the normalized mean square error is the ratio of the square sum to the square sum of the deviation of the original signal from its mean value. The smaller the root mean square error and the normalized mean square error, the better the fitting effect of the wavelet basis function on the signal, and the lower the loss degree of the characteristic information.

[0071] Based on the Nyquist sampling theorem and the inherent frequency of the steel bridge, the wavelet decomposition layer is determined, and after obtaining the target wavelet basis function and the decomposition layer, the target structure response data is subjected to multi-scale decomposition. In the decomposition process, the scale separation of the signal is realized through the low-pass and high-pass filtering of each layer: each layer of decomposition splits the current scale signal into an approximate coefficient representing the low-frequency trend and a detail coefficient representing the high-frequency detail, wherein the approximate coefficient of the low-frequency trend corresponds to the overall vibration law of the signal, such as the long-term wind-induced vibration trend; the detail coefficient of the high-frequency detail corresponds to the local vibration mutation, such as the surge impact or local stress fluctuation of the structure. Secondly, after the iterative decomposition of the determined wavelet decomposition layer, a group of approximate coefficients and a plurality of groups of detail coefficients can be obtained, which together constitute a group of sub-band signals of different frequencies.

[0072] From the above, it can be seen that the embodiment avoids feature distortion caused by insufficient adaptability of the wavelet basis function by selecting the target wavelet basis function according to the data frequency spectrum characteristics and the monitoring requirements; and the decomposition layer is determined based on the Nyquist sampling theorem and the inherent frequency of the steel bridge, which not only ensures the effective separation of key frequency components such as wind-induced and wave-induced vibrations, but also avoids the calculation redundancy of excessive decomposition and the feature omission of under-decomposition; and the finally generated approximate coefficient and detail coefficient sub-band signals improve the accuracy and reliability of subsequent hidden danger evaluation.

[0073] In an embodiment of the present application, based on the Nyquist sampling theorem and the inherent frequency of the steel bridge, the wavelet decomposition layer is determined, comprising:

[0074] The highest frequency of the power spectral density is determined by analyzing the sampling frequency of the multi-source monitoring data according to the Nyquist sampling theorem;

[0075] The first order inherent frequency of the key inherent frequency of the steel bridge is obtained, and the first order inherent frequency is set as the target frequency;

[0076] The highest frequency and the target frequency are calculated based on the formula N_min=ceil(log2(f_max / f_target)) to obtain the theoretical minimum decomposition layer; wherein f_max is the highest frequency, f_target is the target frequency, and ceil is the upward rounding function;

[0077] On the basis of the theoretical minimum decomposition layer, a predetermined engineering margin is added to obtain the wavelet decomposition layer.

[0078] In the embodiment, according to the Nyquist sampling theorem, the sampling frequency is greater than 2 times the highest frequency of the signal, so as to avoid frequency aliasing and completely restore the original signal characteristics. Specifically, for the structural response data and environmental load data of the steel temporary bridge, the actual sampling frequency is first extracted, for example, the sampling frequency of the structural response data in the conventional construction stage is 10 Hz, and is increased to 50 Hz under extreme working conditions, and the environmental load data is continuously collected at a minute level. According to the Nyquist sampling theorem, the highest frequency of the power spectral density is 1 / 2 of the sampling frequency, so the highest frequency of the power spectral density of the structural response data under the conventional working condition is 5 Hz, and the highest frequency of the power spectral density under the extreme working condition is 25 Hz; the sampling frequency of the environmental load data is extremely low, and the highest frequency of the power spectral density can be ignored, and finally the highest frequency of the power spectral density is 25 Hz corresponding to the 50 Hz sampling frequency based on the highest sampling frequency of the structural response data.

[0079] After the steel temporary bridge is built or in the construction process, the structural modal parameters measured by field test means or the first-order natural frequency obtained by theoretical modal calculation through numerical simulation before the steel temporary bridge is constructed or put into use are taken as the target frequency. Secondly, based on the formula: N_min = ceil (log2 (f_max / f_target)) calculation, the theoretical minimum decomposition layer number is obtained. Wherein f_max is the highest frequency, f_target is the target frequency, and ceil is the upward rounding function;

[0080] Due to the uncertain factors such as electromagnetic interference and equipment vibration in the construction site of the steel temporary bridge, only the theoretical minimum decomposition layer number cannot completely adapt to the signal analysis requirements under complex working conditions. Therefore, an engineering margin is added on the basis of the theoretical minimum decomposition layer number. The value of the engineering margin is determined based on the complexity of the construction environment of the steel temporary bridge, the accuracy of the sensor and the level of data noise, and is usually 1-2 layers.

[0081] From the above, it can be seen that, in the embodiment, the Nyquist sampling theorem is used to define the highest frequency of the power spectral density, so as to avoid the risk of frequency aliasing; the theoretical minimum number of layers is obtained by quantitative calculation, so as to avoid feature loss caused by insufficient decomposition and prevent the calculation load from increasing due to redundant decomposition layers; and the engineering margin is added to compensate for the deviation of the site working condition and improve the adaptability to the complex construction environment.

[0082] In an embodiment of the present application, a steel temporary bridge construction hidden danger monitoring method and system further comprises:

[0083] The adjustment of the highest frequency comprises:

[0084] The power spectral density of the target structural response data in the preset time window is calculated.

[0085] identify all energy peaks in the power spectral density to obtain the energy peaks and record the corresponding frequencies;

[0086] compare the energy peaks with a preset first threshold value;

[0087] If the energy peak is greater than the first threshold value, the highest frequency is updated based on the energy peak, otherwise, the highest frequency remains unchanged,

[0088] The adjustment of the target frequency includes:

[0089] Obtain the target structure response data in the time window, and re-identify the target structure response data by the random subspace identification method or the frequency domain decomposition method to obtain the actual first-order natural frequency of the steel temporary bridge;

[0090] Calculate the absolute value of the deviation between the actual first-order natural frequency and the currently used target frequency;

[0091] If the absolute value of the deviation is greater than a preset second threshold value, the target frequency is updated based on the actual first-order natural frequency, otherwise, the target frequency remains unchanged.

[0092] In this embodiment, in the wavelet decomposition process of the steel temporary bridge structure response data, a time window with a preset length, for example, 10 minutes or 30 minutes, is selected, the power spectral density of the target structure response data in the time window is calculated, the time domain signal is converted into the frequency domain signal by Fourier transform, and the energy distribution corresponding to different frequencies is obtained. Subsequently, all energy peaks in the power spectral density curve are identified, the frequency values corresponding to each peak are marked, and then the energy peaks are compared with a preset first threshold value. The first threshold value is determined based on factors such as sensor noise level and environmental interference intensity, if an energy peak is greater than the first threshold value, the highest frequency of the power spectral density is updated to the frequency corresponding to the peak, if all energy peaks are less than the first threshold value, no adjustment is needed, and the original highest frequency remains unchanged.

[0093] The embodiment adjusts the target frequency based on structural response data. Specifically, target structural response data in a time window is extracted, and modal parameter identification is performed on the signal using a random subspace identification method or a frequency domain decomposition method: the random subspace identification method solves the eigenvalue to obtain the structural natural frequency by constructing a state space model of the data matrix; the frequency domain decomposition method directly extracts the natural frequency using the peak value of the power spectral density function, and both methods can realize identification of the actual first-order natural frequency of the steel bridge. Then, the absolute value of the deviation between the actual first-order natural frequency and the currently used target frequency is calculated, and compared with a preset second threshold value. The second threshold value is set according to the accuracy requirement of the steel bridge monitoring, and the second threshold values corresponding to different types of steel bridges are different. If the absolute value of the deviation is greater than the second threshold value, the target frequency is updated to the actual first-order natural frequency; if the absolute value of the deviation is less than or equal to the second threshold value, the current target frequency can be maintained.

