Temperature monitoring and stress monitoring method and system for fan foundation

Abstract

The invention relates to the technical field of construction monitoring, in particular to a temperature monitoring and stress monitoring method and system for a fan foundation. The method comprises the following steps: collecting concrete hydration heat release monitoring signals, performing temperature gradient distribution excavation and finite element numerical simulation, and constructing a concrete pouring process temperature evolution model; concrete surface temperature monitoring parameters are obtained, multi-time-point surface layer temperature calculation is carried out, table environment temperature fluctuation mapping is carried out on the concrete pouring process temperature evolution model, and a real-time area temperature twinborn model is constructed; strain state sensing parameters in the concrete pouring process are identified, multi-scale spectral decomposition is carried out, then temperature strain signal stripping is carried out, and pure mechanical strain component characteristics are obtained; and performing internal micro-pore structure analysis based on pure mechanical strain component characteristics to generate a concrete situation prediction result. The intelligent level of the whole fan foundation construction stage is improved, and the fan construction efficiency and safety are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of construction monitoring, and in particular to a method and system for temperature monitoring and stress monitoring of a wind turbine foundation. BACKGROUND

[0002] During the foundation construction process of high-power wind turbine generators in high-cold regions, temperature control and stress monitoring of concrete face many challenges. Due to the large diurnal temperature difference and extremely low temperature in high-cold environments, concrete is prone to temperature cracks, frost damage, or insufficient early strength during pouring and curing. These adverse factors not only cause micro damage to the internal structure of the concrete, but also may cause deformation and cracking of the overall foundation, thereby affecting the long-term stable operation of the wind turbine and the safety of the project. In addition, due to the large volume and complex structure of wind power foundations, it is difficult to accurately grasp the stress state in low-temperature environments. Once the stress is not released uniformly or not monitored in time, it may cause foundation cracking, structural instability, and other major hidden dangers.

[0003] Currently, the monitoring of the concrete construction process mainly relies on traditional thermometers, data acquisition instruments, and periodic on-site inspections. Although these methods have certain feasibility under general construction conditions, they have significant defects such as response lag, limited monitoring points, and inability to record continuously in the construction environment of wind turbines in high-cold regions. It is difficult to meet the precise perception and dynamic control of temperature changes and stress evolution of concrete throughout its life cycle. At the same time, most existing methods rely on manual reading and experience-based judgment, lack intelligent analysis and early warning mechanisms, and are difficult to provide effective intervention decisions in time under extreme climate conditions.

[0004] With the continuous improvement of the power level of wind power equipment, the requirements for foundation safety performance are becoming increasingly stringent. In order to effectively cope with the complex construction environment and severe climate conditions in high-cold regions, there is an urgent need for a concrete temperature monitoring and stress monitoring method with high sensitivity, real-time performance, and adaptability to low-temperature environments. SUMMARY

[0005] To solve the above technical problems, a method and system for temperature monitoring and stress monitoring of a wind turbine foundation are provided to solve at least one of the above technical problems.

[0006] To achieve the above-mentioned purpose, the present application provides a method for temperature monitoring and stress monitoring of a wind turbine foundation, comprising the following steps:

[0007] Collecting concrete hydration heat release monitoring signals, performing temperature gradient distribution mining and finite element numerical simulation, and constructing a temperature evolution model of the concrete pouring process;

[0008] Obtain concrete surface temperature monitoring parameters, perform multi-time point surface layer temperature calculation, and map surface environment temperature fluctuations to the concrete pouring process temperature evolution model to construct a real-time regional temperature twin model.

[0009] Identify strain state perception parameters in the concrete pouring process, perform multi-scale frequency spectrum decomposition, and then perform temperature strain signal stripping to obtain pure mechanical strain component features.

[0010] Perform internal micro-pore structure analysis based on pure mechanical strain component features, and perform long-term performance deep situation prediction according to the real-time regional temperature twin model to generate concrete situation prediction results.

[0011] Perform strain abnormal cracking risk prediction and comprehensive construction risk level quantification on the concrete situation prediction results to generate a dynamic construction risk level curve.

[0012] Adjust construction time and risk avoidance decisions according to the dynamic construction risk level curve to perform intelligent wind turbine construction optimization and monitoring.

[0013] In the present specification, a wind turbine foundation temperature monitoring and stress monitoring system is provided for performing the wind turbine foundation temperature monitoring and stress monitoring method as described above, comprising:

[0014] A temperature gradient distribution module for collecting concrete hydration heat release monitoring signals, performing temperature gradient distribution mining and finite element numerical simulation, and constructing a concrete pouring process temperature evolution model.

[0015] A temperature fluctuation mapping module for obtaining concrete surface temperature monitoring parameters, performing multi-time point surface layer temperature calculation, and mapping surface environment temperature fluctuations to the concrete pouring process temperature evolution model to construct a real-time regional temperature twin model.

[0016] A strain perception module for identifying strain state perception parameters in the concrete pouring process, performing multi-scale frequency spectrum decomposition, and then performing temperature strain signal stripping to obtain pure mechanical strain component features.

[0017] A situation prediction module for performing internal micro-pore structure analysis based on pure mechanical strain component features, and performing long-term performance deep situation prediction according to the real-time regional temperature twin model to generate concrete situation prediction results.

[0018] A risk prediction module for performing strain abnormal cracking risk prediction and comprehensive construction risk level quantification on the concrete situation prediction results to generate a dynamic construction risk level curve.

[0019] A risk avoidance decision module for adjusting construction time and risk avoidance decisions according to the dynamic construction risk level curve to perform intelligent wind turbine construction optimization and monitoring.

[0020] The beneficial effects of the present application are as follows: accurately grasping the internal temperature variation law caused by heat release in the hydration reaction process of concrete, and extracting the temperature gradient characteristics at different spatial positions. Through numerical simulation of the heat conduction process by the finite element method, the time evolution trend of the temperature field can be effectively predicted, thereby providing basic data support for early crack control, helping to develop reasonable temperature control measures, and improving the safety and durability of the concrete structure during the construction phase. By monitoring the surface temperature of the concrete and external meteorological parameters in real time, a dynamic boundary condition model of the interaction between the concrete structure and the environment is established. The temperature fluctuations in the environment are mapped to the internal temperature evolution process, effectively improving the authenticity and responsiveness of the thermal field simulation. Finally, a regional temperature digital twin is formed, which can provide high-precision support for temperature regulation and stress assessment in a dynamic construction environment. The decomposition and denoising of the multi-source strain signals inside the concrete are realized, and the false strain signals caused by temperature changes are stripped out using multi-scale analysis methods, from which the mechanical strain components that truly reflect the stress state are extracted. This high-precision strain recognition technology can effectively enhance the physical effectiveness of structural response monitoring and provide key foundations for subsequent damage identification and performance evaluation.

[0021] By correlating the pure mechanical strain characteristics with the changes in the material microstructure, the internal deterioration behaviors of concrete during service, such as porosity and crack propagation, can be inverted. Combined with the real-time temperature twin model, the evolution prediction of concrete performance over time under thermal-mechanical coupling is realized, providing a scientific basis for the stability assessment and state prediction of the wind turbine foundation structure during the long-term service period. By comparing the key strain indicators in the situation prediction results with the structure safety threshold, the early identification of potential abnormal strain evolution and crack initiation risk can be achieved. Further, a construction risk level evaluation system is constructed to quantify the risk level of the current construction state and dynamically display the risk change trend in curve form, providing visual reference for on-site decision-making. Based on the dynamic information provided by the risk level curve, reasonable adjustments to the construction pace and optimization of maintenance and insulation measures can effectively avoid high-risk periods and sensitive working conditions. This decision-making mechanism enhances the active control capability of the construction process and improves the intelligent level of the entire wind turbine foundation construction phase, contributing to the realization of high-quality, low-risk, and high-efficiency construction goals. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 A step flowchart for a method for temperature monitoring and stress monitoring of a wind turbine foundation according to the present application;

[0023] Figure 2 A detailed implementation step flowchart for step S1;

[0024] Figure 3 A detailed implementation step flowchart for step S2;

[0025] Figure 4 The detailed implementation step flowchart of step S3 is shown. DETAILED DESCRIPTION

[0026] It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

[0027] The present application provides a method and system for monitoring the temperature and stress of a fan foundation. The execution subject of the method and system for monitoring the temperature and stress of a fan foundation includes but is not limited to mechanical equipment, a data processing platform, a cloud server node, a network upload device, etc., which can be regarded as a general computing node of the present application, and the data processing platform includes but is not limited to an audio image management system, an information management system, and a cloud data management system.

[0028] Referring to Figures 1 to 4 The present application provides a method for monitoring the temperature and stress of a fan foundation, which includes the following steps:

[0029] The concrete hydration heat release monitoring signal is collected, the temperature gradient distribution is excavated, and the finite element numerical simulation is performed to construct a temperature evolution model of the concrete pouring process.

[0030] The concrete surface temperature monitoring parameter is obtained, the multi-time point surface layer temperature is calculated, the temperature evolution model of the concrete pouring process is mapped to the surface environment temperature fluctuation, and a real-time regional temperature twin model is constructed.

[0031] The strain state perception parameter in the concrete pouring process is identified, multi-scale spectrum decomposition is performed, and then temperature strain signal stripping is performed to obtain a pure mechanical strain component feature.

[0032] Internal micro-pore structure analysis is performed based on the pure mechanical strain component feature, long-term performance deep trend prediction is performed according to the real-time regional temperature twin model, and a concrete trend prediction result is generated.

[0033] The strain abnormal cracking risk prediction and comprehensive construction risk grade quantification are performed on the concrete trend prediction result to generate a dynamic construction risk grade curve.

[0034] The construction time adjustment and risk avoidance decision are made according to the dynamic construction risk grade curve to perform intelligent fan construction optimization and monitoring.

