Mass concrete raft foundation three-dimensional temperature field reconstruction method

Abstract

The invention relates to a mass concrete raft foundation three-dimensional temperature field reconstruction method, and relates to the technical field of concrete raft foundation structure three-dimensional temperature field reconstruction. The method comprises the following steps: arranging temperature sensing optical fibers in a plurality of temperature measurement areas of a concrete raft foundation to be measured; constructing an optical fiber temperature measurement monitoring system, and obtaining space-time temperature data; a temperature field prediction model is constructed, the temperature field prediction model is based on a physical information neural network, and a data loss function, a physical loss function and an initial condition loss function are fused; performing multi-objective loss function joint optimization training on the temperature field prediction model based on the space-time temperature data; utilizing the trained temperature field prediction model to predict the temperature value of the concrete raft foundation to be measured at any time and space point coordinates; and reconstructing a three-dimensional temperature field based on a prediction result output by the temperature field prediction model. By adopting the method, high-precision reconstruction of the three-dimensional temperature field of the mass concrete raft foundation can be realized.

Description

Technical Field

[0001] The present application relates to the technical field of three-dimensional temperature field reconstruction of concrete raft foundation structures, and particularly to a method for reconstructing the three-dimensional temperature field of a large-volume concrete raft foundation. Background Art

[0002] At present, with the rapid development of engineering projects, large-volume concrete is widely used in the foundation parts of underground spaces such as buildings and subway stations. During the pouring and maintenance of large-volume concrete, due to the poor heat conduction characteristics of concrete and the large amount of heat generated by cement in the hydration reaction, it is easy to cause excessive temperature differences inside the concrete structure, and thus form large thermal stresses. Projects such as commercial complexes, subway hub projects, and super-high-rise core tube foundations have not monitored the temperature gradient in real time, resulting in warping deformations of raft foundations to varying degrees. Subsequently, groundwater seeps into the foundation pit through cracks, causing leakage, delaying the construction period, and increasing construction costs.

[0003] The current calculation of the temperature field of concrete raft foundations mainly relies on two types of methods. One is the reconstruction method based on the measured data of discrete point thermometers, and the disadvantage is that the reconstruction accuracy is insufficient due to sparse measurement points; the other is the numerical simulation method based on the assumption of isotropic heat conduction, and the disadvantage is that it often ignores the time-varying influence of the degree of hydration on the temperature history curve, resulting in simulation deviation. Therefore, there is an urgent need for a temperature field analysis method with high fidelity and low calculation cost to take into account the time-varying characteristics of hydration heat during the construction period and the material anisotropy, so as to improve the temperature control reliability of large-volume concrete. Summary of the Invention

[0004] Based on this, it is necessary to provide a method for reconstructing the three-dimensional temperature field of a large-volume concrete raft foundation in view of the above technical problems.

[0005] In a first aspect, a method for reconstructing the three-dimensional temperature field of a large-volume concrete raft foundation is provided, and the method includes: Laying temperature sensing optical fibers in multiple temperature measurement areas of the concrete raft foundation to be measured; Constructing an optical fiber temperature measurement monitoring system to obtain spatio-temporal temperature data; Constructing a temperature field prediction model, where the temperature field prediction model is based on a physics-informed neural network and integrates a data loss function, a physical loss function, and an initial condition loss function; Based on the spatio-temporal temperature data, performing joint optimization training on the multi-objective loss function of the temperature field prediction model; Using the trained temperature field prediction model to predict the temperature values at any time and space point coordinates of the concrete raft foundation to be measured; Based on the prediction results output by the temperature field prediction model, reconstructing the three-dimensional temperature field of the concrete raft foundation to be measured.

[0006] As an alternative implementation, laying the temperature sensing optical fiber in multiple temperature measurement areas of the concrete raft foundation to be measured includes: Determining the laying pattern of the temperature sensing optical fiber on the surface of the concrete raft foundation to be measured according to the surface geometry of the concrete raft foundation to be measured; Laying the temperature sensing optical fiber in the internal geometric center area of the concrete raft foundation to be measured.

[0007] As an alternative implementation, determining the laying pattern of the temperature sensing optical fiber on the surface of the concrete raft foundation to be measured according to the surface geometry of the concrete raft foundation to be measured includes: If the surface geometry of the concrete raft foundation to be measured is only vertically symmetric, the surface laying pattern is "┴" or "┬"; If the surface geometry of the concrete raft foundation to be measured is only horizontally symmetric, the surface laying pattern is "├" or "┤"; If the surface geometry of the concrete raft foundation to be measured is centrosymmetric, the surface laying pattern is "└" or "┌" or "┐" or "┘"; If the surface geometry of the concrete raft foundation to be measured is asymmetric, the surface laying pattern is "┼".

[0008] As an alternative implementation, the optical fiber temperature measurement and monitoring system includes a temperature sensing optical fiber, a distributed optical fiber temperature sensor, and an optical fiber demodulator.

[0009] As an alternative implementation, the spatio-temporal temperature data includes time, spatial point coordinates, temperature, and the corresponding relationship among the three.

