A cloud platform-based settlement data analysis and early warning method

Abstract

The application relates to a cloud platform-based settlement data analysis and early warning method, which comprises the following steps: collecting settlement data of a main transformer foundation part in real time through a high-precision electronic settlement observation device array, obtaining deformation stress data through data preprocessing and coordinate calculation, and uploading the deformation stress data to a cloud platform; emitting ultrasonic pulse signals to stress concentration positions in a local area through an ultrasonic detection device, receiving reflected wave data, and identifying the depth and range of a crack initiation position according to the reflected wave data; arranging a sensor array on the surface of the crack initiation position, capturing sound wave signals released in the crack expansion process, and determining a time sequence of crack expansion direction and speed; repairing the crack depth data of a priority area through ultrasonic detection, calculating an adjusted foundation health state value, comparing the adjusted foundation health state value with a preset safety threshold, and generating a corresponding safety warning signal.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital data, in particular to a settlement data analysis and early warning method based on a cloud platform. BACKGROUND

[0002] The settlement data analysis and early warning technology plays a crucial role in the field of infrastructure safety, especially for structures that bear heavy equipment, such as the main transformer foundation of a substation, whose stability and safety are directly related to the reliable operation of the power system. Uneven settlement can cause abnormal internal stress of the foundation, induce crack initiation and propagation, and thus threaten the overall bearing capacity of the structure. Therefore, researching efficient settlement monitoring and crack identification methods is not only an urgent need for engineering safety, but also a key direction for the development of intelligent monitoring technology. Currently, traditional settlement monitoring relies on single measurement methods, such as leveling or conventional sensors, which are difficult to fully capture the internal strain changes and crack dynamics of the foundation. Even if some structural health monitoring technologies are combined, data processing is still mainly offline analysis, lacking real-time and systematicity, making it difficult to accurately predict the initiation location and propagation trend of cracks. The limitations of this method lie in the inability to effectively integrate multi-source data and the difficulty in dealing with complex stress distribution changes under long-term heavy load. The main challenge in the research is how to accurately identify the initiation and propagation process of internal cracks in the foundation. Specifically, although high-precision settlement observation device arrays can provide rich displacement data, it is still a technical difficulty to convert them into reliable indicators reflecting internal strain. In addition, techniques such as ultrasonic detection and acoustic emission monitoring are limited by signal processing capabilities and environmental interference, making it difficult to accurately determine the propagation direction and speed of cracks. These unresolved technical factors lead to a high degree of uncertainty in assessing the impact of cracks on the bearing capacity of the foundation, thus forming a difficult problem that needs to be broken through. Therefore, how to integrate real-time data from high-precision electronic settlement observation device arrays based on a cloud platform, calculate the strain distribution of each part of the foundation, and accurately identify the initiation location, propagation direction, and speed of internal cracks in the main transformer foundation, combined with ultrasonic detection and acoustic emission monitoring technology, to assess their impact on the overall bearing capacity, has become a key problem that needs to be solved in this research. SUMMARY

[0003] In order to solve the problems existing in the prior art, the present application aims to provide a settlement data analysis and early warning method based on a cloud platform.

[0004] The settlement data analysis and early warning method based on a cloud platform described in the present application comprises the following steps:

[0005] S101, real-time acquisition of settlement data of the main transformer foundation by a high-precision electronic settlement observation device array, obtaining the real-time deformation distribution state of the foundation, and uploading to a cloud platform;

[0006] S102, calculate the strain values of each part of the main transformer foundation according to the real-time deformation distribution state and determine the stress concentration position, specifically through grid division of three-dimensional deformation data, strain calculation and stress analysis, obtain the stress distribution data of the foundation structure and identify the local high stress area;

[0007] S103, detect the crack at the stress concentration position of the main transformer foundation by the ultrasonic detection equipment, emit pulse signals and analyze the reflected wave data, identify the initiation position, depth and range of the crack;

[0008] S104, arrange an acoustic emission sensor array at the crack initiation position of the main transformer foundation to capture the acoustic signals when the crack expands, and determine the time sequence of the crack expansion direction and speed through signal processing and time sequence analysis;

[0009] S105, update calculation by combining the time sequence of the crack expansion direction and speed with the strain distribution matrix, analyze the influence of crack expansion on the strain distribution of the main transformer foundation and determine the local strain aggravation trend;

[0010] S106, predict the initiation position, expansion direction and speed of the crack according to the local strain aggravation trend, and calculate the dynamic change of the overall bearing capacity of the main transformer foundation through the structural mechanics simulation model to provide data support for safety evaluation;

[0011] S107, if the dynamic change value of the bearing capacity of the main transformer foundation is lower than the safety threshold, integrate the settlement observation and acoustic emission monitoring data through the cloud platform, generate a three-dimensional visual path of crack expansion and determine the repair priority area to provide a decision basis for engineering maintenance;

[0012] S108, repair the crack depth in the repair priority area through ultrasonic detection and calculate the adjusted health status value of the main transformer foundation, and generate a corresponding safety warning signal after comparing with the preset safety threshold to provide real-time feedback for engineering safety management.

[0013] Preferably, in step S101, it comprises:

[0014] Collect three-axis direction displacement data from the high-precision electronic settlement observation device array through the distributed sensor gateway according to the preset sampling period, linearly correct the original displacement data according to the sensor calibration curve, and eliminate random error interference in the corrected displacement data by using Kalman filtering to obtain filtered displacement data;

[0015] For the filtered displacement data, obtain reference coordinate point information from the pre-calibrated reference pile, establish a coordinate transformation matrix using the least squares method, and perform three-dimensional space coordinate solution on the filtered displacement data according to the coordinate transformation matrix;

[0016] The least square adjustment algorithm is used to optimize the three-dimensional space coordinates after solving, and the three-dimensional space coordinates after optimization are normalized according to the reference coordinate point information, and the normalized coordinate data is obtained.

[0017] According to the normalized coordinate data, the stress and strain parameters are obtained from the preset monitoring point, the stress and strain distribution state is calculated by using the elastic mechanics model, and the deformation stress data is obtained.

[0018] According to the deformation trend curve, the long short-term memory network is used to predict the deformation trend, and the predicted deformation data is obtained.

[0019] According to the predicted deformation data, the basic deformation distribution map is constructed, the corresponding monitoring point threshold is obtained from the pre-set deformation threshold library, and the predicted deformation data and the threshold are compared and judged.

[0020] If the predicted deformation data of the monitoring point exceeds the corresponding threshold, an early warning signal is triggered, and the early warning signal and the predicted deformation data are encrypted and transmitted to the cloud platform database for storage through the data encryption module.

[0021] Preferably, in step S102, the following steps are included:

[0022] The three-dimensional deformation data of the basic part is obtained from the deformation monitoring database, the orthogonal grid division scheme is established according to the deformation data, the hexahedral grid element division of the basic structure is carried out by using the adaptive grid partitioner, and the strain measurement position is calibrated at the grid node;

[0023] According to the grid node, the basic concrete elastic modulus parameters are obtained from the pre-established material library, the initial strain value of the node is calculated by using Hooke's law, and the strain abnormality degree is judged according to the comparison between the strain value and the preset reference strain;

[0024] According to the initial strain value of the node, the strain components are solved by using the finite element strain energy density calculation formula, the numerical integral of the strain energy density equation is carried out by using the fourth-order Runge-Kutta method, and the strain distribution cloud diagram is obtained.

[0025] According to the strain distribution cloud diagram, the node strain gradient is calculated by using the Gauss integral method, the principal value and the principal direction of strain are solved by using the principal strain calculation formula, and the basic structure strain principal direction distribution data is obtained.

[0026] According to the strain principal direction distribution data, the principal stress distribution is calculated by using the elastic mechanics strain stress conversion equation, the maximum shear stress surface is determined by using the Mohr circle criterion, and the basic structure stress distribution data is obtained.