[0035] Referring to Figure 1 The present application provides a step flowchart of a method for monitoring the temperature and stress of a fan foundation, and in the present example, the steps of the method for monitoring the temperature and stress of a fan foundation include:

[0036] The concrete hydration heat release monitoring signal is collected, the temperature gradient distribution is mined, and the finite element numerical simulation is performed to construct a temperature evolution model in the concrete pouring process;

[0037] In this embodiment, first, a distributed optical fiber temperature sensor network is deployed, a distributed optical fiber temperature sensing system based on Raman scattering principle is adopted, the spatial resolution is set to 1 meter, the temperature accuracy reaches ±0.1℃, and the sampling frequency is once every 10 minutes. Before pouring the concrete, a three-dimensional sensor layout scheme is designed according to the geometric size of the foundation, the optical fiber sensors are pre-buried in the foundation according to a grid spacing of 0.5m×0.5m×0.5m, and it is ensured that the three-dimensional distribution characteristics of the internal temperature field of the concrete can be captured. After the pouring of the concrete starts, the hydration heat release process is monitored in real time, the temperature change curve is recorded continuously through the data acquisition system, the peak time and peak temperature of the hydration heat release are identified. The gradient analysis is performed on the collected temperature data, the three-dimensional temperature gradient field is calculated by using the finite difference method, the high temperature gradient area and the temperature stress concentration position are identified. A finite element numerical model based on the heat conduction equation is established, the change law of the concrete thermal conductivity with temperature is considered, the boundary condition is set to be the environmental temperature convection heat exchange, and the internal heat source is the hydration heat release function. The finite element model is parameterized through the measured data, the thermal conductivity of the concrete is determined to be 2.5W / (m·K), the specific heat capacity is 1000J / (kg·K), and the density is 2400kg / m³. The temperature evolution prediction model in the concrete pouring process is constructed, the internal temperature distribution of the concrete in the future 48 hours can be predicted, and a scientific basis is provided for the construction control.

[0038] The concrete surface temperature monitoring parameters are acquired, the surface layer temperature at multiple time points is calculated, the real-time regional temperature twin model is constructed by mapping the surface environment temperature fluctuation to the concrete pouring process temperature evolution model;

[0039] In this embodiment, an array of infrared temperature sensors is deployed on the concrete surface, using a non-contact temperature measurement method. The temperature measurement range is -40°C to +80°C, with an accuracy of ±0.2°C and a response time of less than 1 second. The sensors are deployed according to a grid spacing of 2 meters by 2 meters, covering the entire concrete surface area and achieving comprehensive monitoring of the surface temperature. An environmental meteorological monitoring station is also deployed to monitor environmental temperature, wind speed, wind direction, relative humidity, solar radiation, and other parameters, with a sampling frequency of every 5 minutes. A heat exchange model between the concrete surface and the environment is established, considering convective heat transfer, radiative heat transfer, and evaporative heat transfer. The convective heat transfer coefficient is calculated using an empirical formula based on wind speed, and when the wind speed is 3 m / s, the convective heat transfer coefficient is approximately 15 W / (m²·K). Through multi-time point temperature data analysis, the daily variation of surface temperature and the influence mechanism of environmental temperature fluctuations on surface temperature are identified. The temperature evolution model established in step S1 is coupled with the surface temperature monitoring data to establish the mapping relationship between the internal temperature field and the surface temperature field. A digital twin temperature model is constructed to realize real-time synchronization between the physical concrete structure and the virtual temperature field. The digital twin model can update the concrete internal temperature distribution prediction in real time according to the environmental temperature change, with a prediction accuracy of ±0.5°C. This model provides accurate temperature field basic data for subsequent strain analysis and risk assessment.

[0040] Identify strain state perception parameters in the concrete pouring process, perform multi-scale spectral decomposition, and then perform temperature strain signal stripping to obtain pure mechanical strain component features;

[0041] In this embodiment, a fiber Bragg grating strain sensor network is deployed, with a measurement range of ±3000με, an accuracy of ±1με, and a temperature compensation range of -40°C to +80°C. During the concrete pouring process, strain sensors are deployed on key sections according to the main stress direction, including horizontal, vertical, and 45° oblique directions, forming a three-dimensional strain monitoring network. The strain sensor sampling frequency is set to once per minute, continuously monitoring the transition process of concrete from a plastic state to an elastic state. The collected strain signals are subjected to multi-scale spectral decomposition, and the wavelet transform method is used to decompose the signals into high-frequency, medium-frequency, and low-frequency bands. The high-frequency component reflects short-term strain fluctuations, the medium-frequency component reflects hydration shrinkage effects, and the low-frequency component reflects long-term deformation trends. Based on the temperature field data of step S2, the thermal strain component of the concrete is calculated, and the linear superposition principle is used to decompose the total strain into thermal strain and mechanical strain. The linear expansion coefficient of concrete is taken as 1.0×10⁻ 5 / ℃, and the thermal strain ε t= aAT. Subtract the thermal strain component from the total strain to obtain the pure mechanical strain component, which reflects the real deformation of concrete caused by external force and constraint. Establish a strain component identification algorithm to extract the characteristic parameters of mechanical strain, including strain peak value, strain rate, strain direction, and other key indicators through signal filtering and data fusion technology.

[0042] Based on the characteristics of the pure mechanical strain component, internal micro-pore structure analysis is performed, and long-term performance deep trend prediction is performed according to the real-time regional temperature twin model to generate concrete situation prediction results;

[0043] In this embodiment, the correlation model between strain and micro-pore structure is established based on the concrete meso-mechanics theory. The internal porosity change of concrete is inverted through strain monitoring data. When the strain exceeds 200με, it indicates that the concrete begins to appear micro-damage, and the porosity increases by about 0.1%. A concrete performance degradation model based on damage mechanics is established, considering the influence of porosity, crack density, water-cement ratio and other factors on the long-term performance of concrete. Combined with the temperature twin model of step S2, the influence mechanism of temperature cycle on the microstructure of concrete is analyzed. In the cold region, the daily temperature difference can reach 20℃, and the number of temperature cycles is calculated according to the construction period of 90 days, and about 900 temperature cycles are expected. A temperature-strain-damage coupling analysis model is established, and the finite element method is used to solve the stress-strain state evolution of concrete. Through machine learning algorithm training, the input parameters include current strain state, temperature history, environmental conditions, etc., and the output parameters are concrete strength development curve, elastic modulus evolution curve, damage development trend, etc. The prediction accuracy of the model is verified by historical data, and the strength prediction error is controlled within 5%, and the elastic modulus prediction error is controlled within 8%. The comprehensive situation prediction results of concrete are generated, including performance prediction values of 7 days, 28 days and 90 days, which provide scientific basis for construction quality control and long-term performance evaluation.

[0044] The strain abnormal cracking risk prediction and comprehensive construction risk grade quantification are performed on the concrete situation prediction results to generate a dynamic construction risk grade curve.

[0045] In this embodiment, a concrete cracking risk assessment model is established based on the situational prediction results. The strain anomaly threshold is set to 500με, and when the predicted strain exceeds this threshold, the cracking risk significantly increases. A multi-factor risk assessment matrix is established, considering the comprehensive influence of temperature gradient, strain rate, environmental temperature, wind speed and other factors. When the temperature gradient exceeds 15℃ / m, the cracking risk level increases by one level; when the strain rate exceeds 10με / h, the cracking risk level increases by one level; when the environmental temperature is lower than -10℃, the freezing damage risk level increases by one level. A risk level quantification scoring system is constructed, divided into four levels: low risk (0-3 points), medium risk (4-6 points), high risk (7-9 points), and extremely high risk (more than 10 points). Each risk factor is assigned different weights according to its influence degree, with temperature gradient weight of 0.3, strain state weight of 0.25, environmental condition weight of 0.2, material performance weight of 0.15, and construction technology weight of 0.1. A time-varying model of risk level is established, and the risk assessment results are updated every 2 hours, generating a dynamic construction risk level curve. This curve can intuitively show the trend of risk level change during construction, identify the time and position of risk peak value. When the risk level reaches medium risk or above, the system automatically triggers the early warning mechanism, and notifies the construction management personnel through SMS, email, sound and light alarm and other ways. The corresponding relationship between risk level and construction measures is established, and corresponding response strategies and disposal measures are formulated for different risk levels.

[0046] Construction time adjustment and risk avoidance decisions are made according to the dynamic construction risk level curve to execute intelligent fan construction optimization and monitoring.

[0047] In this embodiment, when the risk level is low, the original construction plan is followed; when the risk level rises to medium, the construction time is adjusted to avoid high-risk periods, such as choosing to pour in the afternoon when the temperature is higher; when the risk level reaches high, the construction work is suspended, and risk response measures are started, including strengthening insulation maintenance, adjusting the concrete mix ratio, and increasing the maintenance time. A dynamic adjustment algorithm for construction parameters is established to automatically adjust key parameters such as pouring speed, layer thickness, and curing temperature based on real-time monitoring data and risk assessment results. The pouring speed is adjusted from the normal 30m³ / h to 20m³ / h, the layer thickness is adjusted from 2m to 1.5m, and the curing temperature is increased from 5℃ to 10℃. An intelligent curing control system is built, including electric heating film heating, steam curing, and insulation covering equipment, which automatically adjusts the curing intensity according to the risk level. A mobile monitoring APP is developed to provide real-time risk information, construction guidance, and emergency disposal suggestions for on-site construction personnel. A construction quality traceability system is established to record the complete process of each risk warning, decision adjustment, and measure implementation, providing experience data for subsequent project optimization. Through intelligent monitoring and decision-making system, precise control of wind turbine foundation construction is realized, ensuring that the concrete construction quality meets the design requirements, and laying a solid foundation for the long-term safe operation of the wind turbine.