[0010] As an alternative implementation, constructing the temperature field prediction model includes: Setting the temperature field prediction model as a model framework based on a physics-informed neural network; Defining the model input variables as spatial point coordinates and time, and the model output variable as the temperature prediction value corresponding to the spatial point coordinates and the time; Constructing the data loss function; Constructing the physical loss function based on the partial differential equation of heat conduction; Constructing the initial condition loss function.

[0011] As an alternative implementation, performing joint optimization training of the multi-objective loss function on the temperature field prediction model based on the spatio-temporal temperature data includes: Construct a training data set based on the spatio-temporal temperature data; Construct a multi-objective joint loss function according to the data loss function, the physical loss function, the initial condition loss function, and a preset weight; During the training process, use the gradient descent algorithm to iteratively update the learnable parameters of the temperature field prediction model, and the training objective is to minimize the total loss function of the multi-objective joint loss function; In the region where the value of the physical loss function is greater than the preset physical residual threshold, increase the sampling points of the spatio-temporal temperature data and update the training data set.

[0012] As an optional implementation manner, after constructing the optical fiber temperature measurement monitoring system and obtaining the spatio-temporal temperature data, the method further includes: Use the wavelet-Kalman hybrid filtering algorithm for denoising; In the region where the spatio-temporal temperature data is missing or the noise is significant, use the interpolation method or the physical simulation method for data enhancement.

[0013] As an optional implementation manner, reconstructing the three-dimensional temperature field of the concrete raft foundation to be measured based on the prediction result output by the temperature field prediction model includes: When the number of prediction results output by the temperature field prediction model is less than the preset data scale threshold, call the fitrsvm support vector regression model through the Matlab software for three-dimensional temperature field fitting; When the number of prediction results output by the temperature field prediction model is greater than the data scale threshold, perform three-dimensional temperature field reconstruction through the griddata function based on GPU or custom CUDA code.

[0014] As an optional implementation manner, the method further includes: Compare the three-dimensional temperature field of the concrete raft foundation to be measured with the finite element simulation result or the measured data; Calculate the temperature prediction error based on a preset error evaluation index; Optimize the temperature field prediction model based on the distribution result of the temperature prediction error.

[0015] In a second aspect, a computer device is provided, including a memory and a processor, where a computer program that can run on the processor is stored on the memory, and when the processor executes the computer program, the method steps described in any item of the first aspect are implemented.

[0016] In a third aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the method steps described in any item of the first aspect are implemented.

[0017] The present application provides a method for reconstructing the three-dimensional temperature field of a large-volume concrete raft foundation, the method comprising: arranging temperature sensing optical fibers in multiple temperature measurement areas of a concrete raft foundation to be measured; constructing an optical fiber temperature measurement monitoring system to obtain spatiotemporal temperature data; constructing a temperature field prediction model, the temperature field prediction model is based on a physical information neural network, and integrates a data loss function, a physical loss function, and an initial condition loss function; based on the spatiotemporal temperature data, the temperature field prediction model is subjected to multi-objective loss function joint optimization training; the trained temperature field prediction model is used to predict the temperature value of the concrete raft foundation to be measured at any time and spatial point coordinates; based on the prediction results output by the temperature field prediction model, the three-dimensional temperature field of the concrete raft foundation to be measured is reconstructed. The technical solution provided by the embodiment of the present application brings at least the following beneficial effects: arranging temperature sensing optical fibers in multiple temperature measurement areas of a concrete raft foundation to be measured, combining the structural characteristics of the raft foundation, and scientifically arranging optical fiber paths on the surface and core areas, can achieve comprehensive coverage of key thermal change areas. A fiber optic temperature monitoring system is constructed, and distributed fiber optic sensing technology is used to obtain spatiotemporal temperature data covering the entire structure, thereby solving the problem of sparse distribution of traditional thermocouples and insufficient spatial continuity. A temperature field prediction model based on physical information neural network is constructed. The model integrates three types of constraints: data loss function, physical loss function, and initial condition loss function. It not only ensures high-precision fitting of the measured temperature data, but also ensures that the prediction results meet the physical laws described by the partial differential equation of heat conduction. At the same time, the initial deviation of the solution is restricted with the initial temperature field as the starting point, thereby enhancing the physical consistency and stability of the model. In the model training stage, a multi-objective loss function joint optimization strategy is adopted. By introducing three independent objectives of measured data error, physical residual, and initial state error, comprehensive constraints are performed, and adaptive sampling is implemented in areas with large physical residuals to increase training points, improve the prediction accuracy of the model in key thermal mutation areas, and solve the problems of inaccurate prediction, overfitting, or physical mismatch in areas with drastic changes in thermal gradients by traditional methods. The trained neural network model is used to predict the temperature of any spatiotemporal point coordinates, and the three-dimensional temperature field of the concrete raft foundation to be tested is completely reconstructed. In summary, the embodiments of the present application can achieve high-precision reconstruction of the global temperature field of a large-volume concrete raft foundation without adding a large number of measuring points, overcome the problems of traditional thermocouple local measurement being unable to obtain a continuous temperature field, the lack of physical consistency of traditional interpolation / simulation models, and the weak generalization ability of data-driven models. It provides an efficient, low-cost, and precision-controlled solution for temperature control monitoring and crack warning of engineering structures during the construction period.