[0027] According to the stress distribution data, a stress intensity calculation formula is used to solve a node stress intensity value, a stress concentration area is determined through a stress intensity threshold value judgment, and basic structure local stress concentration position coordinate data is obtained.

[0028] Preferably, the step S103 comprises:

[0029] According to the stress concentration position coordinates, a longitudinal wave pulse signal with a frequency range of 2-5 MHz is generated by using an array ultrasonic probe, the transmission waveform parameters are set by a probe vibrator controller, the concrete acoustic characteristic parameters are obtained from an ultrasonic wave parameter database, the ultrasonic pulse signal is subjected to acoustic impedance matching, and a matched transmission wave signal is obtained.

[0030] For the matched transmission wave signal, a reflected wave signal is collected by using an ultrasonic receiver, the reflected wave signal is subjected to band-pass filtering to eliminate concrete medium noise, and the envelope feature data is obtained by Hilbert transform to extract the signal envelope of the filtered reflected wave signal.

[0031] According to the envelope feature data, an attenuation compensation coefficient is obtained from a sound wave compensation database, the reflected wave signal strength is subjected to exponential attenuation compensation, and a compensated reflected wave signal is obtained.

[0032] For the compensated reflected wave signal, time-frequency features are extracted by using discrete wavelet transform, and characteristic coefficients are obtained by multi-scale decomposition to obtain a reflected wave time-frequency spectrum.

[0033] According to the reflected wave time-frequency spectrum, a crack standard feature vector is obtained from a defect feature database, a feature matching degree is calculated by using a random forest recognizer, and a crack type discrimination result is obtained.

[0034] For the crack type discrimination result, a crack depth parameter is calculated by using a waveform peak time difference method, and a crack propagation direction is calculated by using a phase spectrum analysis method, and crack spatial distribution feature data is obtained.

[0035] Preferably, the step S104 comprises:

[0036] According to the crack initiation position, an acoustic emission sensor array is arranged in a hexagonal topological structure, the distance between adjacent sensors is 30 cm, the sampling frequency and trigger voltage are set by a sensor matrix controller, the waveform parameter threshold is obtained from an acoustic emission calibration database, the sensitivity of the sensor array is calibrated, and calibrated acquisition parameters are obtained.

[0037] For the calibrated acquisition parameters, an acoustic emission signal is collected by using a data collector according to a preset frequency, environmental noise is eliminated by using a dynamic threshold filter, and acoustic wave event data is obtained by envelope extraction of the filtered signal.

[0038] According to the acoustic wave event data, a Fourier transform is used to extract frequency spectrum features, a frequency segment filter is used to extract feature frequency band signals, time-frequency analysis is performed on the feature frequency band signals, and an acoustic emission feature map is obtained;

[0039] According to the acoustic emission feature map, a triangular positioning algorithm is used to calculate the acoustic wave source position coordinates, a time difference of arrival of acoustic waves is used to calculate the acoustic source direction angle, the acoustic source coordinates are sequentially arranged in time, and crack propagation trajectory data is obtained;

[0040] According to the crack propagation trajectory data, a deep recurrent neural network is used to establish a time series predictor, the crack propagation rate is calculated by acoustic emission energy density, the propagation rate is accumulated in time, and a crack propagation velocity sequence is obtained;

[0041] According to the crack propagation velocity sequence, a Kalman filter is used to smooth the velocity data, the propagation direction is calculated by vector synthesis, the propagation direction and velocity are combined to construct a time series, and crack propagation state parameters are obtained.

[0042] Preferably, the step S104 further comprises:

[0043] According to the original acoustic wave signals collected by the sensor array, frequency band range parameters are obtained from an acoustic wave collection parameter library, frequency screening is performed by a band-pass filter, and self-adaptive threshold denoising is performed on the filtered signals to obtain denoised acoustic wave signals;

[0044] According to the denoised acoustic wave signals, a Hilbert transform is used to extract a signal envelope curve, an amplitude sequence is calculated from the envelope curve, and a frequency spectrum sequence is calculated from a Fourier transform to obtain acoustic wave feature data;

[0045] According to the acoustic wave feature data, a spectral peak recognizer is used to extract a feature frequency band, a wavelet decomposition is used to obtain an amplitude variation trend, and if the amplitude variation exceeds a preset monitoring threshold, a feature triggering time is recorded to obtain feature triggering data;

[0046] According to the feature triggering data, a cross-correlation function is used to calculate a sensor signal time difference sequence, array arrangement positions are obtained from a sensor coordinate library, acoustic source coordinates are calculated by an acoustic wave positioning algorithm, and crack propagation coordinates are obtained;

[0047] According to the crack propagation coordinates, an adjacent time point coordinate difference is used to calculate an expansion displacement vector, an expansion direction is determined by the direction angle of the displacement vector, and an expansion rate is calculated by dividing the expansion displacement amount by the time interval to obtain crack propagation parameters;

[0048] According to the crack propagation parameters, a support vector regression predictor is used to construct a velocity prediction model, continuous prediction results are obtained by time window sliding, and the expansion direction and velocity data are time-synchronized to obtain a crack propagation time sequence.

[0049] Preferably, the step S105 comprises:

[0050] According to the crack propagation time series, the strain redistribution calculator is used to update the basic strain distribution data, the strain measured value is extracted from the strain monitoring database, the expansion influence area boundary is determined through strain difference calculation, and the strain real-time distribution matrix is obtained;

[0051] For the strain real-time distribution matrix, the tensor mapping method is used to calculate the strain change around the crack propagation trajectory, the strain field reconstruction result is obtained from the finite element calculator, the strain distribution update data is established through the grid interpolation method, and the strain influence tensor is obtained;

[0052] According to the strain influence tensor, the recursive neural network is used to establish the strain transfer function, the influence of the expansion parameter on the strain field is calculated from the strain transfer function, the influence coefficient of each monitoring point is determined through strain coupling calculation, and the strain transfer matrix is obtained;

[0053] For the strain transfer matrix, the strain time series is used to extract the strain change rate of the high stress area, the mutation point position is identified from the strain change rate curve, the strain aggravation area range is determined through local strain gradient calculation, and the local strain aggravation trend is obtained.

[0054] Preferably, the step S106 comprises:

[0055] According to the local strain aggravation trend data, the strain gradient field distribution is calculated by using the five-point difference format, the strain growth rate is extracted from the strain gradient field data, the mutation point is identified through the strain rate curve, and the strain mutation position data is obtained;

[0056] For the strain mutation position data, the least square fitting is used to calculate the strain contour line, the crack initiation position is determined from the contour line curvature change, the expansion trend angle is calculated through the principal strain direction, and the crack space parameter is obtained;

[0057] According to the crack space parameter, the deep neural network is used to predict the expansion path, the expansion speed is calculated from the strain growth curve, and the space-time mapping of the expansion path and speed is performed, and the crack propagation prediction data is obtained;

[0058] For the crack propagation prediction data, the concrete elastic modulus, Poisson's ratio and compressive strength parameters are read from the basic parameter library, the structure stiffness degradation amount is calculated through the elastoplastic damage function, and the stiffness damage data is obtained;

[0059] According to the stiffness damage data, the stress redistribution algorithm is used to calculate the basic structure stress distribution, the principal stress direction is extracted from the stress distribution cloud map, the bearing level is calculated through the principal stress and allowable stress ratio, and the bearing capacity change data is obtained;

[0060] For the aforementioned load-bearing capacity variation data, a multi-point regression algorithm is used to fit the load-bearing capacity evolution curve. The load-bearing values ​​at key moments are extracted from the evolution curve, and the overall load-bearing capacity is determined by calculating the load-bearing ratio, thus obtaining the load-bearing capacity assessment results of the foundation structure.

[0061] Preferably, step S107 includes:

[0062] Based on the dynamic change value of the basic bearing capacity, the structural bearing capacity safety threshold is obtained from the safety monitoring database. The threshold is judged based on the change value of the bearing capacity. If it is lower than the safety threshold, the settlement displacement data and acoustic emission data are extracted from the online monitoring platform to obtain the monitoring fusion data.