[0048] In this embodiment, referring to Figure 2 For the collection of concrete hydration heat release monitoring signals, temperature gradient distribution mining and finite element numerical simulation are performed to construct a detailed implementation step flowchart of the concrete pouring process temperature evolution model. In this embodiment, the detailed implementation steps of collecting concrete hydration heat release monitoring signals, performing temperature gradient distribution mining and finite element numerical simulation, and constructing a concrete pouring process temperature evolution model include:

[0049] Deploy distributed optical fiber temperature sensors to monitor the concrete pouring process of the wind turbine construction area in real time, and collect concrete hydration heat release monitoring signals;

[0050] Continuous spatial temperature sampling is performed on the concrete hydration heat release monitoring signals to extract and generate spatiotemporal hydration heat release characteristic parameters;

[0051] The spatiotemporal hydration heat release characteristic parameters are subjected to hydration heat peak value occurrence time calculation to extract heat peak value timestamps;

[0052] Temperature gradient distribution mining is performed on multiple spatiotemporal hydration heat release monitoring parameters to generate temperature gradient distribution rules;

[0053] Based on the heat peak value timestamps and temperature gradient distribution rules, finite element numerical simulation is performed to construct a concrete pouring process temperature evolution model.

[0054] In this example, before the concrete pouring construction begins, to realize real-time monitoring of the whole process, it is necessary to reasonably arrange distributed optical fiber temperature sensors (DTS) in the pouring area. The DTS system is based on the principle of Raman scattering and can realize continuous spatial resolution temperature collection along the length direction of the optical fiber. Its accuracy can generally reach ±0.1℃, and the spatial resolution can be controlled between 0.5-1 meter, which is suitable for internal temperature monitoring of mass concrete. In the actual layout process, the optical fiber layout path should be designed according to the geometric size of the fan foundation (such as diameter 20m, thickness 5m). Common methods include "serpentine" layout, "spiral" embedding or "layered distribution". In order to improve the representativeness of the temperature field perception, it is recommended to set multiple layers of optical fibers at different elevations and different radius positions, with a spacing of 0.5-1.0m. At the same time, the optical fiber fixed points are pre-buried before pouring to avoid abnormal signals caused by construction disturbance. The sensing system should be connected to the data acquisition unit and transmitted to the remote monitoring center in real time, with one sample per minute to ensure the integrity of the temperature evolution over time. Using the optical fiber along-line temperature data collected by the DTS system, systematic processing is carried out through continuous spatial sampling method. Concrete will produce significant temperature rise during hydration reaction, especially within 24-72 hours after pouring, showing obvious heat peak value, so the data processing needs to have good time resolution. Adopting time-space two-dimensional matrix form to organize temperature data, each sampling point (such as every 0.5m interval) forms a temperature sequence changing with time. In order to reduce the noise interference of the original data, sliding average filtering (window width 35 points) or denoising method based on wavelet transform can be used for data smoothing processing. Then, key parameters such as hydration heat rising rate (℃ / h), maximum temperature value (T_max), temperature peak occurrence time (t_peak) and temperature sustained rising time (Δt_rise) are extracted at each spatial node. Through cluster analysis, spatial areas with similar hydration heat characteristics are classified, and representative heat response patterns are extracted.

[0055] After the concrete is poured, the hydration reaction rate is closely related to the properties of raw materials, environmental temperature, and structural geometry. The time of the thermal peak in different regions also has significant differences. Therefore, the temperature time series collected need to be analyzed in detail to extract the thermal peak timestamp of each spatial point. The specific method can be based on the first derivative detection method to identify the inflection point of the temperature curve slope, that is, the local maximum point; or the empirical mode decomposition (EMD) technique can be used to extract the dominant temperature change trend, thereby improving the robustness of peak identification. Set the temperature change threshold (such as ≥2℃ / h) to filter noise false peaks, and record the peak time t_peak(x,y,z) of each sampling point in the spatial distribution. It is found during the experiment that in an environment with an air temperature of -15℃, the thermal peak in the external region generally occurs after 48 hours, while the core region can be delayed to 72-96 hours. Drawing the thermal peak time into a spatial isochrone map can intuitively reflect the propagation path and intensity of the hydration reaction, providing important clues for the internal thermal field evolution of the structure and a time window for thermal crack risk identification. The temperature gradient in the concrete is an important factor that causes early cracks, stress concentration, and reduced durability. Therefore, after obtaining comprehensive temperature data, a three-dimensional spatial analysis of the temperature gradient must be performed. The finite difference method is used to calculate the temperature difference between adjacent points in space, and the gradient where r is the spatial coordinate in any direction. For different time sections (such as 12h, 24h, 48h, 72h) during construction, the gradient evolution analysis is performed to extract the peak gradient time, maximum gradient position, and gradient change rate. The experiment shows that the maximum temperature gradient between the core and edge regions of the concrete can reach 7.5℃ / m, which is more significant when the boundary heat transfer coefficient is low. Further, the high gradient region can be delineated as a potential thermal stress concentration area through cluster analysis, and a dynamic risk level distribution map can be established based on the time evolution trend. This step is of great significance in structural health monitoring, as it not only reveals the non-uniform evolution of the temperature field but also provides a basis for subsequent stress-strain inversion and anti-cracking control strategies.

[0056] To comprehensively understand and predict the temperature variation of concrete under different construction and environmental conditions, a finite element temperature field simulation model based on measured data is needed. The modeling uses simulation software such as ANSYS or COMSOL, which utilizes the heat conduction control equation By combining the peak heat time stamps and temperature gradient patterns obtained from the experiment, the heat source term Q and the boundary heat transfer coefficient h were calibrated. The structural geometry adopted the construction drawing model, with refined meshing (unit size approximately 0.25m), and multi-region material properties (such as cement content, ice content, aggregate thermal conductivity, etc.) were defined. Boundary conditions were set based on meteorological monitoring data, including diurnal temperature variations (e.g., -10℃ during the day, -20℃ at night) and a radiative heat transfer model. The time step was set to 10 minutes to ensure model response accuracy. The model accuracy was verified by comparison with measured DTS data. After parameter correction, it can be used to predict temperature change responses under different pouring methods and insulation measures. The model output includes temperature isosurface maps, heat flux vector maps, and temperature time history curves for each node, providing scientific support for subsequent thermal stress analysis and construction process optimization.

[0057] In this embodiment, the specific steps for continuously sampling the spatial temperature of the concrete hydration heat release monitoring signal and extracting spatiotemporal hydration heat release characteristic parameters are as follows:

[0058] Calculate the spatial location of the distributed fiber optic temperature sensor to obtain the distribution coordinates of each sensor node;

[0059] Identify the acquisition frequency of the distributed fiber optic temperature sensor;

[0060] The frequency difference between multiple devices is calculated for the acquisition frequency to generate the acquisition frequency difference between different sensors;

[0061] Based on the difference in acquisition frequency, time alignment processing is performed to obtain standardized acquisition timestamps and frequencies;

[0062] Abnormal noise is identified in the monitoring signal of heat release from concrete hydration, and abnormal noise points are marked.

[0063] Calculate the waveform and abrupt change range of the abnormal noise points;

[0064] Based on the waveform and the range of abrupt changes, adaptive filtering optimization is performed to construct an optimized signal for heat of hydration release.

[0065] Based on the distribution location coordinates, the optimized signal of heat of hydration release is continuously sampled spatially, and the time stamp is uniformly calibrated based on the standardized acquisition timestamp and frequency to generate spatiotemporal heat of hydration release characteristic parameters.

[0066] In this embodiment, to accurately reflect the spatial temperature field in actual engineering, it is necessary to precisely locate each node in the distributed fiber optic temperature sensing system. Typically, the DTS system provides a "distance scale" along the fiber's length, rather than absolute three-dimensional spatial coordinates. Therefore, it is necessary to combine the construction layout diagram and measurement data to convert the fiber path into (x, y, z) coordinates in the structural geometric coordinate system. Specifically, during the deployment process, a laser rangefinder or total station is used to record the precise location of the fiber's turning points. For example, the bottom layer uses a ring layout with a radius of 5m and a spacing of 0.5m; the upper layer uses a spiral layout, with a height increase of 0.2m per revolution. Using interpolation algorithms (such as spline interpolation or linear piecewise interpolation), the fiber path is refined into sampling points every 0.25~0.5 meters, corresponding to the spatial sampling interval of the DTS system, thus obtaining a complete set of spatial coordinates for the sensing nodes. This step provides a crucial geometric foundation for multi-sensor data spatial fusion and subsequent three-dimensional thermal modeling, and also helps identify the thermal response behavior of key parts of the structure (such as corners, edges, and core areas). Because multiple DTS subsystems may be used in field deployments (such as zone management or multiple fiber optic lines in parallel), the sampling frequencies of each system may not be completely consistent, potentially differing by minutes. Identifying the sampling frequency is a prerequisite for ensuring data timing consistency. In practice, by reading the timestamp information of the original temperature time series, calculating the time difference Δt between two consecutive sampling points, and statistically analyzing the average and standard deviation of all Δt over one hour, the actual sampling period of the system can be determined. For example, a DTS-A system samples once per minute, while a DTS-B system samples once every 90 seconds, with an error of approximately ±5 seconds. To improve the accuracy of sampling frequency identification, a sliding window frequency analysis method can be used, setting the window width to 10 minutes to dynamically detect the sampling period and identify whether frequency drift or data interruption exists. Furthermore, it is necessary to identify whether there are missing samples, i.e., multiple consecutive missing sampling times, which are removed by setting a maximum allowable interval (e.g., 5 minutes). This step ensures the accuracy of frequency matching and data alignment between multiple devices and is the basis for subsequent standardization processing. After completing the sampling frequency identification, it is necessary to systematically compare the frequency differences between each DTS channel or device. In actual engineering testing, three DTS devices in a certain project had sampling periods of 60s, 90s, and 120s, with a maximum frequency difference of 0.008Hz. Further analysis of the impact of these differences on data alignment, using simulation interpolation reconstruction testing, revealed that differences exceeding 0.003Hz would result in a 24-hour data alignment error exceeding 5 minutes, requiring correction. Therefore, these differences were compiled into a frequency difference matrix and combined with timestamp records to identify whether interpolation compensation or time reconstruction was needed. This process provides a decision-making basis for time-series data consistency, especially suitable for data fusion tasks in high-density sensing environments, helping to eliminate the error accumulation caused by sampling differences between multi-source systems.