[0018] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0020] Figure 1 It is a flowchart of a method for reconstructing the three-dimensional temperature field of a large-volume concrete raft foundation provided by an embodiment of the present application; Figure 2 It is a flowchart of a method for arranging temperature sensing optical fibers provided by an embodiment of the present application; Figure 3a It is a schematic diagram of the internal optical fiber arrangement of a raft foundation provided by an embodiment of the present application; Figure 3b It is a schematic plan view of the optical fiber arrangement of an up-and-down symmetric raft foundation provided by an embodiment of the present application; Figure 3c It is a schematic plan view of the optical fiber arrangement of a left-and-right symmetric raft foundation provided by an embodiment of the present application; Figure 3d It is a schematic plan view of the optical fiber arrangement of a centrally symmetric raft foundation provided by an embodiment of the present application; Figure 3e It is a schematic plan view of the optical fiber arrangement of an asymmetric raft foundation provided by an embodiment of the present application; Figure 4 It is a flowchart of a method for processing spatio-temporal temperature data provided by an embodiment of the present application; Figure 5 It is a flowchart of a method for constructing a temperature field prediction model provided by an embodiment of the present application; Figure 6 It is a flowchart of a method for training a temperature field prediction model provided by an embodiment of the present application; Figure 7 It is a flowchart of a method for reconstructing a three-dimensional temperature field provided by an embodiment of the present application; Figure 8 It is a flowchart of a method for optimizing a temperature field prediction model provided by an embodiment of the present application; Figure 9 It is a temperature monitoring comparison chart provided by an embodiment of the present application; Figure 10 It is a schematic structural diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0021] To make the objectives, technical solutions, and advantages of this application more clear and understandable, the following further elaborates on this application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely used to explain this application and are not intended to limit this application.

[0022] Next, in combination with the specific implementation manners, a method for reconstructing the three-dimensional temperature field of a large-volume concrete raft foundation provided in an embodiment of this application will be described in detail. Figure 1 The flowchart of a method for reconstructing the three-dimensional temperature field of a large-volume concrete raft foundation provided in an embodiment of this application is as Figure 1 shown, and the specific steps are as follows: S101, Lay temperature sensing optical fibers in multiple temperature measurement areas of the concrete raft foundation to be measured.

[0023] In implementation, based on the finite element simulation experiment analysis of the large-volume concrete raft foundation, the three-dimensional temperature field of the large-volume concrete raft foundation is symmetrically distributed, and the temperature distribution rule is that the temperature in the central area is the highest, the temperature on the upper surface and around is lower, and the temperature on the lower surface is slightly lower than that in the core area. Therefore, technicians can lay temperature measurement optical fibers on its upper and lower surfaces and the core area according to the geometric structure and symmetry characteristics of the concrete raft foundation to be measured to ensure the spatial continuity and structural stability of temperature acquisition. The laying path can be set along the main axis direction of the raft foundation surface, and the temperature sensing optical fibers can be fixed on the vertical and horizontal bars of the steel reinforcement cage by means of high-strength cable ties or embedded steel bar buckles, etc., to ensure that they do not shift or deform during the concrete pouring process.

[0024] As an optional implementation manner, Figure 2 The flowchart of a method for laying temperature sensing optical fibers provided in an embodiment of this application is as Figure 2 shown, and the specific steps of laying the temperature sensing optical fibers in multiple temperature measurement areas of the concrete raft foundation to be measured in step S101 are as follows: S201, Determine the laying form of the temperature sensing optical fibers on the surface of the concrete raft foundation to be measured according to the surface geometric shape of the concrete raft foundation to be measured.

[0025] In implementation, the computer can select the optical fiber laying scheme corresponding to its structure by analyzing the plane geometric shape and symmetry characteristics of the concrete raft foundation to be measured.

[0026] S202, Lay temperature sensing optical fibers in the internal geometric center area of the concrete raft foundation to be measured.

[0027] In implementation, the position of the internal geometric center area of the raft foundation can be confirmed through the structural design drawings. Figure 3a The schematic diagram of the internal optical fiber laying of the raft foundation provided in an embodiment of this application is as Figure 3aAs shown, a temperature sensing optical fiber 310 can be pre-fixed inside the steel reinforcement cage 330 in an "S" shape path, with the path transversely crossing the entire core area to achieve dynamic capture of the internal temperature rise peak value. To ensure accurate temperature measurement, the temperature sensing optical fiber is fixed to the steel bars by spaced ties. After the layout is completed, the optical fiber can be pre-tightened and the protective shell can be fixed to avoid damage or displacement during concrete pouring and vibration.

[0028] As an alternative implementation, in step S201, the specific method for determining the layout form of the temperature sensing optical fiber on the surface of the concrete raft foundation to be measured according to the surface geometry of the concrete raft foundation to be measured is as follows: If the surface geometry of the concrete raft foundation to be measured is only vertically symmetric, the surface layout form is "┴" or "┬"; If the surface geometry of the concrete raft foundation to be measured is only horizontally symmetric, the surface layout form is "├" or "┤"; If the surface geometry of the concrete raft foundation to be measured is centrosymmetric, the surface layout form is "└" or "┌" or "┐" or "┘"; If the surface geometry of the concrete raft foundation to be measured is asymmetric, the surface layout form is "┼".