[0063] For the aforementioned monitoring fusion data, a time-series matching algorithm is used to align the monitoring data in time, and a spatial coordinate transformation is used to establish a mapping relationship between the settlement monitoring points and the acoustic emission source locations to obtain calibration coordinate data;

[0064] Based on the calibration coordinate data, a deep convolutional neural network is used to extract the spatial features of the crack, the deformation trend is calculated from the surface displacement monitoring data, and a crack propagation mapping function is established by combining features to obtain the crack propagation spatial data.

[0065] For the crack propagation spatial data, a three-dimensional mesh generator is used to construct a crack propagation reference surface. Depth layer data is extracted from the reference surface, and point cloud fitting is performed through an iterative nearest point algorithm to obtain the crack propagation three-dimensional contour.

[0066] Based on the three-dimensional profile of the crack propagation, a regional damage distribution map is established using a load-bearing capacity loss calculator. The hazard level is extracted from the damage distribution map, and the hazard levels are sorted using the analytic hierarchy process to obtain the structural hazard classification.

[0067] For the structural hazard classification, a priority sorting algorithm is used to determine the order of repair areas, the urgency of repair is calculated from the degree of damage to the area, and the area classification identifier is generated through the repair level matrix to obtain repair priority data.

[0068] Preferably, step S108 includes:

[0069] Based on the coordinates of the priority repair area, an ultrasonic detection array is used to emit probe waves. The ultrasonic frequency and amplitude parameters are extracted from the standard waveform library. The echo signal is extracted through a bandpass filter. The filtered signal is then subjected to Hilbert transform to obtain the ultrasonic envelope curve.

[0070] For the ultrasonic envelope curve, a peak identification algorithm is used to extract the peak position of the reflected wave, the crack depth parameter is calculated from the reflected wave time difference, and the actual crack depth value is calculated through acoustic attenuation compensation to obtain crack depth distribution data;

[0071] According to the crack depth distribution data, a stress redistribution function is constructed by using an elastic-plastic mechanics calculator, elastic modulus and strength parameters are extracted from a material parameter library, the structural bearing capacity is calculated through a stress balance equation, and foundation bearing state data is obtained;

[0072] According to the foundation bearing state data, a state evaluation function is established by using a depth feedback neural network, displacement field and strain field features are extracted from monitoring data, the foundation structure comprehensive score is calculated through feature fusion, and the foundation health index is obtained;

[0073] According to the foundation health index, a warning grading standard is established by using a fuzzy evaluation matrix, safety threshold parameters are obtained from a warning level library, if the health index is lower than the safety threshold, a corresponding level warning mark is generated, and the foundation structure safety state is obtained;

[0074] According to the foundation structure safety state, a warning signal is constructed by using a warning information generator, sound and light warning parameters are extracted from a signal feature library, the warning signal type is determined through a warning level matrix, and the foundation safety warning data is obtained.

[0075] The settlement data analysis and warning method based on the cloud platform has the advantages that the foundation settlement data is collected in real time by the high-precision electronic settlement observation device array, the strain distribution matrix is determined by combining the finite element analysis and the material mechanics model, the stress concentration position is identified, the crack initiation position is located by using the ultrasonic detection technology, and the sound wave signals in the crack expansion process are captured by the sensor array.

[0076] The application predicts the crack expansion trend by analyzing the correlation between the crack expansion parameters and the strain distribution matrix, inputs the crack expansion trend into the structural mechanics simulation model, calculates the dynamic change of the foundation bearing capacity, generates the three-dimensional visual path of the crack expansion when the bearing capacity is lower than the safety threshold, determines the repair priority area, and generates the safety warning signal according to the ultrasonic detection result.

[0077] The method realizes the early warning and accurate repair guidance of the main transformer foundation crack, and effectively guarantees the safe operation of the transformer substation. BRIEF DESCRIPTION OF DRAWINGS

[0078] Figure 1 is the flow of the settlement data analysis and warning method based on the cloud platform Figure 1 ;

[0079] Figure 2 is the flow of the settlement data analysis and warning method based on the cloud platform Figure 2 . DETAILED DESCRIPTION

[0080] As Figures 1-2As shown, the settlement data analysis and early warning method based on a cloud platform described in this application includes the following steps:

[0081] S101. Real-time settlement data of the main transformer foundation is collected through a high-precision electronic settlement observation device array to obtain the real-time deformation distribution status of the foundation and upload it to the cloud platform.

[0082] S102. Calculate the strain values ​​of each part of the main transformer foundation based on the real-time deformation distribution state and determine the stress concentration location. Specifically, obtain the stress distribution data of the foundation structure and identify local high stress areas through mesh division, strain calculation and stress analysis of three-dimensional deformation data.

[0083] S103. Use ultrasonic testing equipment to detect cracks at stress concentration points in the main transformer foundation, emit pulse signals and analyze reflected wave data to identify the location, depth and extent of crack initiation;

[0084] S104. An array of acoustic emission sensors is arranged at the location where the crack in the main transformer foundation begins to capture the acoustic signal when the crack expands. The time series of crack expansion direction and velocity is determined by signal processing and time series analysis.

[0085] S105. By combining the time series of crack propagation direction and velocity with the strain distribution matrix, the effect of crack propagation on the strain distribution of the main transformer foundation is analyzed and the local strain intensification trend is determined.

[0086] S106. Based on the trend of local strain intensification, predict the initiation location, propagation direction and speed of cracks, and calculate the dynamic changes of the overall bearing capacity of the main transformer foundation through a structural mechanics simulation model to provide data support for safety assessment.

[0087] S107. If the dynamic change value of the bearing capacity of the main transformer foundation is lower than the safety threshold, the settlement observation and acoustic emission monitoring data are integrated through the cloud platform to generate a three-dimensional visualization path of crack propagation and determine the priority area for repair, so as to provide a basis for decision-making for engineering maintenance.

[0088] S108. The depth of cracks in the priority repair area is detected by ultrasonic testing, and the adjusted health status value of the main transformer foundation is calculated. After comparing it with the preset safety threshold, a corresponding safety warning signal is generated to provide real-time feedback for engineering safety management.

[0089] like Figures 1-2 As shown, in step S101, the settlement data of the main transformer foundation is collected in real time by a high-precision electronic settlement observation device array to obtain the real-time deformation distribution status of the foundation and upload it to the cloud platform.

[0090] Further, in step S101, S1011, the distributed sensor gateway collects displacement data in three-axis direction from the high-precision electronic settlement observation device array at a preset sampling period, linearly corrects the original data based on the sensor calibration curve, and then uses Kalman filtering technology to eliminate random error interference to obtain smooth filtered displacement data.

[0091] Then, the reference coordinate point information is obtained from the pre-calibrated reference pile, the least square method is used to construct the coordinate transformation matrix, the three-dimensional space coordinate of the filtered displacement data is calculated and normalized, and the normalized coordinate data is generated.

[0092] Subsequently, the normalized coordinate data is uploaded to the cloud platform to provide basic data support for analysis. The sampling period can be set according to actual needs, for example, set to 100Hz to realize millimeter-level precision collection, and the sensor gateway is deployed in a star-shaped topology structure, and each gateway connects multiple sensors to form a collaborative observation network.

[0093] S1012, after obtaining the normalized coordinate data, stress and strain parameters are extracted from the preset monitoring points, the stress and strain distribution state of each part of the foundation is calculated by using the elastic mechanics model, and the deformation stress data is generated; the deformation trend curve is constructed according to the deformation stress data, the long short-term memory network is used to predict the deformation trend in a period of time in the future, the predicted deformation data is obtained and the foundation deformation distribution map is constructed; the threshold value of each monitoring point is obtained from the preset deformation threshold value library, compared with the predicted deformation data, if the threshold value is exceeded, a warning signal is triggered, and the warning signal and the predicted deformation data are encrypted and transmitted to the cloud platform database through the data encryption module. The monitoring points can be divided into multiple units, for example, the foundation is divided into 25 regions, and stress and strain sensors are configured in each region to cover the key load-bearing parts.