[0094] From the above, it can be seen that the embodiment avoids the omission of the effective frequency band or the redundancy of the invalid frequency band caused by the fixed highest frequency by monitoring the energy peak value of the power spectral density of the structural response data in the time window; and the first-order natural frequency is calibrated in real time by the random subspace identification method or the frequency domain decomposition method, which can dynamically match the natural frequency offset of the steel bridge caused by load changes and structural state changes, so that the target frequency always matches the structural dynamics characteristics; at the same time, the threshold value is used to determine the on-demand update of the target frequency and the highest frequency, which not only ensures the scientificity and timeliness of the parameter, but also avoids the calculation fluctuation caused by frequent adjustment.

[0095] In an embodiment of the present application, after extracting the multi-scale time-frequency feature vectors related to wind-induced vibration and wave-induced vibration from a group of sub-band signals of different frequencies, the following steps are included:

[0096] Comparing the multi-scale time-frequency feature vectors with a historical reference library;

[0097] If the multi-scale time-frequency feature vectors do not deviate from the normal range in the historical reference library, the multi-scale time-frequency feature vectors are determined as normal data;

[0098] If the multi-scale time-frequency feature vectors deviate from the normal range in the historical reference library, the multi-scale time-frequency feature vectors are determined as preliminary abnormal data.

[0099] In this embodiment, after the multi-scale decomposition of the steel bridge structure response data and the feature vector extraction are completed, the multi-scale time-frequency feature vector is compared with the historical benchmark library. The historical benchmark library is constructed based on long-term monitoring data of the steel bridge in the healthy state, including the normal range of multi-scale time-frequency feature vectors in different construction stages and different environmental load conditions. The data includes typical scenes such as normal construction, extreme weather, and dense vehicle traffic, and the normal fluctuation range of each feature dimension is determined by statistical analysis method.

[0100] In the comparison process, the real-time extracted feature vector is compared with the feature vector in the corresponding condition in the historical benchmark library, and the cosine similarity, Euclidean distance and other quantitative indicators between the vectors are calculated to evaluate the degree of agreement between the real-time feature vector and the normal feature vector in the benchmark library. Specifically, if all dimension components of the multi-scale time-frequency feature vector are within the normal range of the historical benchmark library, and the overall similarity index is greater than the preset agreement threshold, the multi-scale time-frequency feature vector is determined to be normal data. If at least one dimension of the multi-scale time-frequency feature vector is greater than the normal fluctuation range of the historical benchmark library, or the overall similarity index is less than the agreement threshold, the multi-scale time-frequency feature vector is determined to be preliminary abnormal data. The value of the agreement threshold takes into account the matching degree of the feature vector and the engineering safety control requirements, and updates the agreement threshold based on the construction stage of the steel bridge.

[0101] From the above, it can be concluded that the embodiment realizes the standardized verification of the multi-scale time-frequency feature vector based on the normal feature range of multiple conditions in the historical benchmark library, avoids misjudgment caused by single dimension feature fluctuation, and guarantees the objectivity and consistency of state judgment. Through the binary judgment logic of normal range matching, the normal fluctuation and preliminary abnormality of the structure response data can be distinguished, the construction safety risk caused by the failure to identify the hidden damage of the structure in time is effectively avoided, and the timeliness and reliability of the health monitoring of the steel bridge during construction are improved.

[0102] In an embodiment of the present application, after the multi-scale time-frequency feature vector deviates from the normal range in the historical benchmark library, the multi-scale time-frequency feature vector is determined to be preliminary abnormal data, and further comprising:

[0103] Matching the target environmental load feature vector with the historical environmental load interval corresponding to the similar abnormal feature vector in the historical benchmark library;

[0104] If the target environmental load is within the historical environmental load interval, the preliminary abnormal data is determined to be real abnormal data; otherwise, it is determined to be observation data;

[0105] Calculating the Mahalanobis distance between the real abnormal data and the corresponding normal feature vector center point in the historical benchmark library as the deviation.

[0106] When the deviation degree is greater than the preset first threshold, the corresponding real abnormal data is marked as key attention data.

[0107] In the embodiment, first, similar abnormal feature vector records archived in the historical reference library are called, and the corresponding environmental load interval when the abnormalities occur is extracted, for example, the key parameter range of wind load, wave load, and construction vehicle load; then the target environmental load feature vector is matched with the historical environmental load interval, and the consistency of the two in core dimensions such as load peak value, action duration, and coupling strength is compared. If the target environmental load is completely within the historical environmental load interval, it indicates that the abnormal feature vector is not caused by extreme fluctuation of the environmental load, and then it is determined that the preliminary abnormal data is real abnormal data; if the target environmental load is greater than the historical environmental load interval, it indicates that the abnormality is a temporary response fluctuation caused by rare environmental load working conditions, and then it is determined to be observation data.

[0108] In the embodiment, the influence degree of the real abnormal data on the structure health state is evaluated by quantifying the deviation degree. Specifically, the Mahalanobis distance between the multi-scale time-frequency feature vector corresponding to the real abnormal data and the center point of the normal feature vector under the same working condition in the historical reference library is calculated. The distance can eliminate the interference of dimension difference and correlation of each feature dimension, and represent the deviation degree of the abnormal vector from the normal reference. The deviation degree is compared with the preset first threshold: if the deviation degree is less than or equal to the first threshold, it indicates that the structure state of the steel bridge is abnormal but still within a controllable range; if the deviation degree is greater than the first threshold, it indicates that there is a potential safety hazard, and then the real abnormal data is marked as key attention data. The determination logic of the first threshold is based on the load-feature correlation law of the steel bridge during construction, the monitoring data error characteristics, and the engineering safety control requirements.

[0109] From the above, it can be seen that, in the embodiment, the matching of the target environmental load feature vector and the historical abnormal working condition load interval distinguishes between false abnormalities caused by environmental load fluctuation and real abnormalities caused by changes in the structure itself, avoids misjudgment caused by non-structural factors, and ensures the reliability of the abnormality warning. In addition, the deviation degree of the real abnormal data from the normal reference is quantified based on the Mahalanobis distance, which realizes scientific quantification of the abnormality degree. The key attention data is divided by the deviation degree and the first threshold, which improves the quality of the multi-source monitoring data.

[0110] In an embodiment of the present application, a steel bridge construction hidden danger monitoring method further comprises:

[0111] The target environmental load feature vector is input into a predefined load-feature relationship model to obtain a predicted time-frequency feature vector;

[0112] a similarity between the multi-scale time-frequency feature vector and the predicted time-frequency feature vector is calculated as the data credibility;

[0113] When the data credibility is greater than or equal to a preset second threshold, it is determined that the corresponding data is credible data.

[0114] When the data credibility is less than the second threshold, it is determined that the corresponding data is untrustworthy data.

[0115] The health status of the data acquisition device and the data acquisition link corresponding to the untrustworthy data is checked to obtain a health status list.

[0116] In the embodiment, the predefined load-feature relationship model is trained based on a large amount of environmental load and structural time-frequency feature correlation data under the health state of the steel temporary bridge, and the structural time-frequency feature law under the corresponding working condition can be mapped through the environmental load feature vector. For example, the target environmental load feature vector is input into the load-feature relationship model, and the load-feature relationship model outputs the predicted time-frequency feature vector under the load working condition based on the built-in correlation algorithm. Then, the similarity between the time-frequency feature vector and the predicted time-frequency feature vector of the load-feature relationship model is compared by calculating the cosine similarity, Pearson correlation coefficient and other quantitative indicators. The similarity value is the data credibility, and the value range is 0-1. The closer the value is to 1, the higher the degree of agreement between the target environmental load feature vector and the theoretical law environmental load feature vector.