[0029] In implementation, according to the engineering design drawings and on-site measurement results, the symmetry characteristics of the concrete raft foundation in the plane direction can be judged. Symmetry determines the uniformity of the heat conduction path and also affects the surface temperature distribution characteristics. Therefore, the layout form should match its symmetric structure to achieve the optimal coverage of the temperature measurement points. Figure 3b This is a schematic plan view of the optical fiber layout of an up-and-down symmetric raft foundation provided by an embodiment of the present application. As Figure 3b shown, for a raft foundation 320 that is only up-and-down symmetric (such as a rectangular structure with a long longitudinal side and a short transverse side), select a path of laying a "┴" or "┬" (not shown in the figure) temperature sensing optical fiber 310 along the central axis to cover the middle area of the upper surface and the upper and lower edges. Figure 3c This is a schematic plan view of the optical fiber layout of a left-and-right symmetric raft foundation provided by an embodiment of the present application. As Figure 3c shown, for a raft foundation 320 that is only left-and-right symmetric, use a "├" or "┤" (not shown in the figure) layout, and introduce the temperature sensing optical fiber 310 from the right or left to cover the transverse temperature difference area of the structure. Figure 3d This is a schematic plan view of the optical fiber layout of a centrosymmetric raft foundation provided by an embodiment of the present application. As Figure 3dAs shown, for the centrally symmetrical raft foundation 320, a "└" shape, a "┌" shape (not shown in the figure), a "┘" shape (not shown in the figure) or a "┐" shape (not shown in the figure) is selected for layout, and the temperature sensing optical fiber 310 is distributed in the four corners, taking into account both the corners and the central area, so that the temperature distribution under uniform symmetry can be measured. Figure 3e A schematic diagram of the optical fiber layout of an asymmetric raft foundation provided in an embodiment of the present application is shown in FIG. Figure 3e As shown, for an irregular shaped raft foundation 320 without any symmetry, a "┼"-shaped layout is adopted, that is, the temperature sensing optical fiber 310 covers the center and diagonal areas in a cross manner to ensure sampling uniformity and maximum temperature distribution feature coverage. In actual construction, the temperature sensing optical fiber 310 can be fixed with a high temperature resistant cable tie, the spacing is controlled within a preset range, and the joint length is reserved for subsequent welding and signal debugging.

[0030] As an optional implementation, the present application embodiment also provides a method for bundling, routing and protecting a temperature sensing optical fiber, which is as follows: Use high-strength cable ties to cross-tie the temperature sensing optical fiber to the vertical and horizontal bars in the steel cage, the upper and lower surfaces and all around. Lead the temperature sensing optical fiber in the steel cage close to the foundation pit wall, fix the remaining optical fiber coil on the ground, use hemp rope to pull it vertically along the wall, and run it along the foundation pit guardrail to the subsequent monitoring end, and tighten it in sections with cable ties throughout the process.

[0031] S102, constructing an optical fiber temperature measurement monitoring system to obtain spatiotemporal temperature data.

[0032] In implementation, after the fiber optic temperature measurement and monitoring system is started, it can automatically scan the temperature of each point of the temperature sensing optical fiber at a preset time interval (such as 1 to 5 minutes), and store the measurement results together with the corresponding position coordinates and timestamp in a pre-established SQL Server database to form spatiotemporal temperature data for subsequent modeling.

[0033] As an optional implementation, the optical fiber temperature measurement and monitoring system includes a temperature sensing optical fiber, a distributed optical fiber temperature sensor and an optical fiber demodulator.

[0034] In implementation, the optical fiber temperature measurement and monitoring system is constructed based on the principles of optical time domain reflectometry and Raman scattering, and continuous temperature monitoring is achieved by demodulating the intensity ratio of Stokes and anti-Stokes scattered light. The temperature sensing optical fiber is laid in the concrete raft foundation structure and is connected to an optical fiber demodulator at one end. The optical fiber demodulator is used to receive and process the scattered signals reflected by the optical fiber, and calculate the temperature values at each position point on the optical fiber in real time. The distributed optical fiber temperature sensor can achieve continuous temperature measurement over the entire length of the optical fiber, and has the advantages of anti-electromagnetic interference, adaptability to the concrete pouring environment, and strong measurement continuity, and is suitable for monitoring the internal temperature field of large-volume concrete structures in complex environments.

[0035] As an alternative implementation, the spatio-temporal temperature data includes time, spatial point coordinates, temperature, and the corresponding relationships among the three.

[0036] In implementation, the computer combines the optical fiber layout path and the initial positioning data, and for each set of temperature readings T measure its corresponding acquisition time t j , the corresponding spatial point coordinates x i on the optical fiber path to generate record entries. Each record is stored in the form of a triple, and the data form can be T measure (x i ,t j ), indicating the temperature value at the spatial point x i at time t j .

[0037] As an alternative implementation, Figure 4 is a flowchart of a method for processing spatio-temporal temperature data provided by an embodiment of the present application. As Figure 4 shown, the following steps are further included after step S102: S401, denoise using a wavelet-Kalman hybrid filtering algorithm.