[0094] In the embodiment of the present application, the deployment of the high-precision electronic settlement observation device array fully considers the multi-point monitoring needs of the main transformer foundation. The distributed sensor gateway connects 6 sensors through a star-shaped topology structure, the calibration curve uses a piecewise linear fitting method to ensure that the error is less than 0.01mm within the range of 0-50mm. Kalman filtering effectively removes Gaussian white noise through the prediction and update stages, and the smoothness of the filtered data is significantly improved. The reference coordinate points are selected at the four corners and the center position of the foundation, and are fixed by using a 15-meter deep reinforced concrete pile to ensure that the coordinate system conversion accuracy is better than 0.1mm.

[0095] For stress-strain analysis, the elastic mechanics model comprehensively considers the foundation bearing capacity and load distribution to generate stress field distribution data. The deformation trend curve calculates the displacement, velocity and acceleration characteristics based on a 48-hour sliding window. The long short-term memory network adopts a three-layer structure with 128 hidden neurons to predict the deformation trend in the next 24 hours with an error controlled within 0.5 mm. The deformation threshold library sets a 5-mm threshold for load-bearing areas and an 8-mm threshold for non-load-bearing areas. The data transmission uses a 2048-bit asymmetric encryption algorithm to ensure security.

[0096] In practical applications, for example, in the case of monitoring the main transformer foundation of a certain 380kV substation, the system detected that the 48-hour cumulative settlement of the northwest corner point was 4.2mm, with a settlement rate of 0.175mm / h, and predicted that the settlement would be close to 7.8mm after 48 hours. The cloud platform promptly issued a warning signal, and after the engineering personnel implemented reinforcement, the settlement rate decreased to below 0.02mm / h, verifying the effectiveness of the method.

[0097] It can be understood that the embodiments of the present application do not excessively limit the specific parameters of the sampling frequency, the number of sensors or the network structure, which can be adjusted by the technician according to the actual scene to adapt to different monitoring needs. The subsequent steps will further analyze the strain and crack state based on this real-time deformation data, and the specific implementation will be described in detail in subsequent embodiments.

[0098] As shown in Figures 1-2 In step S102, the strain values of each part of the main transformer foundation are calculated according to the real-time deformation distribution state, and the stress concentration position is determined. Specifically, through grid division, strain calculation and stress analysis of three-dimensional deformation data, stress distribution data of the foundation structure are obtained, and local high stress areas are identified.

[0099] Further, in step S102, S1021, three-dimensional deformation data of the main transformer foundation is extracted from the deformation monitoring database stored in the cloud platform, hexahedral grid element division is performed on the foundation structure using an adaptive grid partitioner to generate grid node coordinate data, then the elastic modulus and Poisson's ratio parameters of concrete are read from the pre-constructed material attribute library, the initial strain values of each grid node are calculated using Hooke's law, and compared with the preset baseline strain threshold to determine whether there is an abnormal strain area;

[0100] For abnormal strain regions, the finite element analysis method is used to calculate the strain energy density, and the energy equation is numerically integrated by the fourth-order Runge-Kutta method to generate the strain distribution cloud data. When meshing, the base body adopts orthogonal mesh with a side length of 200 mm, and the key parts such as supports and edges are encrypted to a side length of 100 mm to improve the calculation resolution of stress concentration areas. The strain energy density, as the core index to measure the deformation capacity of the material, considers the elastic potential energy accumulation of the concrete under loading in the calculation process, and the typical value range is between 0.1 and 0.5 kJ / m 3 , and the integral step is set to 0.01 s to ensure accuracy.

[0101] S1022, for the generated strain distribution cloud data, the strain gradient of each grid node is calculated by using the Gaussian integral method, and the gradient change rate is extracted, the principal strain value and its direction are solved by the principal strain calculation formula combined with the three-dimensional characteristic equation, and then the principal stress distribution is calculated by using the stress and strain conversion relationship in elastic mechanics to generate the stress distribution data of the foundation structure;

[0102] Then the maximum shear stress plane and the angle between it and the horizontal plane are analyzed by the Mohr circle criterion, and the stress intensity value of each node is evaluated by the stress intensity calculation formula, if the stress intensity of a node exceeds twice the average value, it is determined as a stress concentration area and its coordinate position is recorded, the stiffness matrix correction is introduced in the principal strain calculation to reflect the anisotropic properties of the concrete in different directions, and the stress distribution data is derived by the generalized Hooke's law to ensure that the results are consistent with the actual stress state.

[0103] In the embodiment of the application, the collection frequency of three-dimensional deformation data is set to 10 Hz, which can record the small displacement changes of the foundation surface in real time. The concrete parameters stored in the material attribute library include an elastic modulus of 30 GPa, a Poisson's ratio of 0.2, and a compressive strength of 30 MPa. These parameters are suitable for stable concrete with an age of more than 90 days, and strain calculation is based on a linear elastic constitutive model. The reference strain threshold is set to 200 micro-strains, which serves as a reference standard for abnormality determination.

[0104] The strain distribution cloud is displayed in the form of color contour lines, graded by 100 micro-strains from 0 to 1000 micro-strains, which facilitates intuitive identification of high-strain areas. The fourth-order Runge-Kutta method improves calculation stability through multi-step iteration in numerical integration, with an error control within 0.1%. During principal stress distribution calculation, special attention is paid to the area below the load application point, as it is often the starting position of stress concentration. The measured case shows that the maximum principal strain value of a 500 kV transformer substation main transformer foundation is about 400 micro-strains, located directly below the support, with a stress intensity of 12 MPa, close to 40% of the design strength.

[0105] S1023According to the stress distribution data, the characteristics of the stress concentration area are further analyzed, the distribution law of the local high stress point is determined by the stress intensity ratio method, and is marked in the basic three-dimensional model, for example, an elliptical stress concentration area is identified within 300mm around the transformer support, the maximum shear stress surface is at an angle of 45 degrees with the horizontal plane and points to the edge; these coordinate data provide accurate spatial reference for the prediction of crack initiation position, and the identification of the stress concentration area also helps to optimize the arrangement of monitoring points. The stress intensity calculation formula considers the superposition effect of principal stress and shear stress, ensuring the reliability of the determination result.

[0106] In the embodiment of the application, the application of the adaptive mesh partitioner improves the calculation efficiency, and the division scheme is dynamically adjusted according to the spatial distribution of the deformation data, avoiding the problem of insufficient precision of the traditional uniform mesh in the high strain area. The application of the Mohr circle criterion provides theoretical support for the direction analysis of the shear stress surface, making the determination of the stress concentration area more scientific. In actual monitoring, the coordinate data of the stress concentration position can be directly imported into the three-dimensional visualization module of the cloud platform, which is convenient for engineers to quickly locate and take targeted measures.

[0107] It can be understood that the application does not excessively limit the specific values of the mesh size, calculation step or material parameters, which can be adjusted by the technician according to the actual engineering requirements to adapt to different analysis scenarios of the foundation structure. Subsequent steps will carry out crack detection and expansion analysis based on the stress concentration position.

[0108] As shown in Figures 1-2 , in step S103, the stress concentration position of the main transformer foundation is detected for cracks by an ultrasonic detection device, a pulse signal is emitted, and reflection wave data is analyzed to identify the initiation position, depth and range of the crack.