[0117] The load-feature relationship model adopts the process of data preprocessing-feature engineering-model training-iterative optimization. The training data of the load-feature relationship model comes from the long-term monitoring database under the health state of the steel temporary bridge, including regular construction, extreme wind and rain, heavy lifting and other working conditions. The sample size of each type of working condition is greater than or equal to 5000 groups, so that the training data includes the full correlation scene of the steel temporary bridge construction period load and structural response. In the data preprocessing stage, the 3σ rule is used to eliminate outliers, and the Z-score standardization is used to eliminate the dimension influence. In the feature engineering link, the statistical features of the environmental load such as peak value, mean value and duration, and the core indexes of the structural time-frequency feature such as main frequency amplitude and modal damping ratio are extracted. The gradient boosting tree algorithm is selected for the main body of the load-feature relationship model, and the key parameters are set as follows: the learning rate is 0.05 to balance the training efficiency and accuracy, the maximum tree depth is 6 to avoid overfitting, the minimum sample weight is 0.1 to adapt to the volatility of construction data, and the model complexity is constrained based on L2 regularization term. In the training process of the load-feature relationship model, 5-fold cross-validation is adopted, the root mean square error between the predicted time-frequency feature and the measured value is taken as the objective function, and after 10 rounds of iterative optimization, the fitting degree of the load-feature relationship model is stable at more than 0.92, which meets the demand of engineering monitoring accuracy.

[0118] Exemplarily, the load-feature relationship model is integrated into the steel temporary bridge construction monitoring scene through the mechanism of working condition adaptation + parameter calibration + dynamic updating. For the time-varying characteristics of the steel temporary bridge during construction, the load-feature relationship model has a working condition recognition module built in, which locates the working condition to which the target environmental load feature vector belongs through feature matching of real-time environmental load data, and calls the corresponding model parameters of the working condition, for example, for the heavy lifting working condition, the model load-feature relationship model strengthens the correlation weight of load peak value and structure strain; for the working condition dominated by wind load, the mapping of vibration frequency characteristics is focused on. In addition, a monthly updating mechanism is established for the model, and new monitoring data in the current month under the health state are included in the training set, and the parameters are optimized through incremental learning, so that the model is always adapted to the structural stiffness change in the construction progress of the steel temporary bridge.

[0119] Secondly, the data credibility is graded and determined based on the second threshold value. Specifically, if the data credibility is greater than or equal to the second threshold value, it means that the multi-source monitoring data corresponding to the target environmental load feature vector is consistent with the theoretical feature rule under the load working condition, and then the data is determined as credible data. If the data credibility is less than the second threshold value, it means that the multi-source monitoring data corresponding to the target environmental load feature vector deviates from the normal load-feature correlation logic, which is invalid data caused by device-level problems such as sensor failure, transmission link interruption, and data acquisition module anomaly. Then the data is determined as untrustworthy data. And for the untrustworthy data, the health state checking process of the data acquisition device and the acquisition link is started.

[0120] From the above, it can be concluded that the load-feature relationship model is used to realize the theoretical prediction of the time-frequency feature vector, and the similarity comparison between the multi-scale time-frequency feature vector and the predicted time-frequency feature vector is quantified as the data credibility, which avoids the interference of false data on the determination of the structural health state. The grading determination rule based on the second threshold value divides the data credibility level, which improves the accuracy of the monitoring result. The health state checking of the device and the link and the generation of the list for the untrustworthy data can locate the fault hidden danger in the data acquisition link, and provide guidance for equipment maintenance and link optimization.

[0121] In an embodiment of the present application, the health state of the data acquisition device and the data acquisition link corresponding to the untrustworthy data is checked to obtain a health state list, including:

[0122] The health state of the data acquisition device corresponding to the untrustworthy data is checked, including:

[0123] It is checked whether the multi-source monitoring data collected by the data acquisition device corresponding to the untrustworthy data has a data segment with continuous zero value or constant value within a preset first time window, and the data segment is taken as an invalid data segment;

[0124] The signal noise level in a stationary environmental load period is calculated and compared with a historical normal noise benchmark level to obtain a comparison result;

[0125] If the noise level is greater than the historical normal noise benchmark level, it is determined that the signal quality of the corresponding data acquisition device is unreliable;

[0126] Based on the invalid data segment and the comparison result, a first health status list of the data acquisition device is obtained;

[0127] The health status of the data acquisition link of the data acquisition device corresponding to the untrusted data is checked, including:

[0128] The internal state parameters of the data acquisition device corresponding to the untrusted data are obtained and judged to obtain a first judgment result;

[0129] A standard test signal is injected into the specified data acquisition channel, and by analyzing the signal characteristics returned by the acquisition, the working state of the acquisition channel is judged to obtain a second judgment result;

[0130] Based on the first judgment result and the second judgment result, a second health status list of the data acquisition link is obtained;

[0131] The first health status list and the second health status list are summarized to obtain a health status list.

[0132] In this embodiment, the health status of the data acquisition device corresponding to the untrusted data is checked, including: first, the multi-source monitoring data collected by the data acquisition device corresponding to the untrusted data in a preset first time window is extracted, and whether there is a continuous zero value or constant value data segment is identified by continuity verification; Therefore, such data segments cannot reflect the true response of the structure, and are determined as invalid data segments, which are usually related to sensor power supply interruption, sensing module sticking and other faults. Secondly, the stationary environmental load period in the data is screened, for example, the period when there is no vehicle passing and the wind load is stable, and the noise level of the monitoring signal in this period is calculated and compared with the historical normal noise benchmark level to obtain a comparison result, which includes: if the current noise level is greater than the benchmark level, it indicates that the corresponding data acquisition device has problems such as signal drift and sensitivity attenuation, and the signal quality is determined to be unreliable. At the same time, based on the invalid data segment and the comparison result, a first health status list including the fault type of the data acquisition device, the abnormal period and the influence range is generated. The length of the first time window is set based on the sampling frequency of the data acquisition device, the load during the construction period of the steel bridge, and the typical duration of the equipment fault.

[0133] The health status of the data acquisition link of the data acquisition device corresponding to the untrusted data is checked, including: after completing the diagnosis of the data acquisition device, the internal state parameters of the device are called, such as the power supply voltage, the communication module signal strength, and the data buffer occupancy rate, and whether the device hardware exists running abnormity is judged through parameter threshold specification to obtain a first judgment result, the first judgment result includes: if the parameter is greater than the normal interval, it indicates that there is a hardware failure in the link source. Secondly, a standard test signal is injected into the specified data acquisition channel, such as a fixed frequency sine signal and an analog strain signal with a known amplitude, the amplitude, frequency and phase characteristics of the returned signal are collected and analyzed, specifically, if the returned signal deviates from the standard signal, it indicates that there is a transmission loss, signal distortion and other problems in the acquisition channel, and a second judgment result is obtained. Based on the above two judgment results, a second health status list including the link fault position, the transmission abnormality degree and the channel available state is generated. The parameter threshold specification is determined based on the factory technical index of the data acquisition device, the construction period on-site environment adaptation requirement and the statistical law of historical operation data.

[0134] Finally, the first health status list and the second health status list are summarized to obtain a health status list. The health status list includes not only the sensor failure, signal drift and other hardware problems of the data acquisition device, but also the transmission loss, channel failure and other link defects of the acquisition link, and the influence degree and priority of each abnormality on the monitoring data are labeled.