[0067] Temporal alignment is a core step in ensuring that data from multiple DTS systems can be compared, analyzed, and modeled on the same time axis. To address the aforementioned frequency differences and sampling drift issues, linear interpolation is used to resample the low-frequency system data, aligning it uniformly to the time axis of the high-frequency system (e.g., using a standard sampling period of 60 seconds). For sequences with actual sampling time t_i, a target time axis T_standard = {t_0, t_0+60s, t_0+2×60s, ...} is constructed, and numerical estimation interpolation is performed on non-overlapping original data points. To improve accuracy, cubic spline interpolation or local polynomial fitting methods can be used to achieve better smoothness and continuity in the interpolation results. Experimental verification shows that time alignment can control the error of multi-channel data within ±10s, significantly improving spatiotemporal synchronization accuracy. This step is crucial for building a unified hydration heat characteristic analysis platform, ensuring the comparability and fusion of data from different channels at the same time point, and further improving the consistency and reliability of model input. In complex field environments, the data collected by the DTS system is inevitably affected by external disturbances (such as electromagnetic interference, construction vibration, fiber optic damage, etc.), resulting in anomalous signals such as abrupt changes, packet loss, and outliers. To eliminate such interference, an automated anomaly identification mechanism needs to be established. First, statistical analysis is performed on the temperature time series, calculating the mean μ and standard deviation σ under a sliding window, for each data point... calculate If the Z-score exceeds a set threshold (e.g., ±3), it is marked as an outlier. Secondly, by comparing the temperature difference between neighboring sampling points, if the temperature difference exceeds a physical set threshold (e.g., 5℃ / min), it is judged as abrupt noise. Experiments revealed that in a -20℃ environment, the instantaneous temperature jumps of some fiber optic nodes exceeded 15℃, significantly deviating from the physical characteristics of hydration heat release, and therefore had to be removed.

[0068] After identifying anomalies, further analysis of their waveform characteristics is needed to distinguish between random noise, periodic disturbances, and systematic distortions. By extracting data segments within a certain range (e.g., ±5 minutes) before and after the anomaly and performing local curve fitting, the waveform structure can be determined. If the waveform exhibits a single-point abrupt change (i.e., the abrupt change amplitude is greater than 10°C and there is no obvious rebound), it is identified as a signal breakage type; if it exhibits periodic disturbances (e.g., a jump every 10 minutes), it may be related to equipment sampling jitter. Furthermore, Fourier transform (FFT) analysis of the spectral characteristics of the anomaly segments can further quantify their amplitude range and frequency characteristics. For example, an abnormal enhancement at a node in the 0.1~0.3Hz frequency band indicates that it may be affected by low-frequency mechanical vibration. Statistical analysis of the abrupt change amplitude range shows that the vast majority of anomalies are within ±7°C, but some can reach 20°C, requiring classified processing. This step helps in developing differentiated filtering strategies to ensure that the physical authenticity of the hydration heat signal is not misfiltered.

[0069] After identifying and classifying anomalous noise points, adaptive filtering is required on the original monitoring signals to ensure the concrete hydration heat release data has engineering analysis value. Adaptive filtering emphasizes dynamically selecting the most suitable filtering strategy when different types of noise occur, thereby removing interference while preserving the original hydration heat change characteristics to the maximum extent. In practice, for low-amplitude, high-frequency fluctuations (such as temperature curves with continuous small ups and downs), smoothing filtering methods, such as moving average filtering or weighted filtering based on historical values, are used to effectively remove sharp fluctuations without losing trend information. For abrupt noise (such as a sudden temperature rise or fall of more than 10°C at a certain point), median filtering or edge-preserving filters are preferred. These methods can handle single-point anomalies without affecting the overall data continuity. In cold environments, due to drastic diurnal temperature variations (such as diurnal temperature variations exceeding 15°C), natural changes may also manifest as rapid heating or cooling. Therefore, the filtering algorithm must have environmental adaptability. In practical applications, a dynamic threshold adjustment mechanism can be introduced, that is, adjusting the filtering sensitivity according to the background temperature fluctuation amplitude within the monitoring period. For example, when the system detects a large overall temperature fluctuation (e.g., >10℃) within 24 hours, it automatically reduces the filtering intensity to avoid mistakenly filtering out the true hydration heat signal. After obtaining high-quality, filtered and optimized temperature monitoring data, it is necessary to combine the spatial coordinate information of each sampling point to reconstruct the continuous spatial temperature field and perform unified time series processing on the entire monitoring area. Specifically, based on the fiber optic cable layout path in the concrete, temperature values ​​are extracted according to a predetermined spatial sampling interval (e.g., every 0.5 meters), and the data of each sensing node is mapped to its three-dimensional coordinates (x, y, z) to construct a complete spatial temperature grid. In the time dimension, based on the standardized time axis unified in the previous processing steps (e.g., one time point every 60 seconds), all sensor data are strictly timestamped and synchronized. This means that regardless of which device or sampling frequency the original data comes from, it must ultimately be mapped to a unified sampling rhythm, ensuring that all nodes have the corresponding temperature value at the same time point. If data is missing or there is no corresponding sampling point at certain times, spatial neighborhood interpolation or temporal interpolation techniques can be used to complete it. Common methods include linear interpolation and spline interpolation. After the above processing, multidimensional feature parameters can be extracted from the integrated spatiotemporal data, including: the duration of temperature rise at each node, the peak value of hydration heat temperature, the rate of temperature rise, the direction of heat diffusion, and the type of temperature change trend. These features not only reflect the spatial non-uniformity of concrete hydration reaction but also serve as a basis for assessing the risk of thermal cracking, judging the effectiveness of curing measures, and predicting structural stress distribution. Experiments show that in extremely cold environments (such as -20℃), the rate of temperature rise in the core area of ​​concrete is significantly lower than that in the peripheral area, with a maximum temperature difference of up to 18℃. The standardized data clearly reveals this thermal field development pattern.The resulting spatiotemporal hydration heat characteristic parameter set provides solid data support for the simulation modeling of concrete thermal processes and the adjustment of on-site construction strategies.

[0070] In this embodiment, see Figure 3 The specific steps for acquiring concrete surface temperature monitoring parameters, calculating surface temperature at multiple time points, mapping surface temperature fluctuations to the concrete pouring process temperature evolution model, and constructing a real-time regional temperature twin model are as follows:

[0071] Continuously collect concrete surface temperature monitoring parameters in the wind turbine construction area;

[0072] The surface temperature of concrete was calculated at multiple time points based on the monitoring parameters, and the temperature fluctuation trend was analyzed to generate a multi-time point surface temperature trend curve.

[0073] Based on the concrete surface temperature monitoring parameters, the monitoring location is determined, and the spatial temperature distribution is analyzed to obtain a surface spatial temperature distribution map.

[0074] Based on the surface temperature trend curve at multiple time points, the surface spatial temperature distribution map is rendered to show temperature trend fluctuations, and a surface temperature distribution field of the slow-concrete concrete is constructed.

[0075] Based on the surface temperature distribution field of slow-setting soil, the temperature evolution model of the concrete pouring process is mapped to the surface temperature fluctuation, and a real-time regional temperature twin model is constructed.

[0076] In this embodiment, during the concrete pouring and curing process of wind turbine construction, the concrete surface is the main area in direct contact with the external environment and is highly susceptible to external factors such as temperature, wind speed, and radiation. Therefore, deploying a dedicated surface temperature acquisition system is crucial. Generally, infrared temperature measurement modules, wireless thermistor sensors (such as the PT1000 type), or integrated thermocouple arrays are used for point installation. The spacing depends on the structural size and accuracy requirements, typically 1-2 meters apart, distributed along the edge and top of the wind turbine foundation. In extremely cold environments (such as below -25℃), the surface temperature sensor must be freeze-resistant and have real-time data transmission capabilities. Using a low-temperature protective housing and a remote data transmission module (such as LoRa or NB-IoT) can effectively ensure data continuity. A data acquisition frequency of 5-10 minutes is recommended to cover the rapid diurnal temperature variation. The data is transmitted to a backend platform for unified storage via a gateway and correlated with other temperature monitoring data (such as internal DTS monitoring) in a time-series manner. This step focuses on achieving multi-point, multi-region, all-weather surface thermal environment sampling, providing a continuous data foundation for identifying changes in heat exchange intensity and mapping the external temperature field. After collecting surface temperature data, it is necessary to perform continuous time-series processing and extract temperature fluctuation trend characteristics. First, the historical temperature data of each sensor node is organized into a time series according to the sampling time. Then, temperature calculations are performed for fixed time periods by setting multiple analysis time points (such as 6:00 AM, 12:00 PM, 6:00 PM, and 0:00 AM the next day). For each node, the temperature trend at multiple key moments throughout the day is plotted, forming a "multi-time point temperature evolution curve." In data analysis, a sliding window method can be used to identify short-term temperature rise or fall trends, for example, calculating the temperature change rate with a 3-hour window. Simultaneously, the highest and lowest temperatures within the daily cycle and their occurrence times are identified to capture the lag in climate response to surface temperature. Experiments showed that under conditions of high wind speed (>8 m / s) and low radiation intensity (overcast and snowy weather), the surface temperature fluctuation range narrows, remaining only within ±2℃; while under clear and windless weather, the daytime temperature rise can reach over 10℃. The resulting trend curve not only reflects the rate of temperature change, but also reveals the correlation between the sensitivity of surface thermal response and external climate change, providing direction for further analysis of spatial thermal distribution.