[0109] Further, in step S103, in the embodiment of the application, for the coordinate data of the stress concentration position, a longitudinal wave pulse signal with a frequency of 2MHz to 5MHz is generated by using an array ultrasonic probe, the transmission waveform parameters are set by a probe vibrator controller, and sound impedance matching is performed in combination with the acoustic characteristics of concrete to generate an optimized transmission wave signal; then, reflection wave data is collected by an ultrasonic receiver, noise is removed by a band-pass filter, and envelope characteristics are extracted by Hilbert transform, and then, in combination with attenuation compensation and time-frequency analysis, the spatial distribution characteristics of the crack are obtained. The array probe is linearly arranged with 8 probes, the spacing is 25mm, the coverage length is 200mm, the transmission frequency is preferably 3MHz to adapt to the wavelength of 1.2mm in concrete, and the sound wave propagation speed is 3600m / s. The sound impedance matching is optimized by a coupling agent, the sound impedance of concrete is 8×10 6 kg / m 2 s, and the sound impedance of the coupling agent is 4×10 6 kg / m2 s, ensuring energy-efficient delivery.

[0110] S1031, after the emission wave signal is generated, a Gaussian modulation pulse form is adopted, the pulse width is set to 1 mu s, the repetition frequency is 1 kHz, the reflected wave signal is captured through an ultrasonic receiver, low-frequency aggregate scattering noise and high-frequency electromagnetic interference are filtered out by using a band-pass filter with a frequency band range of 2.5 MHz to 3.5 MHz, the attenuation band suppression ratio is greater than 40 dB; then the filtered reflected wave signal is subjected to Hilbert transform, time domain envelope feature data are extracted to highlight the echo peak value distribution, the correlation between the peak value amplitude and the crack size provides a basis for subsequent analysis; for the envelope feature data, an attenuation compensation coefficient, for example, 0.8 dB / cm, is read from a sound wave compensation database, and exponential attenuation compensation is performed on the reflected wave signal, so that the deep signal strength is improved, for example, the amplitude at a depth of 80 cm can be improved by 12 dB, thereby enhancing the detection sensitivity of deep cracks. The Hilbert transform decomposes the signal into amplitude and phase components, so that the time domain features of the crack reflected wave are more obvious, and subsequent feature extraction is facilitated.

[0111] S1032, after the compensated reflected wave signal is obtained, a discrete wavelet transform is used to extract time-frequency features, a db4 wavelet basis function is used for 5-layer multi-scale decomposition, and a reflected wave time-frequency spectrum is generated, wherein the main frequency component is concentrated in 2.8 MHz to 3.2 MHz, and the spectrum is widened by about 0.4 MHz; according to the time-frequency spectrum, a crack standard feature vector is extracted from a defect feature database, sample data of different depths and angles are covered, a feature matching degree is calculated by using a random forest recognizer, training samples contain 1000 groups of data, feature dimensions include 15 items of time domain waveform, frequency spectrum and statistical parameters, and a crack type discrimination result is output; further, a crack depth is calculated by using a waveform peak time difference method, the travel time resolution is 0.1 mu s, the depth accuracy is better than 2 mm, a crack extension direction is determined by using a phase spectrum analysis method, the phase sensitivity is 0.5 degrees / degree, and crack spatial distribution feature data are generated. The multi-scale decomposition of the wavelet transform can separate noise and effective signals, the time-frequency spectrum directly reflects the frequency distribution characteristics of the crack reflected wave, and the random forest improves the recognition robustness through multi-decision tree voting.

[0112] In the embodiment of the present application, the implementation of ultrasonic detection aims to solve the problem of insufficient identification accuracy of deep cracks in the traditional method. The acoustic impedance matching improves the detection sensitivity by reducing the sound wave reflection loss, and the band-pass filter effectively isolates the complex interference in the concrete medium. The application of attenuation compensation enables the deep reflected wave signal to be restored, and ensures the detection ability of the through cracks.

[0113] In the actual detection of the main transformer foundation of a 500 kV substation, a penetrating crack with a crack depth of 15 cm, an inclination angle of 35 degrees, and an extension length of 45 cm is found. The reflected wave amplitude is 6 dB higher than that of the complete area, the frequency spectrum is widened by 0.6 MHz, and the phase difference changes by more than 15 degrees. Through time difference method and phase analysis, the crack position and trend are accurately calibrated, providing key data support for reinforcement measures. The extraction and matching process of crack characteristics fully embodies the accuracy of the method, and the detection results can be directly used for subsequent analysis on the cloud platform.

[0114] It can be understood that the present application does not excessively limit the specific settings of the number of probes, frequency range or filtering parameters, which can be adjusted by the technician according to the actual concrete characteristics. For example, if the detection object is high-density concrete, the transmission frequency can be appropriately increased to shorten the wavelength and improve the resolution.

[0115] As shown in Figures 1-2 In step S104, an acoustic emission sensor array is arranged at the crack initiation position of the main transformer foundation to capture acoustic signals during crack propagation, and the time sequence of crack propagation direction and speed is determined through signal processing and time sequence analysis.

[0116] Further, in step S104, in the embodiment of the present application, the acoustic emission sensor array is arranged in a hexagonal topology according to the crack initiation position, the central sensor is placed directly above the crack initiation point, and the surrounding six sensors are distributed in a regular hexagon with an adjacent spacing of 300 mm. The sampling frequency is set to 1 MHz and the trigger voltage is set to 50 mV through the sensor matrix controller, the standard waveform parameters are obtained from the acoustic emission calibration database for sensitivity calibration, and the calibrated acquisition parameters are generated. The acoustic emission signal is collected by the data collector, the 100 kHz to 300 kHz frequency band signal is screened by the band-pass filter, and the environmental noise is removed by the adaptive threshold filter, the threshold is set to 2.5 times the root mean square value of the background noise, and the clear acoustic event data is obtained. The dynamic adjustment of the adaptive threshold can optimize the filtering effect in real time according to the noise level on site, ensuring the signal quality, and the hexagonal topology improves the coverage range and accuracy of acoustic source positioning through multi-point cooperative monitoring.

[0117] S1041, for the collected acoustic wave event data, the Fourier transform is used to calculate the frequency spectrum characteristics, the main frequency band signal of 150 kHz to 250 kHz is extracted, and the time-frequency spectrum is generated by 8-level decomposition through db4 wavelet base function, the frequency distribution and time domain change of acoustic emission event are clearly displayed;Then, the acoustic source position is calculated based on the acoustic wave arrival time difference using the triangular positioning algorithm, the positioning accuracy is better than 5mm, the direction angle resolution is 2 degrees, the crack propagation trajectory data is generated by arranging the acoustic source coordinates in time sequence;Further, a time series prediction model is constructed by using a deep recurrent neural network, the network includes three layers, the input layer has 128 time sequence nodes, the hidden layer has 256 neurons, the crack propagation rate is analyzed by using an acoustic emission energy density calculator, the window width is 10ms, the sliding step is 2ms, and the crack propagation state parameters are output. The triangular positioning algorithm restores the spatial position of the acoustic source by analyzing the time difference between multiple sensors, and the deep recurrent neural network predicts the future expansion trend by learning historical data, thereby improving the prediction ability of the analysis.

[0118] S1042, after obtaining the crack propagation trajectory data, the Hilbert transform is used to extract the envelope curve of the acoustic wave signal to reflect the energy release process, the time difference sequence between adjacent sensors is calculated by using the cross-correlation function, the acoustic source coordinates are calculated combined with the sensor coordinate library, and the expansion displacement vector is determined;The expansion direction is calculated according to the direction angle of the displacement vector, the expansion rate is calculated by the ratio of the displacement amount to the time interval, and a speed prediction model is constructed by using a support vector regression predictor, the kernel function is selected as a radial basis function, the parameter γ is 0.1, the penalty factor C is 100, and the training sample is taken as the latest 100 event data, thereby generating continuous expansion speed prediction results;The speed data is smoothed by using a Kalman filter, the process noise covariance is set as 0.01, the measurement noise covariance is 0.1, the expansion direction is calculated combined with vector synthesis, and the crack expansion time sequence including coordinates, speed and direction is generated. The support vector regression optimizes the accuracy of speed prediction through nonlinear mapping, and the Kalman filter effectively eliminates the random fluctuations in the data, thereby ensuring the smoothness and reliability of the time sequence.