[0135] From the above, it can be concluded that by verifying the data segment effectiveness and comparing the signal noise level, the device jam, drift and other failures can be identified from the data level; and based on the internal parameter judgment and the standard signal injection test, the hardware operation and channel transmission state of the acquisition link can be comprehensively verified; in addition, by summarizing the first health status list and the second health status list, the reliability of the monitoring data is guaranteed, and the operation and maintenance efficiency and stability of the monitoring system are improved.

[0136] In an embodiment of the present application, a steel temporary bridge construction hidden danger monitoring method further comprises:

[0137] Based on the health status list, reliable monitoring data from the health status data acquisition device is screened out from the multi-source monitoring data;

[0138] Based on the reliable monitoring data and the number of health devices corresponding thereto, the to-be-optimized parameters of the hidden danger evaluation model are adjusted, the to-be-optimized parameters including a decision threshold and a feature weight parameter;

[0139] When the number of health devices in the health status list decreases, the decision threshold of the hidden danger evaluation model is increased by a first step;

[0140] re-calibrate feature weight parameters of the hidden danger assessment model according to reliable monitoring data, and compensate for output deviation of the hidden danger assessment model;

[0141] When the number of healthy devices in the health status list is less than a preset third threshold, a conservative early warning mode of the early warning decision model is triggered, and a trigger threshold of the hidden danger level is increased.

[0142] In the embodiment, after completing health status diagnosis of the data acquisition device and the collection link, the multi-source monitoring data is filtered based on the health status list. Specifically, monitoring data collected by the data acquisition device marked as healthy in the health status list is extracted, and monitoring data from the faulty device, the device with unreliable signal quality, and the link abnormal channel is eliminated. If the collection period of the historical data of the data acquisition device in the repair state does not overlap with the device fault period, the historical data is included in the reliable monitoring data after the signal quality is verified again, and the reliable monitoring data is obtained.

[0143] Secondly, based on the reliable monitoring data and the number of healthy devices corresponding thereto, the to-be-optimized parameters of the hidden danger assessment model are optimized. On the one hand, the decision threshold parameter is optimized, and the adjustment logic thereof is associated with the number of healthy devices. For example, when the number of healthy devices in the health status list decreases, it indicates that the dimension and sample size of the effective monitoring data have a risk of decline. In this case, the decision threshold is increased by a first step to reduce the false alarm probability caused by insufficient data. If the number of healthy devices rises to the normal range, the decision threshold is gradually adjusted to the reference level. On the other hand, the feature weight parameters of the hidden danger assessment model are re-calibrated according to the reliable monitoring data, and the output deviation of the hidden danger assessment model is compensated. Specifically, based on the health status list, unreliable monitoring data from the data acquisition device in the unhealthy state is filtered from the multi-source monitoring data, and based on the unreliable monitoring data, a feature weight parameter associated with the unreliable monitoring data in the hidden danger assessment model is determined as an unreliable feature weight parameter. Then, the unreliable feature weight parameter is reduced by a preset second step, and the feature weight parameter corresponding to the reliable monitoring data is increased by the second step, and the output deviation of the hidden danger assessment model is calculated. Based on the output deviation, a compensation term is added to the output value of the hidden danger assessment model. The value of the first step is determined based on the attenuation amplitude of the number of healthy devices, the missing proportion of the monitoring data dimension, and the statistical law of historical false alarm cases. The second step is set based on the physical association between features, the statistical correlation, and the contribution weight of the feature to the hidden danger assessment.

[0144] When the number of healthy devices in the health status list is less than the preset third threshold value, it indicates that the monitoring system has failed to achieve comprehensive coverage of the key parts of the steel bridge, and the integrity and representativeness of the data are defective. At this time, the conservative warning mode is triggered, and the triggering of the conservative warning mode is to increase the triggering threshold of the hidden danger level, for example, the original light hidden danger, that is, the pre-warning logic is adjusted to the severe hidden danger. The third threshold value is set around the coverage requirements of the key monitoring parts of the steel bridge, the minimum effective sample size of the monitoring data, and the risk tolerance of structure safety control.

[0145] From the above, the embodiment adjusts the decision threshold and feature weight parameter of the hidden danger evaluation model based on reliable monitoring data and the number of health devices corresponding thereto. When the number of health devices decreases, the decision threshold is increased to reduce the risk of false warning caused by insufficient data. The accuracy of the evaluation result of the hidden danger evaluation model is ensured by weight recalibration and deviation compensation. When the number of health devices is less than the third threshold value, the conservative warning mode is triggered, and the triggering threshold of the hidden danger level is increased, which effectively avoids excessive warning caused by lack of data.

[0146] In an embodiment of the present application, the feature weight parameter of the hidden danger evaluation model is recalibrated according to the trusted data, and the deviation of the hidden danger evaluation model is compensated, including:

[0147] Based on the health status list, unreliable monitoring data from unhealthy state data acquisition devices is screened out from the multi-source monitoring data;

[0148] Based on the unreliable monitoring data, the feature weight parameter associated with the unreliable monitoring data in the hidden danger evaluation model is determined as the unreliable feature weight parameter

[0149] The unreliable feature weight parameter is reduced by a preset second step, and the feature weight parameter corresponding to the reliable monitoring data is increased based on the second step, wherein the second step is determined based on the physical association and statistical correlation between features;

[0150] According to the statistical distribution change of the reliable monitoring data, the output deviation of the hidden danger evaluation model is calculated;

[0151] Based on the output deviation, a compensation term is added to the output value of the hidden danger evaluation model;

[0152] Wherein, the recalibration of the feature weight parameter and the compensation of the deviation are carried out synchronously.

[0153] In the present embodiment, during the parameter optimization process of the hidden danger assessment model, the calibration of feature weights is completed based on the health status list. First, unreliable monitoring data collected by unhealthy state equipment is screened from the multi-source monitoring data, and through feature correlation mapping analysis, the feature weight parameters corresponding to these unreliable data in the hidden danger assessment model are located and marked as unreliable feature weight parameters. Subsequently, the weight parameters are adjusted in both directions using a preset second step: on the one hand, the proportion of unreliable feature weight parameters is reduced based on the second step to weaken the interference of this data on the assessment results of the hidden danger assessment model; on the other hand, the weight of the feature weight parameters corresponding to the reliable monitoring data collected by healthy equipment is increased based on the second step to strengthen the core role of this data in model decision-making. The second step is set based on the physical correlation and statistical correlation between features. The physical correlation includes the mechanical correlation between load features and structural strain features; the statistical correlation includes the Pearson correlation coefficient of feature data.

[0154] At the same time of recalibrating the feature weights, the output deviation of the hidden danger assessment model is compensated. Specifically, based on the statistical distribution changes of reliable monitoring data, such as mean value shift and variance fluctuation, the output deviation of the hidden danger assessment model is obtained by comparing the output value of the hidden danger assessment model with the structure state check value and performing calculation. Subsequently, according to the direction and amplitude of the output deviation, a reverse compensation term is added to the output value of the hidden danger assessment model, for example, if the output deviation is positive, a negative compensation term is used to offset the deviation; if the deviation is negative, a positive compensation term is added to correct the result.

[0155] From the above, it can be seen that by directionally screening unreliable monitoring data and reducing the feature weights corresponding thereto, while increasing the feature weights of reliable data, the effectiveness and rationality of the input of the hidden danger assessment model are ensured, the interference of invalid data on the assessment results is weakened, and the decision weight of valid data is strengthened; and based on the calculation of the output deviation amount from the statistical distribution changes of reliable monitoring data and the synchronous addition of a reverse compensation term, the weight adjustment can be corrected in time, thereby reducing the model assessment error caused by data fluctuations; the synchronous execution of weight calibration and deviation compensation realizes the coordination and unity of parameter optimization and result correction of the hidden danger assessment model, and improves the robustness and decision-making scientificity of the hidden danger assessment model under dynamic data state conditions.