[0077] Based on the analysis of time trends, it is further necessary to understand the temperature distribution differences at different locations at the same time. Therefore, it is necessary to correlate the spatial locations of each surface sensor with its real-time temperature values ​​to construct a two-dimensional temperature distribution map of the outer surface of the structure. This process first requires determining the geometric coordinates of the sensors in the structural plane, which can usually be achieved through laser ranging or by pre-setting the installation positions on drawings, such as a circular arrangement or a staggered strip layout. Based on the sensor coordinates and the corresponding temperature values ​​at the time, spatial interpolation algorithms (such as inverse distance weighted interpolation, Kriging interpolation, or spline interpolation) are used to fill in the temperature values ​​in the unmarked areas, thereby forming a continuous temperature distribution map. In the experiment, eight sensor points were placed at the edge of the wind turbine base. The results showed that the temperature on the sunlit side was 6-8°C higher than that on the shaded side, reaching its peak at noon. This spatial distribution map can be used to identify local hot spots, edge cold zones, and areas of uneven heating, providing a key reference for assessing the thermal stress distribution of the concrete surface and adjusting antifreeze measures. By integrating the time trend curve and the spatial temperature distribution map, a dynamic temperature evolution field of the concrete surface, i.e., the surface temperature distribution field, can be constructed. The specific operation involves using spatial temperature distribution maps at multiple key time points as a sequence of frames. Based on the heating or cooling trends at each point, color gradients and rendering effects are applied to create a dynamic visualization scene. This not only displays the current temperature status but also visually presents the path and direction of temperature change over time. Temperature trend fluctuation rendering typically employs dynamic isotherm changes, hotspot migration trajectory markings, and frame-by-frame animation of color heatmaps. A 24-hour continuous rendering experiment on a wind turbine pile foundation platform showed that changes in solar radiation intensity significantly affect the surface temperature distribution pattern. Overall temperatures are low in the early morning, with a significant temperature rise on the south side in the afternoon, and a general temperature drop at night, although the windward area maintains a relatively high residual temperature. This surface temperature distribution field not only helps construction personnel identify local thermal anomaly areas but also allows for comparative analysis with the internal temperature field, thereby identifying the direction of heat flux and potential surface-interior temperature difference risk areas.

[0078] A temperature twin model is a digital mirror reflecting the interaction between a construction structure and its ambient temperature. By coupling the surface temperature distribution field with an internal temperature model, it achieves full-field perception and dynamic prediction of the concrete temperature state. In this step, surface temperature distribution data is input as boundary conditions into the concrete temperature evolution model to simulate the impact of external temperature fluctuations on the internal temperature conduction and diffusion of the structure. Through a time-dynamic mapping mechanism, the temperature change of each surface grid node is synchronously input into the finite element simulation model or the thermal conduction analytical model, driving the model to respond in real time. By constructing a data-physics coupling mechanism, this temperature twin model can automatically correct simulation prediction errors while continuously acquiring new sensor data, achieving model self-learning and adaptive evolution. In engineering tests, this method has been applied to high-altitude and cold construction sites, maintaining a model prediction error within ±1.5℃ even under sudden temperature changes (such as a drop of -30℃), significantly outperforming traditional static models. The final temperature twin model can be used to display the temperature status of each area in the construction area in real time, predict future temperature change trends, and provide intelligent decision support for temperature control measures (such as starting and stopping heating blankets and adjusting insulation covers), which greatly improves the thermal control management level and construction safety of wind turbine foundation concrete construction in cold regions.

[0079] In this embodiment, reference Figure 4 The specific steps for identifying strain state sensing parameters during concrete pouring, performing multi-scale spectral decomposition, and then stripping temperature strain signals to obtain pure mechanical strain component features are as follows:

[0080] A fiber Bragg grating strain sensor array is simultaneously implanted during the concrete pouring process to extract the strain state sensing parameters of the concrete in real time.

[0081] The strain state sensing parameters are subjected to multi-scale spectral decomposition to obtain signal components in different frequency bands;

[0082] The low-frequency components of the signal are identified, and the amplitude and frequency are calculated to obtain the low-frequency temperature strain characteristics.

[0083] The signal components are subjected to high-frequency signal detection, and transient impact and resonance feature analysis is performed to extract high-frequency mechanical strain features.

[0084] A temperature-mechanical strain coupling model is constructed based on the low-frequency temperature strain characteristics and the high-frequency mechanical strain characteristics.

[0085] Temperature strain signal stripping was performed on the temperature-mechanical strain coupling model to obtain the pure mechanical strain component characteristics.

[0086] In this embodiment, during the initial stage of concrete pouring, when the concrete is still in a plastic state, the pre-deployed FBG fiber optic strain sensor array is synchronously embedded into key parts of the structure, which is a prerequisite for achieving high-precision strain sensing. Sensor deployment is typically carried out around different areas of the construction structure, such as the base center, edges, corners, and longitudinal and transverse shear zones, to form a strain monitoring network covering the entire structure. FBG sensors respond to microscopic deformations of the structure through minute changes in reflected wavelength, offering advantages such as resistance to electromagnetic interference, high sensitivity, and long-term embedding capability. In actual high-altitude and cold construction environments (below -25℃), low-temperature stable FBG strain gauges must be selected. These are typically encapsulated in metal sheaths and fixed to the inner wall of the formwork or the surface of the reinforcing steel frame using precise prestressed anchoring to ensure stable transmission of microscopic strain information even after the concrete has hardened. The sensor sampling frequency is generally set between 1 and 10 Hz to balance the monitoring requirements of temperature strain and mechanical dynamic response. The acquisition system uses a spectral analyzer to acquire changes in reflected wavelength in real time and converts them into strain values, forming a set of original strain state sensing parameters. This dataset covers the entire process of concrete from initial setting to final hardening, providing a foundation of first-hand information for analyzing hydration expansion, shrinkage settlement, temperature deformation, and external load response. The strain signal generated during concrete hardening exhibits significant multi-scale characteristics, including both slowly evolving temperature expansion strain and potentially high-frequency responses from short-term impacts, disturbances, and vibrations. Therefore, multi-scale spectral decomposition of the original strain time series is necessary to distinguish the physical meaning of different frequency components. Commonly used methods include wavelet transform, empirical mode decomposition (EMD), or variational mode decomposition (VMD). In practice, a multi-scale wavelet decomposition method with excellent time-frequency localization characteristics (such as the Daubechies wavelet) is selected to perform multi-level decomposition of the strain signal, breaking down the original signal into several sub-signal components, each corresponding to a different frequency range. Taking the strain data inside a wind turbine foundation within a single day as an example, after five layers of wavelet decomposition, layers 12-1 are identified as high-frequency disturbance signals (above 10Hz), layers 3-4 as mid-to-low frequency signals (0.1~1Hz), and layer 5 as low-frequency temperature trend components (<0.01Hz). This frequency band classification preliminarily enables the identification of the physical behavior of the strain signals, laying the foundation for subsequent extraction of specific temperature and mechanical strain characteristics.

[0087] Low-frequency signals typically correspond to thermally induced strain caused by changes in concrete temperature. To extract this feature, a detailed analysis of the low-frequency components is necessary, including their overall trend, rate of change, and amplitude range. First, baseline fitting is performed on the low-frequency signals to identify their slow rise or fall over time; then, the maximum strain amplitude and occurrence time are statistically analyzed. Combined with internal temperature data from the same period, the temperature-strain response coefficient can be further calculated. Under external conditions of -20℃, the peak low-frequency strain in the central region of the concrete occurs 36 hours after pouring, with a maximum strain of 150 με, and the trend is completely consistent with the hydration heat release process. This type of strain change is not accompanied by high-frequency oscillations and has a relatively long duration (usually greater than several hours), clearly attributable to temperature-induced expansion or contraction. By comparing the low-frequency strain responses of different locations laterally, the degree of influence of the thermal gradient on the strain distribution of the structure can be further determined, providing key input parameters for crack prevention design. High-frequency signals are mostly mechanical strain changes caused by short-term impacts, micro-vibrations, resonances, etc., characterized by transient time-domain changes, high amplitude but short duration. Such signals are particularly important during the concrete hardening process, potentially reflecting issues such as formwork displacement, microcracks in the structure, improper curing measures, or external construction interference. In the analysis, an event-triggered window approach is used to detect high-frequency signal abrupt changes, identifying segments with abrupt changes exceeding a threshold (e.g., 30 με) and an extremely rapid rate of change (e.g., above 100 με / s). Subsequently, short-time Fourier transform or Hilbert-Huang transform analysis is used to analyze the frequency distribution and energy concentration areas to determine if resonance characteristics are present. For example, during a construction project, two high-frequency surge events were detected one hour after concrete vibration, with frequencies concentrated between 20 and 30 Hz and a duration of approximately 3 seconds. Analysis indicated that this was caused by a minor external impact on the formwork structure. This high-frequency strain feature extraction helps in the sensitive identification of early structural disturbances, allowing for timely adjustments to the construction schedule or structural reinforcement measures. Furthermore, by statistically analyzing the frequency, density, and waveform characteristics of high-frequency events, the overall compaction degree and stress release of the concrete can be inferred, improving the accuracy of structural forming quality assessment.

[0088] To accurately assess the true structural stress on concrete under temperature-free conditions, it is essential to strip the temperature strain component from the coupled signal, extracting the mechanical strain component representing the pure stress response. Based on the established coupled model, the predicted temperature-induced strain component is subtracted from the original strain data using either inverse signal modeling or residual fitting. Specifically, the model generates a temperature strain estimate sequence for each time point, which is then subtracted point-by-point from the original full strain sequence to obtain the net mechanical strain signal curve. This stripping result was verified by comparison with artificial loading tests (such as standard load hammer tests), showing a deviation of less than 10 μɛ, indicating that the temperature-influenced component has been effectively isolated. The stripped pure mechanical strain curve more intuitively reflects the mechanical processes such as concrete self-weight deformation, internal stress release, and formwork reaction force, serving as the core basis for subsequent stress evolution analysis, structural safety assessment, and intelligent monitoring threshold setting.