[0119] In the embodiment of the application, the collection bandwidth of the acoustic emission signal is set to 100 kHz to 400 kHz, the preamplification multiple is 40 dB, and the single-channel buffer depth is 32 MB, so as to meet the real-time storage requirements of high-frequency signals. The measured data shows that the amplitude of the acoustic emission signal during crack propagation fluctuates between 100 mV and 500 mV, the main frequency is concentrated near 150 kHz, the spectrum is widened by about 50 kHz, and the duration is 0.5 ms to 2 ms. The monitoring of the main transformer foundation of a certain 500 kV substation shows that the crack expands at an angle of 45 degrees under lateral load, 458 acoustic emission events are recorded within 24 hours, 80% of which are in the rapid expansion stage, and the event density within 50 mm of the crack initiation point reaches 12 / cm 2The expansion speed is up to 2mm / s, the average speed is about 0.5mm / s, and the total length is 45mm.

[0120] It is understood that this invention does not strictly limit the specific values ​​of sensor spacing, sampling frequency, or filtering parameters, which can be adjusted according to the crack size and monitoring environment. For example, in high-noise scenarios, the threshold factor can be appropriately increased to enhance anti-interference capabilities. Subsequent steps will further evaluate the impact of the cracks on the foundation based on this time series.

[0121] like Figures 1-2 As shown, in step S105, the strain distribution matrix is ​​updated by combining the time series of crack propagation direction and velocity, and the influence of crack propagation on the strain distribution of the main transformer foundation is analyzed and the local strain intensification trend is determined.

[0122] Further, in step S105, in this embodiment of the invention, measured strain values ​​are extracted from the strain monitoring database and the strain distribution of the foundation is updated using a strain redistribution calculator. The acquisition frequency is set to 10Hz, and strain data is acquired using a fiber optic grating sensor. The measurement range covers ±5000 microstrains, with a resolution of 1 microstrain. The foundation is divided into 200 monitoring units of 100mm × 100mm to record strain changes in real time. The strain difference is calculated based on the crack propagation time series to determine the boundary of the affected area, generating a real-time strain distribution matrix. Subsequently, the tensor mapping method is used to calculate the strain change around the crack propagation trajectory. Spatial discretization is performed using 9-node quadrilateral elements, and the strain field is reconstructed using a bilinear interpolation function. Combined with the results from the finite element calculator, the radial basis function interpolation method is used to establish the updated strain distribution data. The interpolation radius is set to 100mm, and the weight decays exponentially with distance, generating a strain influence tensor. Tensor mapping transforms the local effects of crack propagation into global strain field changes through discretization, ensuring the spatial continuity of the calculation results.

[0123] S1051, for the strain influence tensor, a recurrent neural network is used to construct a strain transfer function to quantify the impact of crack propagation on the strain field, the network contains 128 time nodes, the hidden layer uses long short-term memory units, the input features include crack propagation speed, direction angle and strain increment, the training data consists of 1000 sets of 60s long strain field evolution sequences, the strain influence coefficient of each monitoring point is determined by strain coupling calculation, and the strain transfer matrix is generated; then the strain time series of the high stress area is extracted from the strain transfer matrix, the strain change rate is calculated by using a 10s sliding window and a 1s step, the mutation point is identified by wavelet transform, the position with increased wavelet coefficient amplitude is the strain mutation point, the local strain gradient is calculated by the central difference format, the grid spacing is 50mm, the strain aggravation area range is determined and the strain concentration trend data is generated. The recurrent neural network can effectively simulate the dynamic process of strain transfer by capturing the time dependence, and the wavelet transform enhances the detection sensitivity of the mutation point.

[0124] In the embodiments of the present application, the influence of crack propagation on the basic strain distribution presents significant spatial heterogeneity. The measured data shows that the strain gradient in the range of 50mm in front of the crack tip rises sharply, the maximum strain value increases from 300 microstrain to 800 microstrain, and the increment shows an exponential distribution, reflecting the amplification effect of crack propagation on the local stress field. The monitoring of the main transformer foundation of a certain 500kV substation shows that the crack propagation causes significant changes in the strain field within 80mm, the influence coefficient decays from the crack center to the outside, to the boundary, to 20% of the initial value, and for every increase of 0.1mm / s in the propagation speed, the influence range expands by about 5mm.

[0125] S1052, in the strain concentration trend analysis, the strain change rate of the high stress area increases rapidly from the initial 5 microstrain / s to 20 microstrain / s, the strain aggravation area presents an elliptical distribution, the long axis is consistent with the crack propagation direction, and the short axis is perpendicular to the propagation direction, the area expands 2.5 times within 24 hours, and the maximum strain gradient reaches 15 microstrain / mm, especially at the crack intersection. Through the analysis of the strain transfer matrix, it is found that the crack propagation speed is positively correlated with the strain transfer coefficient, and this correlation provides data support for predicting the local strain aggravation, and the generation of strain concentration trend data helps to identify potential high-risk areas. The strain gradient calculation quantifies the spatial variation rate of strain by the difference method, which provides accurate basis for dynamic tracking of the crack influence range.

[0126] It can be understood that the present application does not strictly limit the specific settings of the monitoring unit division, interpolation radius or network parameters, which can be flexibly adjusted according to the foundation scale and crack characteristics, for example, the monitoring unit density can be increased in complex stress areas to improve the resolution. The subsequent steps will further evaluate the foundation bearing capacity based on this strain trend data.

[0127] AsFigures 1-2 As shown in step S106, the initiation position, expansion direction and speed of the crack are predicted according to the local strain aggravation trend, and the dynamic change of the overall bearing capacity of the main transformer foundation is calculated through a structural mechanics simulation model to provide data support for safety evaluation.

[0128] Further, in step S106, in the embodiment of the present application, the strain gradient field distribution is calculated using five-point difference format with local strain aggravation trend data, the grid spacing is set to 50 mm, and in the high strain area, the grid spacing is encrypted to 25 mm to improve the resolution; the strain growth rate curve is extracted from the strain gradient field, and the strain mutation position data is obtained by identifying the curve mutation point. The strain rate at the mutation point can jump from 15 micro-strain / s to 35 micro-strain / s within 0.1 s; for the mutation position data, the strain contour is fitted using the least squares method, the contour interval is 0.5 mm, and in the crack initiation area, the interval is reduced to 0.1 mm. The crack initiation position is determined by the area with a curvature radius less than 100 mm, and the expansion trend angle is calculated combined with the principal strain direction to generate the crack spatial parameters. The five-point difference format improves the accuracy of gradient calculation through multi-point numerical approximation, and the contour curvature analysis intuitively reflects the local abnormality of the strain field, providing a reliable basis for crack positioning.

[0129] S1061, for the crack spatial parameters, a deep neural network is used to predict the crack expansion path and speed, the network contains a three-layer structure, the input layer integrates 15 features of strain field, stress field and displacement field, the hidden layer is configured with 256 neurons, and the output layer generates expansion path and speed prediction results; the expansion speed is calculated from the strain growth curve, the path and speed are combined through space-time mapping to generate crack expansion prediction data, the prediction shows that the angle between the initial expansion direction and the maximum principal stress is about 45 degrees, and the speed can reach 0.8 mm / s; then the concrete material parameters are read from the foundation parameter library, including elastic modulus 30 GPa, Poisson's ratio 0.2, compressive strength 30 MPa and tensile strength 2.5 MPa, the stiffness degradation amount is calculated using the elastoplastic damage function, the damage factor increases from 0 to 0.35, and the degradation area is distributed in a fan shape with an apex angle of 60 degrees, and the influence range is twice the length of the crack. The deep neural network optimizes the accuracy of path prediction through multi-dimensional feature learning, and the elastoplastic damage function quantifies the weakening effect of the crack on the material stiffness.