[0156] In an embodiment of the present application, based on multi-source monitoring data, a structure performance index system is established, and the preliminary hidden danger grade is corrected to obtain the final hidden danger grade, including:

[0157] Based on the design specifications and safety control standards of the steel temporary bridge construction stage, the benchmark control limit value of the structure performance is determined;

[0158] A finite element calculation model of the steel temporary bridge during construction is established, and a theoretical response value thereof is calculated;

[0159] The reference control limit is fused with the theoretical response value of the finite element calculation model to obtain a structure performance index system.

[0160] According to the statistical characteristics of the multi-source monitoring data and the structure performance index system, a preliminary hidden danger grade is quantitatively corrected to obtain a final hidden danger grade.

[0161] The construction of the structure performance index system and the correction process of the preliminary hidden danger grade are optimized for the time-varying structure characteristics and load conditions during the construction period of the steel temporary bridge.

[0162] In this embodiment, the reference control limit complies with the design specification and safety control standard of the steel temporary bridge during the construction stage, and based on the design load level of the steel temporary bridge, the material performance of the component, the construction process requirements and other core elements, the safety threshold of the key performance indicators such as deflection, strain, natural frequency and support reaction force is respectively determined. For example, for the main beam component, based on the deflection limit requirement of the temporary steel structure in the “Highway Bridge and Culvert Construction Technical Specification”, and according to the dynamic impact coefficient of the vehicle load during the construction period, the maximum allowable deflection is determined as 1 / 500 of the span; for the connecting node, according to the shear bearing capacity design value of the high-strength bolt, the warning threshold of the node strain is determined. At the same time, the reference control limit also takes into account the temporary working condition characteristics during the construction stage, and adds temporary control thresholds for performance indicators under special procedures such as foundation pit excavation and component hoisting.

[0163] In this embodiment, a finite element calculation model of the steel temporary bridge during construction is first established. Specifically, in the modeling process, the actual structure form of the temporary bridge is fully restored, including the geometric size, material parameters and connection mode of the main beam, secondary beam and support, and the boundary conditions of the finite element calculation model are updated according to the construction progress, for example, the stiffness change of the foundation soil needs to be simulated during the pile foundation construction stage, and the distribution characteristics of the additional load need to be considered during the bridge deck paving stage.

[0164] After the finite element calculation model is established, based on the typical load working conditions during the construction period, through static analysis, modal analysis and dynamic time history analysis, the theoretical response values of the key parts of the structure under each working condition are obtained, including component stress, structure deflection, natural frequency and the like.

[0165] Secondly, the benchmark control limit value and the theoretical response value are fused to obtain the structural performance index system of the steel temporary bridge during the construction period. In the fusion process, the weighted fusion algorithm is adopted to take the benchmark control limit value as the safety bottom line index, and the theoretical response value of the finite element model is taken as the performance matching index. In addition, the time-varying structural characteristics and load conditions of the steel temporary bridge during the construction period are optimized for the index system, for example, a stiffness mutation monitoring index is added during the structural system conversion stage, and the stress concentration monitoring index of the local component is strengthened under the heavy equipment hoisting working condition. Finally, the structural performance index system is obtained, which includes not only the absolute safety threshold of each performance parameter, but also the relative deviation range of the theoretical response and the monitoring response, providing a scientific and comprehensive basis for the quantitative correction of the hazard level.

[0166] Based on the structural performance index system, the key performance index associated with the preliminary hazard level and the corresponding measured value are determined; the measured value of the key performance index is matched with the preset hazard level threshold interval of each level in the structural performance index system to determine the independent hazard level based on the index system, and the level difference between the preliminary hazard level and the independent hazard level is calculated, and the confidence level of the preliminary hazard level is calculated according to the statistical characteristics of the multi-source monitoring data in the preset time window; based on the level difference and the confidence level of the preliminary hazard level, a correction function is constructed. Finally, the preliminary hazard level is quantitatively adjusted based on the correction function to obtain the final hazard level.

[0167] From the above, it can be seen that the embodiment determines the benchmark control limit value based on the design specification and safety standard of the construction stage, and forms the index system by fusing the theoretical response value of the finite element model during the construction period, which not only anchors the structural safety bottom line, but also takes into account the structural dynamics characteristics under the construction working condition; the index system and the correction process are optimized for the time-varying structural characteristics and load conditions during the construction period, which can adapt to the construction scenes such as component erection and dynamic load change; the preliminary hazard level is quantitatively corrected based on the statistical characteristics of the multi-source monitoring data and the index system, which effectively eliminates the level misjudgment caused by the construction working condition fluctuation, improves the accuracy and pertinence of the final hazard level determination, and provides a reliable quantitative basis for the safety control of the steel temporary bridge structure during the construction period.

[0168] In an embodiment of the present application, a finite element calculation model of the steel temporary bridge during the construction period is established, and the theoretical response value thereof is calculated, including:

[0169] According to the construction drawings and material properties of the steel temporary bridge, an initial finite element model is established, and a load value equivalent to the environmental load data is applied to the initial finite element model;

[0170] The solver calculates the structural response of the initial finite element model under the action of the load to obtain a calculation result, and compares the calculation result with the target structural response data to obtain a comparison result.

[0171] Based on the comparison result, a model updating algorithm is used to inversely correct the material parameters and boundary conditions of the initial finite element model to obtain a finite element calculation model in the construction period, and the finite element calculation model in the construction period is calculated to obtain a theoretical response value.

[0172] In the initial stage of building the finite element calculation model in the construction period of the steel temporary bridge, the structural topology and component geometric parameters of the steel temporary bridge are restored based on the construction drawings as the core basis. At the same time, the real material properties of the steel temporary bridge are imported, such as the elastic modulus, Poisson's ratio, density and other mechanical parameters of the steel material, so that the finite element calculation model is consistent with the actual mechanical properties. After the definition of the model geometry and material parameters is completed, load equivalent conversion is performed based on the environmental load data obtained by monitoring. Specifically, the wind load data is converted into uniform or local wind pressure load, the construction vehicle load data is converted into moving concentrated load or equivalent uniform load, and the temperature load data is converted into node temperature strain load. Then, the equivalent load values are applied to the corresponding parts of the initial finite element model based on the working condition.

[0173] In this embodiment, the initial model under the action of the load is analyzed by the finite element solver to obtain the structural response of the key parts of the structure, and the calculation result is obtained. The structural response includes the core response indexes such as the stress, strain, deflection, natural frequency and vibration mode of the component. Then, the calculation result is compared with the target structural response data in multiple dimensions. For example, in the numerical level, the response amplitude difference of the same monitoring point is compared; in the trend level, the consistency of the response change with the load is compared; and in the distribution level, the distribution characteristic matching degree of the response in the structure space is compared. The deviation rate and root mean square error of the two are calculated to obtain the comparison result.