[0038] In implementation, first, the collected original spatio-temporal temperature data is wavelet decomposed to extract its high-frequency and low-frequency components. The high-frequency part corresponds to random noise or abnormal fluctuations, and the low-frequency part reflects the true temperature change trend. Then, the extended Kalman filter is used to perform dynamic modeling and estimation on the low-frequency part, comprehensively considering the system state transition and measurement error covariance, to achieve smooth estimation of temperature changes and noise suppression. The processing result is the filtered spatio-temporal temperature data.

[0039] S402, in areas where spatio-temporal temperature data is missing or the noise is significant, use interpolation or physical simulation methods for data enhancement.

[0040] In implementation, for local temperature value missing or distorted areas caused by individual time point loss, fiber optic joint failure, abnormal reflection signals, etc. in the spatio-temporal temperature data acquisition, the noise significant areas can be identified through anomaly detection algorithms (such as z-score or moving variance analysis) and marked as areas to be enhanced. Subsequently, select applicable methods according to the data distribution and region type: For boundary continuous regions, use interpolation methods to complete the missing values in the time or space direction. For non-boundary isolated missing regions or temperature mutation regions, construct one-dimensional or two-dimensional finite difference models of heat conduction, set boundary conditions and initial fields, and simulate to generate local temperature data. The enhanced spatio-temporal temperature data can directly replace the missing points in the original data set, or be used as an auxiliary training set to input into the neural network to improve the model's adaptability to extreme scenarios.

[0041] S103. Construct a temperature field prediction model, where the temperature field prediction model is based on a physics-informed neural network and integrates a data loss function, a physical loss function, and an initial condition loss function.

[0042] In implementation, technicians can construct a temperature field prediction model through a computer. The temperature field prediction model is based on a physics-informed neural network and integrates a data loss function, a physical loss function, and an initial condition loss function.

[0043] As an alternative implementation manner, Figure 5 is a flowchart of a method for constructing a temperature field prediction model provided by an embodiment of the present application. As Figure 5 shown, the specific steps for constructing the temperature field prediction model in step S103 are as follows: S501. Set the temperature field prediction model as a model framework based on a physics-informed neural network.

[0044] In implementation, a physics-informed neural network is a modeling method that embeds physical prior knowledge into a deep learning framework. Its core architecture is a deep neural network. By introducing physical constraint terms into the loss function, the model prediction results strictly follow the established physical laws.

[0045] S502. Define the model input variables as spatial point coordinates and time, and the model output variables as the temperature prediction values corresponding to the spatial point coordinates and time.

[0046] In implementation, the measured point position coordinates (such as three-dimensional coordinates) in the concrete raft foundation to be measured and the corresponding acquisition time are jointly used as the model input variables of the temperature field prediction model. The model output variables are the temperature prediction values corresponding to the spatial point coordinates and time. The mapping relationship between the model input variables and the model output variables can be learned and trained end-to-end through a neural network, supporting temperature prediction at any position and any time point.

[0047] S503. Construct a data loss function.

[0048] In implementation, based on the spatio-temporal temperature data obtained from the optical fiber temperature measurement and monitoring system, a data loss function is constructed to constrain the deviation between the output value and the measured value of the temperature field prediction model. The specific form is: L data =(1 / N)∑ i=1 N ∣T NN (x i ,t j )-T measure (x i ,t j )∣ 2 ; Among them, L data is the data loss function, representing the deviation between the output value and the measured value of the temperature field prediction model, T NN (x i ,t j ) is the output value, T measure (x i ,t j ) is the measured value, and N is the total number of measurement points.

[0049] S504. Construct a physical loss function based on the partial differential equation of heat conduction.

[0050] In implementation, a one-dimensional or three-dimensional heat conduction equation can be selected as the model physical constraint, such as the one-dimensional form: ; Among them, α is the thermal diffusion coefficient. Through automatic differentiation, the time derivative and the second spatial derivative of the output value T NN (x,t) of the temperature field prediction model are calculated to construct a physical residual term, and the mean square is calculated as the physical loss function. The specific form is: .

[0051] S505. Construct an initial condition loss function.

[0052] In implementation, using the temperature data at the initial pouring moment of the concrete, an initial condition loss function is defined to limit the output of the model at the initial moment to be consistent with the true initial temperature field. The loss function can also adopt the mean square error form. The specific form is: .

[0053] S104. Based on the spatio-temporal temperature data, perform joint optimization training on the multi-objective loss function of the temperature field prediction model.

[0054] In implementation, technicians can perform joint optimization training of the multi-objective loss function on the temperature field prediction model based on spatio-temporal temperature data. Specifically, the Adam optimizer can be used to iteratively optimize the learnable parameters of the model to minimize the multi-objective loss function. During the training process, the gradient of each loss term is calculated in each iteration, and backpropagation and update are performed on the learnable parameters of the model.

[0055] As an alternative implementation Figure 6 is a flowchart of a training method for a temperature field prediction model provided by an embodiment of the present application. As shown in Figure 6 the specific steps of performing joint optimization training of the multi-objective loss function on the temperature field prediction model based on spatio-temporal temperature data in step S104 are as follows: S601, construct a training data set based on spatio-temporal temperature data.

[0056] In implementation, the preprocessed spatio-temporal temperature data can be extracted from the SQLServer database to construct a training data set. Among them, the spatial point coordinates include the layout positions along the path of the temperature sensing optical fiber cloth, and the time is sampled at a fixed time interval (such as every 5 minutes). When constructing the training data set, normalization processing can be adopted to unify the temperature value, time, and spatial point coordinates within the interval [0, 1] to accelerate the convergence speed of the neural network. The finally generated training data set structure is T measure (x i ,t j ), i = 1, 2…N, j = 1, 2…M.