[0130] S1062, after obtaining the crack propagation prediction data, a stress redistribution algorithm is used to calculate the stress distribution of the foundation structure, the principal stress direction is extracted from the stress distribution nephogram, and the principal stress to allowable stress ratio is calculated to evaluate the bearing level, and the bearing capacity change data after stiffness damage is generated; a multiple point regression algorithm is used to fit the bearing capacity evolution curve, 50 feature points on the foundation surface are selected for calculation, the bearing value at the key moment is extracted and the overall bearing capacity is determined through the bearing ratio, the foundation is divided into 25 calculation units, and the sectional accumulation method is used to evaluate the dynamic change of bearing capacity. The stress redistribution algorithm simulates the stress transfer process caused by cracks through iterative calculation, and the multiple point regression captures the nonlinear downward trend of the bearing capacity through curve fitting.

[0131] In the embodiment of the present application, the main transformer foundation monitoring of a certain 500kV substation shows that the crack propagation significantly increases the stress within 50mm, the principal stress direction deviates by 25 degrees, forming a new stress concentration area, and the bearing capacity decreases in steps, about 2% for every 15mm of expansion, a total of 12% within 24 hours, of which the rapid expansion stage accounts for 80%. The maximum decrease of the local unit bearing capacity is 35%, the decrease of the adjacent unit is between 15% and 15%, the final bending bearing capacity decreases by 18%, and the shear bearing capacity decreases by 22%.

[0132] It can be understood that the present application does not make fixed restrictions on the specific values of the grid spacing, the number of neural network layers or the material parameters, which can be adjusted according to actual engineering needs, for example, the grid can be further densified in the crack dense area to improve the calculation accuracy. The subsequent steps will generate an early warning signal based on this bearing capacity data.

[0133] As shown in Figures 1-2 , in step S107, if the dynamic change value of the main transformer foundation bearing capacity is lower than the safety threshold, the settlement observation and acoustic emission monitoring data are integrated through the cloud platform to generate a three-dimensional visual path of crack propagation and determine the repair priority area, providing a decision basis for engineering maintenance.

[0134] Further, in step S107, in the embodiment of the application, the structural bearing capacity safety threshold is extracted from the safety monitoring database, which is usually set to 80% of the design bearing capacity, for example, the design bearing capacity of the foundation of a certain 500 kV substation is 24 MPa, and the safety threshold is 19.2 MPa, when the measured value drops to 17.8 MPa, the early warning is triggered; for the case below the threshold, the settlement displacement data and acoustic emission data are obtained from the online monitoring platform, the settlement monitoring frequency is 10 Hz, and the acoustic emission monitoring frequency is 1 MHz, through the time sequence matching algorithm, 1 s sliding window and 0.1 s step are used for time alignment, the alignment accuracy is better than 0.1 ms, then the least square method is used for spatial coordinate transformation, the mapping relationship between the settlement monitoring point and the acoustic emission source is established, the registration accuracy reaches 2 mm, and the calibration coordinate data is generated. Time sequence matching ensures the synchronicity of multi-source data, and coordinate transformation realizes the accurate correspondence of spatial position through matrix calculation.

[0135] S1071, for the calibration coordinate data, a deep convolutional neural network is used to extract crack spatial features, the network is designed as a 5-layer structure, the input features include displacement field, strain field and acoustic emission signal features, and the output is a three-dimensional feature vector of the crack; the deformation trend is calculated from the surface displacement data, it is found that the maximum settlement amount reaches 12 mm, the settlement basin is elliptical, the long axis is consistent with the crack trend, the crack propagation mapping function is constructed through feature combination, the surface deformation and internal crack are associated, the mapping accuracy reaches 85%, the crack propagation spatial data is generated; then a three-dimensional grid divider is used to construct a crack propagation reference surface with 50 mm hexahedral elements, and the crack area is encrypted to 25 mm, the depth layered data is extracted from the reference surface, the point cloud fitting is performed through the iterative closest point algorithm combined with KD tree acceleration search, the iteration is 50 times, the convergence threshold is 0.1 mm, the crack propagation stereo profile is generated, the crack depth is 180 mm, the width is 0.8 mm, and the crack extends at an angle of 45 degrees. The deep convolutional neural network extracts the spatial pattern of the crack through multiple layers of convolution and pooling, and the iterative closest point algorithm optimizes the efficiency and accuracy of point cloud fitting.

[0136] S1072、After generating the crack propagation stereoscopic profile, a bearing capacity loss calculator is used to construct a regional damage distribution map, with red, yellow and blue colors representing damage levels, with red being greater than 0.6, yellow being 0.3 to 0.6, and blue being less than 0.3. The risk level is sorted by the analytic hierarchy process, considering crack depth, width and propagation speed, with weights of 0.5, 0.3 and 0.2 respectively. Three dangerous areas are calculated, with the area under the load-bearing support being the most dangerous with a risk level of 0.82. Further, a comprehensive scoring method is used to prioritize, with scores combining damage level, location importance and expansion trend. The repair urgency is calculated from the damage growth rate, with a growth rate of more than 5% being considered urgent. The foundation is divided into 25 areas by the repair level matrix, and the repair priority data is generated according to the damage level, location coefficient and load coefficient. The four areas around the support have a damage level of 0.75, a daily growth rate of more than 8%, and a load coefficient of 1.5, and are listed as the first repair area. The analytic hierarchy process improves the scientificity of the sorting by quantifying the indicators, and the repair level matrix provides a systematic evaluation framework for the area division.

[0137] In the embodiments of the present application, the data fusion and visualization capabilities of the cloud platform significantly improve the intuitiveness of the crack propagation path and the pertinence of the repair decision. The monitoring data shows the high risk of the load-bearing area, providing a key basis for timely intervention.

[0138] It can be understood that the present application does not strictly limit the grid size, the number of neural network layers or the weight distribution, which can be adjusted according to actual monitoring needs, for example, the grid can be further encrypted or the input features can be increased in high-risk areas to improve accuracy. Subsequent steps will be based on repair priority data to carry out verification and early warning.

[0139] As shown in Figures 1-2 , in step S108, the crack depth of the repair priority area is detected by ultrasonic waves and the adjusted main transformer foundation health status value is calculated. After comparing with the preset safety threshold, the corresponding safety warning signal is generated to provide real-time feedback for engineering safety management.

[0140] Further, in step S108, in the embodiment of the application, for the coordinate position of the repair priority area, a Gaussian modulated pulse probe wave is emitted by using an 8-channel ultrasonic detection array, the center frequency is 3 MHz, the pulse width is 1 μs, the probe spacing is 25 mm, the coverage length is 200 mm, the frequency and amplitude parameters are extracted from the standard waveform library, the signal is enhanced by a preamplification multiple of 40 dB, the 2.5-3.5 MHz frequency band is screened by using a band-pass filter to filter out concrete aggregate scattering noise, and the envelope curve is generated by Hilbert transform on the filtered signal to clearly reflect the echo amplitude distribution; the peak recognition algorithm is used to extract the reflected peak position with the noise root mean square value of 3 times as the threshold, the crack depth is calculated by time difference analysis, the actual depth is quantified by combining the sound wave propagation speed of 3600 m / s and the attenuation coefficient of 0.8 dB / cm, the crack depth distribution data is generated, the measurement error is controlled within 5 mm, and the time difference accuracy is better than 0.1 μs. The ultrasonic detection combines high-frequency pulse and precise filtering to ensure the reliability of deep crack detection.

[0141] S1081, for the crack depth distribution data, an elastoplastic mechanics calculator is used to construct a stress redistribution function, the concrete elastic modulus of 30 GPa, the compressive strength of 30 MPa and the tensile strength of 2.5 MPa are extracted from the material parameter library, the nonlinear characteristics of the material are considered, the structural bearing capacity is calculated by the stress balance equation, the foundation bearing state data is generated, the stress in the range of 50 mm around the crack is increased by 35%, and the local bearing capacity is decreased; then a depth feedback neural network is used to establish a state evaluation function, the input features include displacement field, strain field and crack parameters, a total of 15 items, the long short-term memory unit is used in the hidden layer to capture the time sequence dependence, the features are fused by the weighted average method, the weights are optimized by the BP algorithm, and the foundation health index is output. In a certain monitoring case, the health index decreases from 0.95 to 0.68. The elastoplasticity quantifies the weakening effect of the crack on the bearing capacity, and the neural network improves the comprehensiveness of the state evaluation through multi-dimensional feature analysis.