[0174] After obtaining the comparison results, a model updating algorithm is used to carry out reverse correction of material parameters and boundary conditions. Specifically, in terms of material parameter correction, based on the comparison results, parameters such as the elastic modulus of steel and the stiffness coefficient of nodes are adjusted; in terms of boundary condition correction, according to the measured displacement and counterforce data at the support, the constraint type of the support is optimized, for example, the ideal hinge support is adjusted to an elastic constraint support, the stiffness coefficient of the foundation, etc., so that the boundary of the finite element calculation model matches the engineering conditions. In the correction process, through multiple rounds of iterative calculation and deviation verification, until the deviation between the calculation results of the finite element calculation model and the target structure response data meets the preset accuracy requirement, and finally the finite element calculation model of the steel bridge that fits the actual state during construction is obtained. Based on the finite element calculation model, mechanical solving is carried out again to obtain the theoretical response value. Among them, the preset accuracy is based on the deviation between the calculation results and the target structure response data as the core judgment basis, and needs to meet the multi-dimensional error threshold requirements, including multi-dimensional error threshold including numerical deviation threshold, fitting threshold, working condition adaptation threshold and engineering experience threshold, etc.

[0175] Among them, the input and output data of the model updating algorithm are related around the deviation-correction-precision model. The input data includes three types of key information: first, the quantitative comparison results of the previous sequence; second, the core parameters of the initial finite element model, including the steel elastic modulus, the node stiffness coefficient, the support constraint type, etc. to be corrected; third, the preset accuracy requirement, which is the termination standard of model correction, including the numerical deviation threshold, the working condition adaptation threshold and other multi-dimensional indicators as the correction quantitative scale. The algorithm adjusts the parameters to be corrected based on the comparison results through the reverse correction logic, and after multiple rounds of iterative calculation, outputs the corrected finite element calculation model during construction and the corresponding theoretical response value. At this time, the deviation between the theoretical response value and the target structure response data meets the preset accuracy, realizing the causal correlation of deviation input-parameter adjustment-precision output.

[0176] The parameters of the model updating algorithm include control parameters, material correction parameters and boundary condition parameters. The control parameters include the iteration step factor, the convergence judgment threshold and the parameter constraint boundary; the material correction parameters determine the initial value of the elastic modulus based on the steel grade, for example, the initial value of the elastic modulus of Q345 steel is 2.06x10 11 Pa. Among them, the iteration step factor is 0.1-0.5, and the default is 0.3; the convergence judgment threshold needs to meet the multi-dimensional indicators such as the stress / strain deviation rate less than or equal to 5%, the deflection deviation less than or equal to 3mm, and the correlation coefficient greater than or equal to 0.9, and the upper limit of the iteration number is 10 times; the parameter constraint boundary limits the steel elastic modulus in the interval of 2.0x10 11 -2.1x10 11 Pa,

[0177] From the above, the embodiment establishes a finite element calculation model based on construction drawings and material properties, and applies equivalent environmental load to it, which guarantees the basic reduction degree of the finite element calculation model to the actual structure and working condition; by comparing the calculation response of the finite element calculation model with the field monitoring data, the deviation of the model and the mechanical properties of the actual structure can be identified at the initial time; at the same time, the material parameters and boundary conditions are corrected in reverse based on the model updating algorithm, so that the theoretical response value output by the final construction period finite element model has scientificity and accuracy.

[0178] In an embodiment of the present application, the preliminary hidden danger grade is quantitatively corrected according to the statistical characteristics of the multi-source monitoring data and the structure performance index system, to obtain the final hidden danger grade, including:

[0179] Based on the structure performance index system, the key performance indicators associated with the preliminary hidden danger grade and their corresponding measured values are determined;

[0180] The measured values of the key performance indicators are matched with the preset threshold intervals of each hidden danger grade in the structure performance index system to determine the independent hidden danger grade based on the index system;

[0181] The grade difference between the preliminary hidden danger grade and the independent hidden danger grade is calculated;

[0182] According to the statistical characteristics of the multi-source monitoring data in the preset second time window, the confidence level of the preliminary hidden danger grade is calculated;

[0183] Based on the grade difference and the confidence level, a correction function is constructed;

[0184] The preliminary hidden danger grade is quantitatively adjusted based on the correction function to output the final hidden danger grade.

[0185] In the embodiment, first, based on the structure performance index system that has been constructed, the key performance indicators that are strongly associated with the preliminary hidden danger grade are selected. These indicators are closely related to the core dimensions of the structure safety of the steel bridge during the construction period, such as deflection, strain, natural frequency, node displacement, etc., and are directly related to the structure abnormal type corresponding to the preliminary hidden danger grade. The structure abnormal type includes stiffness attenuation or local stress concentration. Second, the measured values of the key performance indicators in the current monitoring period are extracted, including the mean value, peak value, fluctuation amplitude and change trend characteristic value of the real-time monitoring data, etc., and the measured data are processed for outlier rejection and smoothing filtering.

[0186] Secondly, based on the preset threshold interval of each level of hidden danger in the structure performance index system, the measured values of the key performance indicators are matched one by one. The threshold interval is divided according to the time-varying working condition characteristics during the construction period, for example, for the deflection index of the main beam, the threshold value of the normal state is set to within 1 / 600 of the span, the mild hidden danger is 1 / 600-1 / 500, the moderate hidden danger is 1 / 500-1 / 400, and the severe hidden danger is greater than 1 / 400. In the matching process, if the measured value of a single indicator falls into a certain hidden danger level interval, it is temporarily recorded as the independent level corresponding to the indicator; if multiple indicators correspond to different levels, the comprehensive matching result is calculated based on the safety weight of each indicator, and the independent hidden danger level based on the index system is finally determined.

[0187] Subsequently, by constructing a level quantization mapping table, for example, assigning values of 0, 1, 2, and 3 to normal, mild, moderate, and severe hidden dangers respectively, and calculating the numerical difference between the preliminary hidden danger level and the independent hidden danger level, according to the proportion of the difference value to the total span of the level, the standardized level difference degree is obtained, wherein the value of the level difference degree ranges from 0 to 1, and the larger the value, the higher the degree of deviation between the two levels. At the same time, based on the statistical characteristics of the multi-source monitoring data within the preset second time window, the calculation includes the integrity of the data (the proportion of valid data), the stability (the size of the variance), the consistency (the degree of agreement of multi-device monitoring data), etc. For example, if the proportion of valid data is greater than or less than 90% and the variance is less than or equal to 5%, the confidence level is assigned a value of 0.9; if the data integrity is less than 60%, the confidence level is reduced to below 0.3, indicating the reliability of the preliminary hidden danger level determination result, and the confidence level of the preliminary hidden danger level is obtained. The preset second time window is set based on the working condition characteristics of the steel temporary bridge during the construction period, the effectiveness requirements of multi-dimensional monitoring data, and the time scale law of hidden danger development.

[0188] Finally, based on the level difference degree and the confidence level, a nonlinear correction function is constructed, and the form of the correction function takes into account the coupling effect of the two, and the expression of the correction function is: D1=D2+α×ΔD×C L where D1 is the final hidden danger level, D2 is the preliminary hidden danger level, a is the correction coefficient, AD is the level difference degree, and CL is the confidence level. The correction coefficient a is set based on the structure time-varying characteristics during the construction stage of the steel temporary bridge, the statistical reliability of the monitoring data, and the safety redundancy requirements of hidden danger level determination, and its value range is adapted to the coupling strength of the level difference degree and the confidence level.

[0189] It can be concluded from the above that the embodiment effectively avoids the limitations of single decision logic by anchoring the key performance indicators in the structure performance index system and carrying out independent hidden danger grade matching. The confidence level based on the grade difference and the confidence level based on the statistical characteristics of the monitoring data are used to construct the correction function, which realizes the dynamic quantitative adjustment of the preliminary hidden danger grade, taking into account the influence of data reliability on the determination result, and reducing the grade misjudgment through the difference degree calibration.