[0057] S602, construct a multi-objective joint loss function according to the data loss function, physical loss function, initial condition loss function, and preset weights.

[0058] In implementation, the total loss function L can be defined in a weighted combination manner, which consists of three loss functions: the data loss function L data , the physical loss function L physics , and the initial condition loss function L IC . Each loss is weighted by hyperparameters to obtain: L = λ1L data + λ2L physics + λ3L IC ; Among them, the weight values λ1, λ2, and λ3 can be set according to task requirements. For example, when mainly fitting data in the initial stage, take λ1 > λ2, and the weight of the physical constraint term can be gradually increased in the later stage. The weights can be optimized by static setting (such as empirical values) or dynamic adjustment strategies (such as GradNorm).

[0059] S603. During the training process, the gradient descent algorithm is used to iteratively update the learnable parameters of the temperature field prediction model, and the training objective is to minimize the total loss function of the multi-objective joint loss function.

[0060] In implementation, the Adam optimizer is used to iteratively update the learnable parameters (such as weights and biases) in the temperature field prediction model. In each iteration, the total loss function can be calculated based on the current model parameters, and then the automatic backpropagation calculation of its gradient with respect to the model parameters is performed to update the parameter values to minimize the loss.

[0061] S604. In the region where the physical loss function value is greater than the preset physical residual threshold, increase the sampling points of the spatio-temporal temperature data and update the training dataset.

[0062] In implementation, the physical residuals of the model at different positions can be monitored during the training process. When the residuals in some regions are greater than the preset physical residual threshold, it indicates that the model fails to effectively satisfy the heat conduction equation in these regions. To improve physical consistency, the sampling points of the spatio-temporal temperature data can be increased in this region, added to the training dataset and retrained.

[0063] S105. Use the trained temperature field prediction model to predict the temperature values of the concrete raft foundation to be measured at any time and space point coordinates.

[0064] In implementation, the computer can input the position coordinates and time into the trained temperature field prediction model, and the temperature field prediction model automatically outputs the corresponding temperature prediction values.

[0065] As an optional implementation manner, the Bayesian physics-informed neural network or Monte Carlo Dropout can be used to evaluate the prediction confidence.

[0066] S106. Based on the prediction results output by the temperature field prediction model, reconstruct the three-dimensional temperature field of the concrete raft foundation to be measured.

[0067] In implementation, the computer can perform three-dimensional interpolation reconstruction using methods such as support vector regression based on the large number of predicted temperature results output by the temperature field prediction model at three-dimensional grid points. The three-dimensional temperature field can be presented as a temperature isosurface map, a temperature-time animation, or a structural thermal field profile.

[0068] As an optional implementation manner, Figure 7 is a flowchart of a method for reconstructing a three-dimensional temperature field provided by an embodiment of the present application. As Figure 7 shown, the specific steps for reconstructing the three-dimensional temperature field of the concrete raft foundation to be measured based on the prediction results output by the temperature field prediction model in step S106 are as follows: S701. When the number of prediction results output by the temperature field prediction model is less than the preset data scale threshold, the fitrsvm support vector regression model is called through Matlab software for three-dimensional temperature field fitting.

[0069] In implementation, when the number of predicted temperature points output by the PINN model is less than the preset data scale threshold, the fitrsvm (support vector regression) function in the Matlab environment can be used to perform regression fitting on the three-dimensional temperature field. The specific operation is as follows: The prediction results are used as training samples and input into the fitrsvm model. The radial basis kernel function (RBF) is selected to construct a non-linear mapping model for fitting the temperature values in the entire three-dimensional region. The regression output results are used to generate temperature isosurface maps, sectional temperature distribution maps, or three-dimensional animations, thereby achieving high-precision temperature reconstruction in small and medium-scale data scenarios.

[0070] S702. When the number of prediction results output by the temperature field prediction model is greater than the data scale threshold, three-dimensional temperature field reconstruction is performed through the griddata function based on GPU or custom CUDA code.

[0071] In implementation, when the prediction results output by the PINN model exceed the preset data scale threshold (such as 1M), to improve processing efficiency, a GPU parallel computing method can be used for three-dimensional temperature field reconstruction. The specific methods include: calling the GPU-accelerated griddata function in Matlab to perform interpolation calculations on unstructured point cloud data, or using Numba + Cupy, etc. to construct a custom CUDA interpolation kernel on the Python platform and performing large-scale batch interpolation calculations on the temperature field on the GPU.

[0072] As an alternative implementation manner, Figure 8 This is a flowchart of an optimization method for a temperature field prediction model provided by an embodiment of the present application. As Figure 8 shown, the specific steps are as follows: S801. Compare the three-dimensional temperature field of the concrete raft foundation to be measured with the finite element simulation results or the measured data.

[0073] In implementation, the three-dimensional temperature field prediction results output by the temperature field prediction model can be compared one by one with the temperature field simulated by the finite element method or the temperature data measured at the on-site layout points. The comparison area can be the entire model or limited to the thermally sensitive area or the structural key parts (such as the core area, the middle and lower surfaces). The temperature differences are exported in the form of grids to generate an error distribution map.