[0089] In this embodiment, the specific steps for analyzing the internal micropore structure based on the characteristics of pure mechanical strain components and predicting the long-term performance depth based on the real-time regional temperature twin model to generate concrete condition prediction results are as follows:

[0090] Dynamic wavelet transform decomposition is performed on the pure mechanical strain component characteristics to extract the time-frequency domain data of mechanical strain;

[0091] Strain continuity analysis is performed based on time-frequency domain data of mechanical strain to extract mechanical strain continuity parameters at multiple locations;

[0092] The internal micropore structure was analyzed based on the multi-position mechanical strain continuity parameters, and the micropore structure parameters were obtained.

[0093] The time-series evolution of micropore structure parameters under freeze-thaw cycles was investigated, and a mathematical model of the time-series evolution of freeze-thaw cycles was constructed.

[0094] A cumulative freeze-thaw damage analysis was performed on the mathematical model of freeze-thaw cycle time evolution to obtain the cumulative law of freeze-thaw damage.

[0095] Based on the real-time regional temperature twin model and the cumulative law of freeze-thaw damage, a deep neural network is used to learn the development trend of concrete and perform long-term performance deep trend prediction to generate concrete trend prediction results.

[0096] In this embodiment, pure mechanical strain signals typically contain multi-scale and multi-frequency dynamic characteristics, especially under extreme environments such as freeze-thaw cycles, where the signals exhibit significant time-varying non-stationarity. To comprehensively capture the dynamic changes in mechanical strain, decomposing the pure mechanical strain signal using Dynamic Wavelet Transform (DWT) is crucial. DWT, by adaptively adjusting the window function width, can more accurately analyze the instantaneous frequency and amplitude changes in the signal, balancing time and frequency domain resolution. In practice, a suitable multi-scale wavelet basis, such as Symlet or Coiflet, is selected for the structural signal, combined with multi-level decomposition to process the concrete strain time-series data. For example, performing a 7-level dynamic wavelet decomposition on a mechanical strain sequence with a sampling frequency of 10Hz can obtain detail coefficients and approximation coefficients covering frequency bands from 0.01Hz to 5Hz. The coefficients in each frequency band reflect the influence of different physical processes on mechanical strain, such as microcrack development, pore water pressure changes, and transient impacts from freeze-thaw cycles. After dynamic wavelet transform processing, a time-frequency domain dataset of mechanical strain signals was formed, revealing the energy concentration region and frequency migration law in the time-varying process, laying a data foundation for subsequent multi-point continuity analysis and microstructure inference. Multi-location mechanical strain continuity is an important indicator for evaluating the integrity and homogeneity of the internal structure of concrete. Based on the time-frequency domain data obtained in the first step, the strain continuity parameter was quantitatively extracted by calculating the time correlation, phase synchronization, and energy coupling degree between different sensor nodes. This process can employ time-frequency correlation coefficient analysis, cross-spectral analysis, and coherence function calculation methods. Mechanical strain time-frequency data from four key monitoring points of the wind turbine foundation were selected to calculate the synchronous response of each node in different frequency bands. A high continuity parameter value indicates that the structure exhibits consistent mechanical behavior in that frequency band, reflecting a relatively uniform structure without significant cracks or delamination; conversely, a decrease in continuity may indicate internal microscopic damage or uneven pore structure. In the experiment, after dozens of freeze-thaw test samples, the continuity parameter showed significant fluctuations in the high-frequency band, especially in the 2-4 Hz frequency range, indicating the gradual propagation of local microcracks, leading to local strain response desynchronization. This analysis helps to identify and locate early microscopic damage.

[0097] The continuous change in mechanical strain is directly related to the evolution of the internal pore and crack structure of concrete. By establishing an inverse model between strain continuity and pore structure parameters, quantitative analysis of microscopic pore characteristics can be achieved. Specific methods include statistical modeling based on multi-point continuity parameters and machine learning regression algorithms (such as support vector regression and random forests) to estimate porosity, pore size distribution, and pore connectivity. In freeze-thaw cycle experiments, this method, combined with experimentally measured porosity and scanning electron microscopy (SEM) images, revealed a significant negative correlation between the decrease in continuity parameters and the increase in porosity. Through training on a large amount of freeze-thaw cycle data, a pore structure parameter prediction model was established to accurately reflect the destructive evolution process of the internal microstructure of concrete caused by freeze-thaw cycles. Typical pore sizes increased from an initial 50 micrometers to over 100 micrometers, and connectivity indices decreased with increasing cycle count. Time-series analysis of the dynamic changes in pore structure parameters with the number of freeze-thaw cycles was conducted to construct a mathematical model of the freeze-thaw cycle time-series evolution. The model typically employs a nonlinear recursive or exponential decay model to reflect the gradual accumulation process of pore expansion and structural degradation under freeze-thaw action. The model is based on experimental data fitting, integrating multiple parameters such as porosity, pore size, and connectivity to reveal their changing trends with the number of cycles. For example, porosity exhibits a three-stage change with freeze-thaw cycles: a slow increase in the initial stage, a rapid increase in the middle stage, and a tendency to saturate in the later stage. The fitting uses a piecewise function combined with a smooth transition technique to accurately characterize the evolution process.

[0098] Based on a time-series evolution model, the gradual accumulation effect of freeze-thaw damage in concrete is calculated using cumulative damage theory. This analysis typically uses the damage variable D as a measure of damage severity, with 0 representing no damage and 1 representing complete failure. The number of freeze-thaw cycles is used as a time variable to map the degree of pore structure deterioration to a function of the damage variable, quantifying the damage accumulation law. In practical engineering, combining measured mechanical strain continuity parameters and pore structure evolution models, the Miner linear cumulative damage criterion or nonlinear cumulative model is used to reveal the growth rate and critical threshold of freeze-thaw damage. The freeze-thaw damage accumulation law shows that initial cyclic damage is relatively slow, entering an accelerated damage stage after 20-30 cycles, and approaching a critical value at 50 cycles, indicating a significant decline in the function of the concrete structure. Using the aforementioned temperature twin model and freeze-thaw damage accumulation data as input features, a deep neural network (DNN) is constructed for time-series prediction of concrete performance. The DNN, through its multi-layer nonlinear mapping capabilities, uncovers the complex intrinsic relationships between temperature, strain, and freeze-thaw damage, achieving accurate simulation and prediction of long-term performance trends. The network structure design typically involves convolutional layers to extract spatiotemporal features, recurrent neural networks (such as LSTM) to capture temporal dependencies, and fully connected layers to achieve the final output. During training, real-world multi-time-point data on temperature, strain, freeze-thaw cycles, and damage indicators are used, with mean squared error (MSE) as the loss function, to iteratively optimize the network weights. The prediction results not only cover the future development trend of concrete damage under freeze-thaw cycles but also dynamically reflect the real-time stress state based on changes in ambient temperature, forming an accurate concrete condition prediction model. In a field application, the model successfully predicted the intensification trend of freeze-thaw damage in a wind turbine foundation, providing early guidance for the implementation of local reinforcement measures and significantly extending the service life of the structure.

[0099] In this embodiment, the specific steps for predicting the risk of strain abnormal cracking and quantifying the comprehensive construction risk level based on the concrete condition prediction results to generate a dynamic construction risk level curve are as follows:

[0100] High stress concentration areas are identified and marked based on the concrete condition prediction results.

[0101] Strain abnormal cracking risk prediction is performed in high stress concentration areas to obtain strain cracking risk characteristics;

[0102] Abrupt temperature anomaly amplitude analysis was performed on the concrete condition prediction results, and the characteristics of the abrupt temperature anomaly were marked.

[0103] A risk assessment of frost damage during construction is conducted based on the characteristics of abnormal temperature changes, in order to obtain a frost damage risk assessment value.

[0104] Based on the characteristics of strain cracking risk and the assessment value of frost damage risk, the comprehensive construction risk level is quantified, and a dynamic construction risk level curve is generated.

[0105] In this embodiment, based on the strain and temperature field data output by the concrete condition prediction model, finite element post-processing technology is used to calculate the strain gradient and stress distribution characteristics of each measuring point and its neighborhood. The strain gradient is obtained by differential calculation of the strain values ​​of spatially discrete nodes, reflecting the rate of local stress change. Considering the low-temperature brittleness unique to concrete in cold regions, a strain gradient threshold is typically set to 80% of the maximum value as the identification standard. Combining the elastic modulus and Poisson's ratio parameters of concrete, the constitutive relation model is used to convert and obtain the local stress value, and the stress concentration factor (SCF) is calculated to assess the severity of local stress peaks. In this step, a high-resolution stress field is constructed using spatial interpolation technology to achieve precise localization of micro-region stress changes. This method can effectively extract high-stress concentration areas and mark them using a spatial coordinate system, laying the foundation for subsequent strain anomaly and cracking risk prediction. In the experiment, for wind turbine foundation concrete, the spatial grid size was set to 5cm, and the maximum strain gradient threshold was approximately 1500με / m, meeting the requirements for capturing structural details. For the identified high-stress concentration areas, the strain time-series signals in these areas were further analyzed. A multi-scale anomaly detection algorithm, including sliding window statistical analysis and a support vector machine (SVM)-based classifier, was employed to automatically identify anomalous strain events. First, the coefficient of variation was calculated by statistically analyzing the strain mean and variance within the sliding window to detect anomalous fluctuations. Then, an SVM classification model was applied to divide the time-series signals into normal and anomalous categories. Combined with the critical fracture strain threshold (typically set at 300–350 microstrain, based on material tensile strength and microcrack propagation research results), potential cracking risks were identified. The duration and peak amplitude of anomalous strain events were simultaneously included in the risk weight calculation, and dynamic risk features were extracted using a time-weighted averaging method. In the experiment, fiber Bragg grating sensor data with a sampling frequency of 100 Hz was used to ensure the capture of transient strain abrupt changes, greatly improving the timeliness and accuracy of cracking risk prediction. The risk feature vector can be used in structural health monitoring and intelligent early warning systems.