[0190] The steel bridge construction hidden danger monitoring method corresponding to the above embodiment, Figure 2 The structural block diagram of the steel bridge construction hidden danger monitoring system provided by an embodiment of the present application is shown. For the sake of convenience, only the parts related to the embodiments of the present application are shown. For reference Figure 2 The steel bridge construction hidden danger monitoring system 20 includes a multi-source monitoring data acquisition module 21, an environmental load data processing module 22, a structure response data preprocessing module 23, a time-frequency feature vector extraction module 24, a preliminary hidden danger grade evaluation module 25, and a early warning decision instruction generation module 26.

[0191] The multi-source monitoring data acquisition module 21 is used to acquire multi-source monitoring data in the construction process through data acquisition equipment deployed on the steel bridge, and the multi-source monitoring data includes structure response data and environmental load data;

[0192] The environmental load data processing module 22 is used to preprocess and feature process the environmental load data to obtain a target environmental load feature vector;

[0193] The structure response data preprocessing module 23 is used to preprocess the structure response data to obtain target structure response data;

[0194] The time-frequency feature vector extraction module 24 is used to use the wavelet transform algorithm to perform time-frequency domain decomposition on the target structure response data to obtain a group of sub-band signals of different frequencies; and extract a multi-scale time-frequency feature vector related to wind-induced vibration and wave-induced vibration from the group of sub-band signals of different frequencies;

[0195] The preliminary hidden danger grade evaluation module 25 is used to input the multi-scale time-frequency feature vector and the target environmental load feature vector into a pre-trained hidden danger evaluation model to obtain a preliminary hidden danger grade; a structure performance index system is established based on the multi-source monitoring data, and the preliminary hidden danger grade is corrected to obtain a final hidden danger grade;

[0196] The early warning decision instruction generation module 26 is used to input the final hidden danger grade and the acquired sensor health state data into a fuzzy logic-based early warning decision model to generate corresponding early warning instructions.

[0197] Referring to Figure 3 , Figure 3A schematic block diagram of an electronic device is provided for an embodiment of the present application. As shown in the figure Figure 3 The electronic device 300 in the embodiment can include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The above-mentioned processor 301, input device 302, output device 303, and memory 304 complete communication with each other through a communication bus 305. The memory 304 is used to store a computer program, and the computer program includes program instructions. The processor 301 is used to execute the program instructions stored in the memory 304. Among them, the processor 301 is configured to call the program instructions to execute the functions of each module in the above-mentioned device embodiments, for example Figure 2 The functions of the multi-source monitoring data acquisition module 21, the environmental load data processing module 22, the structural response data preprocessing module 23, the time-frequency feature vector extraction module 24, the preliminary hidden danger grade evaluation module 25, and the early warning decision instruction generation module 26 shown in the figure.

[0198] It should be understood that in the embodiments of the present application, the processor 301 can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), ready-to-program gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0199] The input device 302 can include a touchpad, a fingerprint collection sensor (used to collect fingerprint information and direction information of the fingerprint of a user), a microphone, etc., and the output device 303 can include a display (LCD, etc.), a loudspeaker, etc.

[0200] The memory 304 can include read-only memory and random access memory, and provide instructions and data for the processor 301. A part of the memory 304 can also include non-volatile random access memory. For example, the memory 304 can also store device type information.

[0201] In specific implementations, the processor 301, the input device 302, and the output device 303 described in the embodiments of the present application can perform the implementation described in any embodiment of the steel bridge construction hidden danger monitoring method provided by the embodiments of the present application, and can also perform the implementation of the electronic device described in the embodiments of the present application, which will not be described here.

[0202] In another embodiment of the present application, a computer-readable storage medium is provided, which stores a computer program. The computer program includes program instructions, which are executed by a processor to implement all or part of the processes of the above-mentioned embodiments. The computer program can also be used to instruct related hardware to complete the implementation. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned method embodiments can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0203] The computer-readable storage medium can be an internal storage unit of the electronic device of any of the above-mentioned embodiments, such as a hard disk or a memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the computer-readable storage medium can include both the internal storage unit and the external storage device of the electronic device. The computer-readable storage medium is used to store the computer program and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0204] Those skilled in the art can clearly understand that, for the convenience and conciseness of description, the specific working process of the electronic device and the units described above can refer to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0205] Those skilled in the art can clearly understand that, for the convenience and conciseness of description, the specific working process of the electronic device and the units described above can refer to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0206] In several embodiments provided in the present application, it should be understood that the disclosed electronic device and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface or unit, and can also be electrical, mechanical or other forms of connection.

[0207] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0208] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0209] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for monitoring potential construction hazards of steel temporary bridges, characterized in that, include: Data acquisition equipment deployed on the steel temporary bridge collects multi-source monitoring data during the construction process, including structural response data and environmental load data. The environmental load data is preprocessed and feature-processed to obtain the target environmental load feature vector; The structural response data is preprocessed to obtain the target structural response data; The wavelet transform algorithm is used to decompose the response data of the target structure in the time and frequency domain to obtain a set of sub-band signals of different frequencies; From the set of sub-band signals of different frequencies, extract multi-scale time-frequency feature vectors related to wind-induced vibration and wave-induced vibration; The multi-scale time-frequency feature vector and the target environmental load feature vector are input into a pre-trained hazard assessment model to obtain a preliminary hazard level; Based on the design specifications and safety control standards for the construction phase of the steel temporary bridge, the benchmark control limits for structural performance are determined. At the same time, a finite element calculation model for the construction phase of the steel temporary bridge is established, and its theoretical response value is calculated. By integrating the benchmark control limits with the theoretical response values ​​of the finite element calculation model, a structural performance index system is obtained. Based on the statistical characteristics and structural performance index system of multi-source monitoring data, the preliminary hazard level is quantitatively corrected to obtain the final hazard level. The final hazard level and the acquired sensor health status data are input into a fuzzy logic-based early warning decision model to generate corresponding early warning instructions. The pre-trained hazard assessment model was trained and validated based on historical data of the structural health monitoring of the temporary steel bridge. The hazard assessment model architecture adopts a hybrid model of gradient boosting decision tree and deep neural network: the gradient boosting decision tree module is responsible for capturing the nonlinear correlation and hierarchical logic between features, while the deep neural network module realizes deep fitting of complex load-vibration coupling relationship. The early warning decision model based on fuzzy logic is constructed based on the fuzzy reasoning rule base of the steel temporary bridge project. The fuzzy reasoning rule base is formulated based on the structural safety control requirements of the steel temporary bridge and the reliability constraints of sensor data.

2. The method for monitoring potential construction hazards of a steel temporary bridge according to claim 1, characterized in that, The wavelet transform algorithm is used to perform time-frequency domain decomposition on the target structure response data to obtain a set of sub-band signals of different frequencies, including: Based on the spectral characteristics of the target structure response data and monitoring requirements, a wavelet basis function is selected to obtain the target wavelet basis function; Based on the Nyquist sampling theorem and the natural frequency of the steel temporary bridge, the number of wavelet decomposition layers is determined. Based on the target wavelet basis function and the wavelet decomposition level, the target structural response data is decomposed into multi-scale decomposition to obtain approximation coefficients and detail coefficients. The set of sub-band signals with different frequencies includes approximation coefficients and detail coefficients.

3. The method for monitoring potential construction hazards of a steel temporary bridge according to claim 2, characterized in that, The determination of the wavelet decomposition level based on the Nyquist sampling theorem and the natural frequency of the steel temporary bridge includes: The highest frequency of the power spectral density was determined by analyzing the sampling frequency of the multi-source monitoring data using the Nyquist sampling theorem. Obtain the first natural frequency of the key natural frequency of the steel temporary bridge, and set the first natural frequency as the target frequency; The theoretical minimum number of decomposition layers is obtained by calculating the highest frequency and the target frequency based on the formula N_min=ceil(log2(f_max / f_target)); where f_max is the highest frequency, f_target is the target frequency, and ceil is the floor function. The number of wavelet decomposition layers is obtained by adding a preset engineering margin to the theoretical minimum decomposition layer.