[0142] S1082, after obtaining the basic health index, the early warning level is divided by using a fuzzy evaluation matrix, the safety threshold is set to 0.75 based on Monte Carlo simulation, the yellow early warning threshold is 0.65, the orange early warning threshold is 0.55, the red early warning threshold is 0.45, if the index is lower than 0.75, the early warning is triggered, for example, when the index is 0.68, the yellow early warning is generated; the sound and light parameters are extracted from the signal feature library by the early warning information generator, the yellow early warning is sent once an hour, the sound frequency is 500Hz to 2kHz, the LED flicker frequency is low, the orange early warning is sent once every 30 minutes, the red early warning is sent once every 10 minutes, the frequency is higher, the basic safety early warning data containing the level identification is generated, 32 early warnings are recorded in the continuous monitoring, 25 times of yellow early warning, 5 times of orange early warning, 2 times of red early warning, and the accuracy reaches 95%. The fuzzy evaluation optimizes the threshold setting through the probability distribution, and the grading early warning mechanism ensures the timeliness and differentiation of the response.

[0143] In the embodiment of the application, the base is divided into 25 calculation units for bearing state evaluation, the crack area health index decreases significantly, and the generation of the early warning signal effectively prompts the repair effect and potential risks, thereby providing an intuitive basis for subsequent maintenance.

[0144] It can be understood that the application does not make fixed limitations on the number of probes, the filtering frequency band or the early warning frequency, and can be adjusted according to the characteristics of the repair area, for example, the filtering range can be appropriately widened in a high-noise environment to improve the signal quality.

[0145] For those skilled in the art, other various corresponding changes and deformations can be made according to the above-described technical solutions and concepts, and all these changes and deformations should belong to the protection scope of the claims of the application.

Claims

1. A cloud platform-based sedimentation data analysis and early warning method, characterized in that, The application relates to a method for monitoring and repairing cracks in a structure, and a system thereof. The method comprises the following steps: Collecting triaxial displacement data of an electronic settlement observation device array through a distributed sensor gateway, performing three-dimensional spatial coordinate calculation and calculating deformation stress data, and uploading the deformation stress data to a cloud platform; Based on the deformation stress data, a hexahedral grid element is established, the node strain energy density is calculated, and the stress concentration position of the foundation structure is determined through strain gradient analysis; An ultrasonic pulse signal is emitted to the stress concentration position to identify the crack initiation position, an acoustic emission sensor array is arranged in the crack area to capture the expanding acoustic wave signal, and a time sequence of the crack expansion direction and speed is constructed; A dynamic correlation model of the strain distribution matrix and the crack expansion parameters is established to determine the local strain aggravation trend, and the method further comprises the following steps: according to the crack expansion time sequence, a strain redistribution calculator is used to update the basic strain distribution data, strain measured values are extracted from a strain monitoring database, an expansion influence area boundary is determined through strain difference calculation, and a strain real-time distribution matrix is obtained; According to the strain real-time distribution matrix, a strain change amount around the crack expansion track is calculated by using a tensor mapping method, strain field reconstruction results are obtained from a finite element calculator, strain distribution update data are established by using a grid interpolation method, and a strain influence tensor is obtained; According to the strain influence tensor, a strain transfer function is established by using a recurrent neural network, the influence of the expansion parameters on the strain field is calculated from the strain transfer function, influence coefficients of each monitoring point position are determined through strain coupling calculation, and a strain transfer matrix is obtained; According to the strain transfer matrix, a strain change rate of a high stress area is extracted from a strain time sequence, a mutation point position is identified from a strain change rate curve, a strain aggravation area range is determined through local strain gradient calculation, and a local strain aggravation trend is obtained; The local strain aggravation trend is input into a structure mechanics simulation model to calculate a dynamic change value of the basic bearing capacity, when the dynamic change value is lower than a safety threshold value, a three-dimensional visual path of the crack is generated and a repair priority area is determined; 2. The method according to claim 1, wherein, An ultrasonic depth detection is performed on the repair priority area to calculate the structure bearing capacity and generate a graded warning signal. The method for collecting triaxial displacement data of an electronic settlement observation device array through a distributed sensor gateway, performing three-dimensional spatial coordinate calculation and calculating deformation stress data, and uploading the deformation stress data to a cloud platform comprises the following steps: 3.The settlement data analysis and early warning method based on the cloud platform according to claim 1, characterized in that, Displacement data are collected through a sensor gateway and are subjected to Kalman filtering and smoothing processing; least square method is used for coordinate calculation to generate normalized coordinate data; the coordinate data are uploaded to a cloud platform; stress and strain parameters are extracted and deformation stress data are calculated by using an elastic mechanics model; a deformation trend curve is constructed and a long short-term memory network is used for prediction; a preset threshold value is compared to trigger a warning and the data are encrypted and transmitted to the cloud platform. The method for establishing a hexahedral grid element based on deformation stress data, calculating node strain energy density, and determining the stress concentration position of the foundation structure through strain gradient analysis comprises the following steps: Three-dimensional deformation data is extracted from the cloud platform, adaptive mesh partitioning is performed, and initial strain values are calculated; finite element analysis and Runge-Kutta method are used to calculate strain energy density and generate a cloud chart in the abnormal strain area; strain gradient and principal strain value are calculated to determine the principal stress distribution; stress intensity is evaluated through Mohr's circle criterion and stress intensity formula, and nodes exceeding twice the average value are determined as stress concentration positions.

4. The method according to claim 1, wherein, The stress concentration position emits an ultrasonic pulse signal to identify the crack initiation position, and an acoustic emission sensor array is arranged in the crack area to capture the expanding acoustic wave signal, construct the time series of crack propagation direction and speed, including: An acoustic emission sensor array with a hexagonal topology is used to collect data, and an adaptive threshold filter is used to remove noise; Fourier transform, wavelet decomposition and triangular positioning algorithm are used to process acoustic event data to generate crack propagation trajectories; deep recurrent neural network and Hilbert transform are used to analyze crack propagation state parameters; cross-correlation function and support vector regression are used to calculate propagation direction and speed, and Kalman filter is used to smooth the data to obtain the crack propagation time series containing coordinates, speed and direction.

5. The method of claim 1, wherein the cloud-based sediment data analysis and early warning method is characterized by, The local strain aggravation trend is input into the structural mechanics simulation model to calculate the dynamic change value of the bearing capacity of the foundation, and when the dynamic change value is lower than the safety threshold, a three-dimensional visual path of the crack is generated and the repair priority area is determined, including: The safety threshold of the structure bearing capacity is extracted and a warning is triggered; the settlement displacement and acoustic emission data are obtained, and time-space alignment is performed through time series matching and coordinate transformation; deep convolutional neural network is used to extract crack spatial features and construct crack propagation mapping function; three-dimensional mesh division and iterative closest point algorithm are used to generate crack propagation stereo profile; regional damage distribution map is constructed and hazard level is sorted; comprehensive scoring method and repair level matrix are used to generate repair priority data. 6.The settlement data analysis and early warning method based on the cloud platform according to claim 1, characterized in that, The repair priority area is subjected to ultrasonic depth detection, and a graded warning signal is generated based on the calculation of the structure bearing capacity, including: An ultrasonic detection array is used to obtain crack depth distribution data; an elastoplasticity calculator is used to construct a stress redistribution function and calculate the structure bearing capacity; a depth feedback neural network is used to establish a state evaluation function and generate a foundation health index; a fuzzy evaluation matrix is used to divide the warning level and trigger the warning; a warning information generator is used to send a graded warning signal.

Images