[0106] Abnormal temperature changes in concrete surface and internal temperature data often foreshadow frost damage or other construction anomalies. A sliding window-based change point detection algorithm is used to monitor temperature time-series signals in real time. The window width is typically set to 10 minutes. The temperature change rate is calculated, and when the instantaneous change rate exceeds 1.5℃ / minute and the amplitude change exceeds 3℃, it is identified as a temperature anomaly. To improve detection robustness, a mean-shift filtering algorithm is added to eliminate interference from low-frequency ambient temperature fluctuations. Event localization is achieved using the spatial coordinates of a synchronous fiber optic sensor, facilitating association with specific construction locations. Experiments demonstrate that in simulated freeze-thaw cycles, this method can accurately capture sudden temperature drops caused by abrupt changes in cooling rate, successfully providing early warning of potential frost damage risks. This feature extraction provides crucial foundational data for the quantitative assessment of frost damage risks. The frost damage risk assessment model is based on fuzzy comprehensive evaluation theory, integrating multi-dimensional information such as temperature anomaly amplitude, duration, frequency, and concrete physical parameters (moisture content, porosity) for comprehensive evaluation. The model first normalizes each input parameter to the [0,1] interval and defines a fuzzy membership function to describe the risk level attribution of the parameters. By assigning weights based on expert experience, the weights of each risk factor are comprehensively calculated, and the final frost damage risk assessment value is obtained using fuzzy matrix operations. In the typical frost damage risk threshold settings, a sudden temperature drop greater than 5°C lasting for more than 30 minutes and a porosity exceeding 12% significantly increase the risk value. This model supports dynamic updates of input data, enabling real-time monitoring of frost damage risks during construction. The model was validated through multi-point on-site temperature and humidity monitoring and laboratory freeze-thaw performance tests. The assessment results are highly correlated with actual frost damage manifestations, significantly improving the scientific rigor and accuracy of risk warnings.

[0107] This method integrates the strain cracking risk feature vector with the frost damage risk assessment value, employing a weighted scoring mechanism to quantify comprehensive construction risks. The weighting design is based on engineering experience and statistical analysis, typically assigning a weight of 0.6 to strain risk and 0.4 to frost damage risk. The risk index is calculated through a weighted summation, and the result is mapped to a construction risk level (safe, warning, and dangerous). The dynamic risk level curve uses time-series data to reflect risk fluctuations throughout the construction process. Curve generation, combined with real-time monitoring data, enables continuous and dynamic updates of risk in time and space. Experimental applications show that this method can accurately reflect the safety status of wind turbine foundation concrete under different freeze-thaw environments and construction stages, effectively capturing risk peaks and providing construction managers with a scientific basis for adjusting construction plans and implementing risk mitigation measures. This significantly improves the risk control capabilities and engineering safety assurance level of concrete construction in high-altitude and cold regions.

[0108] In this embodiment, the specific steps for adjusting construction time and making risk avoidance decisions based on the dynamic construction risk level curve to perform intelligent wind turbine construction optimization and monitoring are as follows:

[0109] Obtain construction execution logs; analyze the construction project progress based on the construction execution logs, and extract the project progress from multiple construction cycles;

[0110] The preset construction time points are calculated based on the preset progress of multiple construction cycle projects;

[0111] The construction time is adjusted according to the preset construction time point based on the dynamic construction risk level curve to obtain the optimal construction time point.

[0112] Risk avoidance decisions are made based on dynamic construction risk level curves, and risk avoidance strategies are constructed.

[0113] Optimize and monitor the construction of smart wind turbines based on the optimal construction time and risk avoidance strategies.

[0114] In this embodiment, construction execution logs are typically recorded automatically by the construction management system (such as a BIM platform or construction scheduling system), containing information such as timestamps, durations, task status, and personnel involved in construction nodes such as concrete pouring, vibration, thermal insulation curing, and surface treatment. To ensure data integrity and consistency, the log format must be a structured data format, such as CSV or JSON. The data collection frequency is set to update every 5 minutes to ensure dynamic reflection of the construction status. Time series clustering algorithms (such as K-means or DBSCAN) are used to classify and analyze logs from multiple construction cycles, extracting the actual start time, execution duration, and time interval of key processes to form a standardized project progress template. By comparing and analyzing construction behaviors under different time periods and climatic conditions, the changing patterns of construction rhythm can be identified, providing a reference for subsequent time prediction and dynamic adjustment. In the experiment, the logs of six consecutive wind turbine foundation construction cycles were analyzed, identifying an average pouring duration of 4.5 hours and a curing preparation interval of approximately 1.2 hours, forming a standard multi-cycle progress map. Based on established standard progress maps of multiple construction cycles, the system further extrapolates the preset construction time points. This step involves modeling the time relationships between different processes, including start and end dates, buffer times, and critical path analysis, to form a construction time network. Using the Critical Path Method (CPM) or a Petri net model, logical constraints and time relationships are applied to each process, automatically generating the theoretically optimal construction schedule. Considering the impact of temperature fluctuations in high-altitude and cold regions, the preset time points are adjusted for adaptability using historical meteorological data (such as temperature and wind speed). The system can calculate the preset time points for each critical process based on the planned start time and the standard execution cycle of each node. For example, in the construction of wind turbine concrete foundations, if the pouring is preset to start at 07:00, considering the rising temperature trend and temperature sensitivity, the insulation covering node is automatically preset to be completed before 11:45 to avoid early hydration heat loss. These preset time points form a process time baseline table, serving as a reference baseline for subsequent dynamic adjustments.

[0115] The dynamic construction risk level curve is generated in real time by the pre-monitoring system, comprehensively considering factors such as temperature, strain, frost damage, and cracking risk to reflect the risk change trend between the current and predicted time periods. Construction time optimization requires cross-matching analysis of this curve with preset time points. A multi-objective optimization algorithm (such as particle swarm optimization or genetic algorithm) is used to search for the time node with the lowest risk as the optimal construction time point, while satisfying the construction window and resource constraints. For example, if the original planned concrete pouring time is 07:00, but the risk curve shows a sharp drop in external temperature between 07:00 and 09:00, with a frost damage risk assessment value exceeding 0.75 (high risk), the system automatically shifts the construction time to the low-risk segment between 10:30 and 13:00. In experimental deployment, by interfacing with the on-site temperature prediction system and dynamically adjusting the construction window, the average daily avoidance rate of high-risk construction periods was increased by 42%, significantly improving the safety and stability of the concrete initial setting and curing environment. The final output optimal construction time point serves as the time benchmark for construction execution. To achieve dynamic risk avoidance throughout the entire construction process, a highly operable risk avoidance strategy system needs to be constructed. Strategy design uses a dynamic construction risk level curve as the decision-making basis, combining process type, risk tolerance, and emergency response mechanisms to set strategy rules. Strategies are divided into multi-level responses, including time-based avoidance (adjusting start time), process-based avoidance (changing curing methods), and resource-based avoidance (adding insulation equipment). The system is driven by a rule engine (such as Drools or a custom rule base), automatically triggering corresponding decisions based on the risk level. For example, if the risk level of the pouring process is "high," the system automatically adjusts the construction sequence and instructs the on-site use of heated vibrators and rapid-insulating blankets for coordinated handling. In applications at high-altitude, cold-weather wind turbine construction sites, the constructed avoidance strategy successfully avoided six instances of erroneous construction behavior under extreme low temperatures, significantly improving construction quality and initial structural stability, and reducing the cracking rate by 28%. This strategy base can continuously learn and optimize itself, gradually forming a more adaptable intelligent construction response system.

[0116] The calculated optimal construction time and the established risk avoidance strategies are integrated into the construction scheduling system to achieve intelligent scheduling and execution control of wind turbine foundation construction. The system links with on-site PLC equipment, construction control terminals, and monitoring platforms via API interfaces, realizing a closed-loop control process from issuing construction start commands and controlling equipment switching to real-time data feedback and dynamic risk curve updates. In terms of intelligent monitoring, a distributed fiber optic temperature and strain sensing system, combined with timestamp synchronization and event triggering mechanisms, tracks the execution status of key processes and changes in environmental parameters in real time, achieving full digital monitoring and scheduling feedback at the construction site. Experimental applications show that this intelligent scheduling mode improves the construction efficiency of wind turbine foundations in cold regions by 12% and reduces the construction accident rate by 34%, effectively supporting intelligent decision-making and robust execution of large-volume concrete construction tasks in extreme environments. This marks a significant step forward for wind power foundation construction in cold regions, entering a stage of intelligent collaboration and proactive risk avoidance.

[0117] In this embodiment, a system for monitoring the temperature and stress of wind turbine foundations is provided, for performing the method for monitoring the temperature and stress of wind turbine foundations as described above, including:

[0118] The temperature gradient distribution module is used to collect monitoring signals of concrete hydration heat release, mine the temperature gradient distribution and perform finite element numerical simulation to construct a temperature evolution model of the concrete pouring process.

[0119] The temperature fluctuation mapping module is used to acquire concrete surface temperature monitoring parameters, calculate surface temperature at multiple time points, and map the surface environment temperature fluctuation to the temperature evolution model of the concrete pouring process to build a real-time regional temperature twin model.

[0120] The strain sensing module is used to identify strain state sensing parameters during concrete pouring, perform multi-scale spectral decomposition, and then strip the temperature strain signal to obtain the pure mechanical strain component characteristics.

[0121] The situation prediction module is used to analyze the internal micro-pore structure based on the characteristics of pure mechanical strain components, and to perform long-term performance in-depth situation prediction based on the real-time regional temperature twin model, generating concrete situation prediction results.

[0122] The risk prediction module is used to predict the risk of strain abnormal cracking and quantify the comprehensive construction risk level based on the concrete condition prediction results, and generate a dynamic construction risk level curve.