4. The method for monitoring construction hazards of a steel temporary bridge according to claim 3, characterized in that, Also includes: Adjustment of the highest frequency includes: Calculate the power spectral density of the target structure response data within a preset time window; Identify all energy peaks in the power spectral density, obtain the energy peaks, and record their corresponding frequencies; The energy peak value is compared with a preset first threshold. If the energy peak value is greater than the first threshold, the highest frequency will be updated based on the energy peak value; otherwise, the highest frequency will remain unchanged. Adjusting the target frequency includes: The target structure response data within the time window is obtained, and the target structure response data is re-identified using the random subspace identification method or the frequency domain decomposition method to obtain the actual first-order natural frequency of the steel temporary bridge. Calculate the absolute value of the deviation between the actual first-order natural frequency and the currently used target frequency; If the absolute value of the deviation is greater than a preset second threshold, the target frequency is updated based on the actual first-order natural frequency; otherwise, the target frequency remains unchanged.

5. The method for monitoring potential construction hazards of a steel temporary bridge according to claim 1, characterized in that, After extracting the multi-scale time-frequency feature vectors related to wind-induced vibration and wave-induced vibration from the set of sub-band signals of different frequencies, the process includes: The multi-scale time-frequency feature vectors are compared with a historical benchmark database; If the multi-scale time-frequency feature vector does not deviate from the normal range in the historical benchmark library, then the multi-scale time-frequency feature vector is determined to be normal data; If the multi-scale time-frequency feature vector deviates from the normal range in the historical benchmark database, then the multi-scale time-frequency feature vector is determined to be preliminary abnormal data.

6. The method for monitoring construction hazards of a steel temporary bridge according to claim 5, characterized in that, After determining that the multi-scale time-frequency feature vector is preliminary anomalous data if it deviates from the normal range in the historical benchmark database, the method further includes: The target environmental load feature vector is matched with the historical environmental load intervals corresponding to when similar abnormal feature vectors were generated in the historical benchmark library; If the target environmental load is within the historical environmental load range, the preliminary abnormal data is determined to be real abnormal data; otherwise, it is determined to be data to be observed. Calculate the Mahalanobis distance between the real abnormal data and the center point of the corresponding normal feature vector in the historical benchmark database, as the deviation. When the deviation exceeds a preset first threshold, the corresponding real abnormal data is marked as data of key concern.

7. The method for monitoring construction hazards of a steel temporary bridge according to claim 6, characterized in that, Also includes: The target environmental load feature vector is input into a predefined load-feature relationship model to obtain the predicted time-frequency feature vector; The similarity between the multi-scale time-frequency feature vector and the predicted time-frequency feature vector is calculated as the data reliability. When the data credibility is greater than or equal to a preset second threshold, the corresponding data is determined to be credible data. When the credibility of the data is less than the second threshold, the corresponding data is determined to be unreliable data. The health status of the data acquisition devices and data acquisition links corresponding to the untrusted data is checked to obtain a health status list.

8. The method for monitoring construction hazards of a steel temporary bridge according to claim 7, characterized in that, The health status of the data acquisition device and data acquisition link corresponding to the untrusted data is checked to obtain a health status list, including: The health status of the data acquisition device corresponding to the untrusted data is checked, including: Check the data segments of the multi-source monitoring data collected by the data acquisition device corresponding to the untrusted data that have continuous zero values ​​or constant values ​​within a preset first time window, and regard the data segments as invalid data segments; Calculate the signal noise level during a period of stable environmental load and compare it with the baseline level of historical normal noise to obtain the comparison results; If the noise level is greater than the baseline level of historical normal noise, then the signal quality of the corresponding data acquisition device is determined to be unreliable. Based on the invalid data segments and the comparison results, a first health status list of the data acquisition device is obtained; The health status of the data acquisition link of the data acquisition device corresponding to the untrusted data is checked, including: Obtain the internal state parameters of the data acquisition device corresponding to the untrusted data, make a judgment, and obtain a first judgment result; A standard test signal is injected into the specified data acquisition channel. By analyzing the characteristics of the acquired signal, the working status of the acquisition channel is determined, and a second judgment result is obtained. Based on the first judgment result and the second judgment result, a second health status list of the data acquisition link is obtained; The first health status list and the second health status list are combined to obtain the health status list.

9. A method for monitoring construction hazards of a steel temporary bridge according to claim 8, characterized in that, Also includes: Based on the health status list, reliable monitoring data from the health status data acquisition device are selected from the multi-source monitoring data; Based on the reliable monitoring data and the corresponding number of health devices, the parameters to be optimized in the hazard assessment model are adjusted. The parameters to be optimized include decision threshold and feature weight parameters. When the number of healthy devices in the health status list decreases, the decision threshold of the hazard assessment model is increased in the first step. The feature weight parameters of the hazard assessment model are recalibrated based on the reliable monitoring data, and the output deviation of the hazard assessment model is compensated. When the number of healthy devices in the health status list is less than a preset third threshold, the conservative early warning mode of the early warning decision model is triggered, raising the trigger threshold for the hidden danger level.

10. A monitoring system for potential construction hazards of steel temporary bridges, characterized in that, include: The multi-source monitoring data acquisition module is used to collect multi-source monitoring data during the construction process through data acquisition equipment deployed on the steel temporary bridge. The multi-source monitoring data includes structural response data and environmental load data. An environmental load data processing module is used to preprocess and feature process the environmental load data to obtain a target environmental load feature vector. The structural response data preprocessing module is used to preprocess the structural response data to obtain the target structural response data; The time-frequency feature vector extraction module is used to perform time-frequency domain decomposition on the target structure response data using a wavelet transform algorithm to obtain a set of sub-band signals of different frequencies; and to extract multi-scale time-frequency feature vectors related to wind-induced vibration and wave-induced vibration from the set of sub-band signals of different frequencies. The preliminary hazard level assessment module is used to input the multi-scale time-frequency feature vector and the target environmental load feature vector into a pre-trained hazard assessment model to obtain the preliminary hazard level; Based on the design specifications and safety control standards for the construction phase of the steel temporary bridge, the benchmark control limits for structural performance are determined. At the same time, a finite element calculation model for the construction phase of the steel temporary bridge is established, and its theoretical response value is calculated. By integrating the benchmark control limits with the theoretical response values ​​of the finite element calculation model, a structural performance index system is obtained. Based on the statistical characteristics and structural performance index system of multi-source monitoring data, the preliminary hazard level is quantitatively corrected to obtain the final hazard level. The early warning decision instruction generation module is used to input the final hazard level and the acquired sensor health status data into the early warning decision model based on fuzzy logic to generate corresponding early warning instructions. The pre-trained hazard assessment model was trained and validated based on historical data of the structural health monitoring of the temporary steel bridge. The hazard assessment model architecture adopts a hybrid model of gradient boosting decision tree and deep neural network: the gradient boosting decision tree module is responsible for capturing the nonlinear correlation and hierarchical logic between features, while the deep neural network module realizes deep fitting of complex load-vibration coupling relationship. The early warning decision model based on fuzzy logic is constructed based on the fuzzy reasoning rule base of the steel temporary bridge project. The fuzzy reasoning rule base is formulated based on the structural safety control requirements of the steel temporary bridge and the reliability constraints of sensor data.

Images