[0074] S802. Calculate the temperature prediction error based on the preset error evaluation index.

[0075] In implementation, an error calculation method including but not limited to the maximum absolute error, mean square error, and relative error is used to quantitatively evaluate the temperature prediction error. After the error calculation, an error heat map can be generated in combination with a visualization tool to clarify the spatial distribution characteristics of the model prediction error.

[0076] S803, optimize the temperature field prediction model based on the distribution result of the temperature prediction error.

[0077] In implementation, the temperature field prediction model can be optimized in the following ways. For example, additional training points or local interpolation points can be added in high-error regions (such as boundaries and the junction of thermal cores). Or dynamically adjust the weights of the physical loss and data loss in the multi-objective loss function to enhance physical consistency. Or increase the number of neural network layers or width and adopt local weighted training. Or within the error-concentrated feature region, freeze the model parameters in other regions and only locally fine-tune the sub-model in the high-error region to improve the local prediction accuracy of the model.

[0078] As an alternative implementation manner, the embodiment of the present application also provides a method for reconstructing a three-dimensional temperature field through an indoor similarity simulation experiment to verify that the temperature sensing optical fiber layout method provided by the embodiment of the present application can monitor the temperatures of the upper and lower surfaces and the middle core region of a large-volume concrete raft foundation without adding a large number of measurement points, specifically as follows: According to the method of technicians for laying temperature sensing optical fibers on the surface and inside of the concrete raft foundation to be measured, an indoor similarity simulation experiment is carried out. The experiment is carried out in the laboratory, and the concrete pouring model is a cuboid pouring model, and the model size is: b×h×l = 120mm×200mm×2000mm; Lay the temperature sensing optical fiber along the inner side of the steel cage in an "S" shape, pre-tension the optical fiber, tie and fix the optical fiber with a cable tie every other stirrup. After the temperature sensing optical fiber layout is completed, splice the jumper wire, connect it to the optical fiber demodulator for debugging, then pour the concrete, and conduct temperature monitoring for n days.

[0079] Through finite element analysis, simulate the hydration heat temperature field of the large-volume concrete raft foundation, and then compare the monitoring results of the indoor similarity simulation experiment with the finite element analysis results. The comparison indicators can include the curve trend and the error range. Figure 9 A temperature monitoring comparison chart provided by the embodiment of the present application, and the comparison result is as Figure 9 shown. It can be concluded that by laying the temperature sensing optical fiber according to the scheme provided by the embodiment of the present application, the temperatures of the upper and lower surfaces and the middle core region of the large-volume concrete raft foundation can be accurately monitored.

[0080] The embodiment of the present application provides a method for reconstructing the three-dimensional temperature field of a large-volume concrete raft foundation. By laying temperature-sensing optical fibers in multiple temperature-measuring areas of the concrete raft foundation to be measured, combined with the structural characteristics of the raft foundation, the optical fiber paths are scientifically laid out on the surface and core areas to achieve full coverage of the key thermal change areas. A fiber-optic temperature measurement and monitoring system is constructed, and distributed fiber-optic sensing technology is used to obtain spatio-temporal temperature data covering the entire structure, thus solving the problems of sparse distribution of traditional thermocouples and insufficient spatial continuity. A temperature field prediction model based on a physics-informed neural network is constructed. The model integrates three types of constraints: a data loss function, a physical loss function, and an initial condition loss function, which not only ensures high-precision fitting of the measured temperature data, but also ensures that the prediction results satisfy the physical laws described by the partial differential equation of heat conduction. At the same time, the initial deviation of the solution is restricted starting from the initial temperature field, thereby enhancing the physical consistency and stability of the model. In the model training stage, a multi-objective loss function joint optimization strategy is adopted. By introducing three independent objectives: the measured data error, the physical residual, and the initial state error, comprehensive constraints are imposed, and adaptive sampling is implemented in areas with large physical residuals to increase the training points, improving the prediction accuracy of the model in key thermal mutation areas and solving the problems of inaccurate prediction, overfitting, or physical mismatch in traditional methods in areas with drastic changes in thermal gradients. The trained neural network model is used to predict the temperature at any spatio-temporal point coordinates, realizing the complete reconstruction of the three-dimensional temperature field of the concrete raft foundation to be measured. In summary, the embodiment of the present application can achieve high-precision reconstruction of the three-dimensional temperature field of a large-volume concrete raft foundation without adding a large number of measurement points, overcoming the problems of the inability to obtain a continuous temperature field by local measurement of traditional thermocouples, the lack of physical consistency in traditional interpolation or simulation models, and the weak generalization ability of data-driven models.

[0081] It should be understood that although Figure 1 , Figure 2 , Figures 4 to 8 the steps in the flowchart are shown sequentially in the order indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise clearly stated in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, Figure 1 , Figure 2 , Figures 4 to 8 at least a part of the steps in

[0082] It can be understood that the same / similar parts among the various embodiments of the above methods in this specification can be referred to each other. Each embodiment focuses on the differences from other embodiments. For the relevant parts, refer to the descriptions of other method embodiments.