[0123] The risk avoidance decision-making module is used to adjust the construction time and make risk avoidance decisions based on the dynamic construction risk level curve, so as to implement intelligent wind turbine construction optimization and monitoring.

[0124] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0125] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for monitoring the temperature and stress of wind turbine foundations, characterized in that, Includes the following steps: Collect monitoring signals of heat release from concrete hydration, mine temperature gradient distribution and perform finite element numerical simulation to construct a temperature evolution model of the concrete pouring process; The surface temperature monitoring parameters of concrete were obtained, the surface temperature at multiple time points was calculated, and the surface temperature fluctuation of the concrete pouring process temperature evolution model was mapped to the surface temperature fluctuation model to construct a real-time regional temperature twin model. The strain state sensing parameters during the concrete pouring process are identified, multi-scale spectral decomposition is performed, and then temperature strain signal stripping is performed to obtain the pure mechanical strain component characteristics. The internal microstructure of concrete is analyzed based on the characteristics of pure mechanical strain components, and the long-term performance depth is predicted based on the real-time regional temperature twin model, generating concrete condition prediction results. The results of concrete condition prediction are used to predict the risk of strain abnormal cracking and quantify the comprehensive construction risk level, generating a dynamic construction risk level curve. Based on the dynamic construction risk level curve, construction time adjustments and risk avoidance decisions are made to optimize and monitor the construction of intelligent wind turbines.

2. The method for monitoring the temperature and stress of wind turbine foundations according to claim 1, characterized in that, The specific steps for collecting monitoring signals of concrete hydration heat release, mining temperature gradient distribution, and conducting finite element numerical simulation to construct a temperature evolution model of the concrete pouring process are as follows: Deploy distributed fiber optic temperature sensors to monitor the concrete pouring process in the wind turbine construction area in real time and collect monitoring signals of concrete hydration heat release. Continuous spatial temperature sampling is performed on the concrete hydration heat release monitoring signal to extract and generate spatiotemporal hydration heat release characteristic parameters; Calculate the occurrence time of the hydration heat peak based on the spatiotemporal hydration heat release characteristic parameters, and extract the heat peak timestamp; Temperature gradient distribution is mined from multi-temporal and spatial hydration heat release monitoring parameters to generate temperature gradient distribution patterns; A finite element numerical simulation model of the temperature evolution during concrete pouring was constructed based on the hot peak timestamp and temperature gradient distribution.

3. The method for monitoring the temperature and stress of wind turbine foundations according to claim 2, characterized in that, The specific steps for continuously sampling the spatial temperature of the concrete hydration heat release monitoring signal and extracting spatiotemporal hydration heat release characteristic parameters are as follows: Calculate the spatial location of the distributed fiber optic temperature sensor to obtain the distribution coordinates of each sensor node; Identify the acquisition frequency of the distributed fiber optic temperature sensor; The frequency difference between multiple devices is calculated for the acquisition frequency to generate the acquisition frequency difference between different sensors; Based on the difference in acquisition frequency, time alignment processing is performed to obtain standardized acquisition timestamps and frequencies; Abnormal noise is identified in the monitoring signal of heat release from concrete hydration, and abnormal noise points are marked. Calculate the waveform and abrupt change range of the abnormal noise points; Based on the waveform and the range of abrupt changes, adaptive filtering optimization is performed to construct an optimized signal for heat of hydration release. Based on the distribution location coordinates, the optimized signal of heat of hydration release is continuously sampled spatially, and the time stamp is uniformly calibrated based on the standardized acquisition timestamp and frequency to generate spatiotemporal heat of hydration release characteristic parameters.

4. The method for monitoring the temperature and stress of wind turbine foundations according to claim 1, characterized in that, The specific steps for acquiring concrete surface temperature monitoring parameters, calculating surface temperature at multiple time points, mapping surface temperature fluctuations to the concrete pouring process temperature evolution model, and constructing a real-time regional temperature twin model are as follows: Continuously collect concrete surface temperature monitoring parameters in the wind turbine construction area; The surface temperature of concrete was calculated at multiple time points based on the monitoring parameters, and the temperature fluctuation trend was analyzed to generate a multi-time point surface temperature trend curve. Based on the concrete surface temperature monitoring parameters, the monitoring location is determined, and the spatial temperature distribution is analyzed to obtain a surface spatial temperature distribution map. Based on the surface temperature trend curve at multiple time points, the surface spatial temperature distribution map is rendered to show temperature trend fluctuations, and a surface temperature distribution field of the slow-concrete concrete is constructed. Based on the surface temperature distribution field of slow-setting soil, the temperature evolution model of the concrete pouring process is mapped to the surface temperature fluctuation, and a real-time regional temperature twin model is constructed.

5. The method for monitoring the temperature and stress of wind turbine foundations according to claim 1, characterized in that, The specific steps for identifying strain state sensing parameters during concrete pouring, performing multi-scale spectral decomposition, and then stripping the temperature strain signal to obtain the pure mechanical strain component features are as follows: A fiber Bragg grating strain sensor array is simultaneously implanted during the concrete pouring process to extract the strain state sensing parameters of the concrete in real time. The strain state sensing parameters are subjected to multi-scale spectral decomposition to obtain signal components in different frequency bands; The low-frequency components of the signal are identified, and the amplitude and frequency are calculated to obtain the low-frequency temperature strain characteristics. The signal components are subjected to high-frequency signal detection, and transient impact and resonance feature analysis is performed to extract high-frequency mechanical strain features. A temperature-mechanical strain coupling model is constructed based on the low-frequency temperature strain characteristics and the high-frequency mechanical strain characteristics. Temperature strain signal stripping was performed on the temperature-mechanical strain coupling model to obtain the pure mechanical strain component characteristics.

6. The method for monitoring the temperature and stress of wind turbine foundations according to claim 1, characterized in that, The specific steps for analyzing the internal micropore structure based on the characteristics of pure mechanical strain components and predicting the long-term performance depth based on the real-time regional temperature twin model to generate concrete condition prediction results are as follows: Dynamic wavelet transform decomposition is performed on the pure mechanical strain component characteristics to extract the time-frequency domain data of mechanical strain; Strain continuity analysis is performed based on time-frequency domain data of mechanical strain to extract mechanical strain continuity parameters at multiple locations; The internal micropore structure was analyzed based on the multi-position mechanical strain continuity parameters, and the micropore structure parameters were obtained. The time-series evolution of micropore structure parameters under freeze-thaw cycles was investigated, and a mathematical model of the time-series evolution of freeze-thaw cycles was constructed. A cumulative freeze-thaw damage analysis was performed on the mathematical model of freeze-thaw cycle time evolution to obtain the cumulative law of freeze-thaw damage. Based on the real-time regional temperature twin model and the cumulative law of freeze-thaw damage, a deep neural network is used to learn the development trend of concrete and perform long-term performance deep trend prediction to generate concrete trend prediction results.

7. The method for monitoring the temperature and stress of wind turbine foundations according to claim 1, characterized in that, The specific steps for predicting strain abnormal cracking risk and quantifying comprehensive construction risk level based on concrete condition prediction results to generate a dynamic construction risk level curve are as follows: High stress concentration areas are identified and marked based on the concrete condition prediction results. Strain abnormal cracking risk prediction is performed in high stress concentration areas to obtain strain cracking risk characteristics; Abrupt temperature anomaly amplitude analysis was performed on the concrete condition prediction results, and the characteristics of the abrupt temperature anomaly were marked. A risk assessment of frost damage during construction is conducted based on the characteristics of abnormal temperature changes, in order to obtain a frost damage risk assessment value. Based on the characteristics of strain cracking risk and the assessment value of frost damage risk, the comprehensive construction risk level is quantified, and a dynamic construction risk level curve is generated.

8. The method for monitoring the temperature and stress of wind turbine foundations according to claim 1, characterized in that, The specific steps for adjusting construction time and making risk avoidance decisions based on the dynamic construction risk level curve to optimize and monitor intelligent wind turbine construction are as follows: Obtain construction execution logs; analyze the construction project progress based on the construction execution logs, and extract the project progress from multiple construction cycles; The preset construction time points are calculated based on the preset progress of multiple construction cycle projects; The construction time is adjusted according to the preset construction time point based on the dynamic construction risk level curve to obtain the optimal construction time point. Risk avoidance decisions are made based on dynamic construction risk level curves, and risk avoidance strategies are constructed. Optimize and monitor the construction of smart wind turbines based on the optimal construction time and risk avoidance strategies.

9. A system for monitoring the temperature and stress of wind turbine foundations, characterized in that, For performing the method for monitoring the temperature and stress of a wind turbine foundation as described in claim 1, comprising: The temperature gradient distribution module is used to collect monitoring signals of concrete hydration heat release, mine the temperature gradient distribution and perform finite element numerical simulation to construct a temperature evolution model of the concrete pouring process. The temperature fluctuation mapping module is used to acquire concrete surface temperature monitoring parameters, calculate surface temperature at multiple time points, and map the surface environment temperature fluctuation to the temperature evolution model of the concrete pouring process to build a real-time regional temperature twin model. The strain sensing module is used to identify strain state sensing parameters during concrete pouring, perform multi-scale spectral decomposition, and then strip the temperature strain signal to obtain the pure mechanical strain component characteristics. The situation prediction module is used to analyze the internal micro-pore structure based on the characteristics of pure mechanical strain components, and to perform long-term performance in-depth situation prediction based on the real-time regional temperature twin model, generating concrete situation prediction results. The risk prediction module is used to predict the risk of strain abnormal cracking and quantify the comprehensive construction risk level based on the concrete condition prediction results, and generate a dynamic construction risk level curve. The risk avoidance decision-making module is used to adjust the construction time and make risk avoidance decisions based on the dynamic construction risk level curve, so as to implement intelligent wind turbine construction optimization and monitoring.

Images