[0083] In one embodiment, a computer device is provided. As Figure 10 shown, it includes a memory and a processor. A computer program that can run on the processor is stored on the memory. When the processor executes the computer program, the method steps for reconstructing the three-dimensional temperature field of the large-volume concrete raft foundation are implemented.

[0084] In one embodiment, a computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the method for reconstructing the three-dimensional temperature field of the large-volume concrete raft foundation are implemented.

[0085] Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in this application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0086] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the existence of additional identical elements in the process, method, article or device comprising the said element.

[0087] It should also be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0088] Each embodiment in this specification is described in a related manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and reference can be made to the corresponding parts of the method embodiments for the relevant content.

[0089] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as within the scope described in this specification.

[0090] The above-described embodiments merely represent several implementation manners of this application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of this application patent shall be subject to the appended claims.

Claims

1. A three-dimensional temperature field reconstruction method for a large-volume concrete raft foundation, characterized in that, The method includes: Laying temperature sensing optical fibers in multiple temperature measurement areas of the concrete raft foundation to be measured; Constructing an optical fiber temperature measurement monitoring system to obtain spatio-temporal temperature data; Constructing a temperature field prediction model, which is based on a physics-informed neural network and integrates a data loss function, a physical loss function, and an initial condition loss function; Based on the spatio-temporal temperature data, performing joint optimization training on the multi-objective loss function of the temperature field prediction model; Using the trained temperature field prediction model to predict the temperature values at any time and spatial point coordinates of the concrete raft foundation to be measured; Based on the prediction results output by the temperature field prediction model, reconstructing the three-dimensional temperature field of the concrete raft foundation to be measured.

2. The method according to claim 1, wherein The laying of the temperature sensing optical fibers in multiple temperature measurement areas of the concrete raft foundation to be measured includes: According to the surface geometric shape of the concrete raft foundation to be measured, determining the laying form of the temperature sensing optical fibers on the surface of the concrete raft foundation to be measured; Laying the temperature sensing optical fibers in the internal geometric center area of the concrete raft foundation to be measured.

3. The method according to claim 2, characterized in that, The determining of the laying form of the temperature sensing optical fibers on the surface of the concrete raft foundation to be measured according to the surface geometric shape of the concrete raft foundation to be measured includes: If the surface geometric shape of the concrete raft foundation to be measured is only symmetric up and down, the surface laying form is "┴" or "┬"; If the surface geometric shape of the concrete raft foundation to be measured is only symmetric left and right, the surface laying form is "├" or "┤"; If the surface geometric shape of the concrete raft foundation to be measured is centrosymmetric, the surface laying form is "└" or "┌" or "┐" or "┘"; If the surface geometric shape of the concrete raft foundation to be measured is asymmetric, the surface laying form is "┼".

4. The method according to claim 1, wherein The optical fiber temperature measurement monitoring system includes a temperature sensing optical fiber, a distributed optical fiber temperature sensor, and an optical fiber demodulator.

5. The method according to claim 1, wherein The spatio-temporal temperature data includes time, spatial point coordinates, temperature, and their corresponding relationships.

6. The method according to claim 1, wherein The constructing of the temperature field prediction model includes: Setting the temperature field prediction model as a model framework based on a physics-informed neural network; Defining the model input variables as spatial point coordinates and time, and the model output variable as the temperature prediction value corresponding to the spatial point coordinates and the time; Constructing the data loss function; Constructing the physical loss function based on the partial differential equation of heat conduction; Constructing the initial condition loss function.

7. The method according to claim 1, characterized in that, The performing of joint optimization training on the multi-objective loss function of the temperature field prediction model based on the spatio-temporal temperature data includes: Based on the spatio-temporal temperature data, constructing a training data set; Constructing a multi-objective joint loss function according to the data loss function, the physical loss function, the initial condition loss function, and a preset weight; During the training process, using the gradient descent algorithm to iteratively update the learnable parameters of the temperature field prediction model, and the training objective is to minimize the total loss function of the multi-objective joint loss function; In the region where the physical loss function value is greater than the preset physical residual threshold, increase the sampling points of the spatio-temporal temperature data and update the training dataset.

8. The method according to claim 1, characterized in that, After constructing the optical fiber temperature measurement and monitoring system and obtaining the spatio-temporal temperature data, the method further includes: Using the wavelet-Kalman hybrid filtering algorithm for denoising; In the region where the spatio-temporal temperature data is missing or the noise is significant, use the interpolation method or the physical simulation method for data enhancement.

9. The method according to claim 1, characterized in that, Reconstructing the three-dimensional temperature field of the concrete raft foundation to be measured based on the prediction result output by the temperature field prediction model, including: When the number of prediction results output by the temperature field prediction model is less than the preset data scale threshold, call the fitrsvm support vector regression model through Matlab software for three-dimensional temperature field fitting; When the number of prediction results output by the temperature field prediction model is greater than the data scale threshold, reconstruct the three-dimensional temperature field through the griddata function based on GPU or custom CUDA code.

10. The method according to claim 1, characterized in that The method further includes: Comparing the three-dimensional temperature field of the concrete raft foundation to be measured with the finite element simulation result or the measured data; Calculating the temperature prediction error based on the preset error evaluation index; Optimizing the temperature field prediction model based on the distribution result of the temperature prediction error.

Images