Optimization control method for construction joint spacing of mass concrete based on finite element simulation

Abstract

This invention relates to the field of electronic digital data processing technology and discloses a method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation. The method includes: acquiring real-time environmental state information; constructing a strategy generation module based on a Gaussian process to generate candidate values ​​for construction joint spacing through a data acquisition function; establishing a finite element verification module to perform thermo-mechanical coupling simulation on the candidate values ​​and introduce a model parameter self-calibration mechanism to calculate a multi-objective reward vector including crack resistance, efficiency, and cost; and using this reward feedback to update the surrogate model of the strategy generation module. By iteratively executing the above generation and verification steps until a termination condition is met, the optimal construction joint spacing and auxiliary decision-making information are finally provided by the decision output module. This invention, by constructing a data-driven closed-loop optimization framework, achieves dynamic, intelligent, and optimized decision-making for construction joint spacing, thereby improving the quality and efficiency of engineering construction.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology, specifically to a method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation. Background Technology

[0002] Currently, large-volume concrete structures are widely used in water conservancy and hydropower, bridge engineering, and large building foundations. Due to the heat of hydration of cement, these structures generate enormous internal temperature stress. If this stress exceeds the tensile strength of concrete at the same age, temperature cracks will occur. These cracks not only affect the aesthetics of the structure but also threaten its durability and safety. Therefore, the rational design of construction joints and segmented pouring are among the most crucial technical means to control temperature stress and prevent cracks. The decision regarding the spacing of construction joints directly determines the crack control effect, construction cost, and construction period of the project, making it a key technical aspect of the entire project management process.

[0003] Regarding the aforementioned issues, the engineering community has developed several common methods for determining construction joint spacing. One method is the standard guidance method, where technicians, based on industry standards such as the "Code for Construction of Hydraulic Concrete," and considering the project scale and concrete grade, consult recommended value ranges to initially determine the spacing. Another method is the experience-based decision-making method, where senior engineers, based on successful cases of similar past projects, adjust the standard values ​​according to their personal experience to provide a decision spacing. Additionally, there is the method using forward finite element analysis, where engineers preset a set of material and environmental parameters, perform thermo-mechanical coupling simulations on one or a few pre-selected spacing schemes, and assess the crack resistance risk of the schemes by analyzing the stress results.

[0004] Existing technologies have several shortcomings in practical applications. Standardized and empirical methods are essentially static decision-making processes, providing fixed spacing values ​​that cannot respond to dynamically changing environmental parameters at the construction site, such as sudden temperature drops or batch fluctuations in raw materials. This leads to significant uncertainty in the decision-making results. Existing methods struggle to handle conflicts between multiple objectives. To ensure foolproof crack prevention, excessively dense construction joints are chosen, increasing construction costs and time. Overemphasizing economy amplifies the risk of cracking. Existing technologies lack a quantitative model to systematically balance these interdependent objectives. Even with finite element simulation, its accuracy is significantly challenged. The material parameters relied upon in the simulation are typically ideal values ​​from the laboratory, deviating from actual field conditions. Once established, the model lacks a closed-loop mechanism for back-calibrating and correcting its parameters using field monitoring data, resulting in a disconnect between simulation results and engineering reality, greatly diminishing its guiding value.

[0005] Therefore, this invention provides a method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation, in order to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an optimization and control method for the spacing of construction joints in large-volume concrete based on finite element simulation. This method solves the problems in existing technologies, such as reliance on experience for determining the spacing of construction joints in large-volume concrete, low efficiency of the optimization process, difficulty in balancing multiple engineering objectives, and insufficient accuracy of decisions due to deviations between the finite element simulation model and actual construction conditions.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation, comprising the following steps:

[0008] S1. Obtaining Environmental Status Information: This involves acquiring environmental status information during the construction of large-volume concrete structures. This information includes material property parameters, environmental parameters, construction progress parameters, real-time monitoring data, and structural geometric parameters. This information collectively forms the input basis for optimization decisions.

[0009] S2. Generate candidate values ​​for construction joint spacing: Based on the environmental state information obtained in step S1, one or more candidate values ​​for construction joint spacing to be evaluated are generated through a strategy generator. The strategy generator is a decision module built based on a specific algorithm, whose function is to output a decision scheme to be evaluated based on the input environmental state information.

[0010] S3. Perform finite element self-calibration verification: Perform finite element self-calibration verification on the candidate construction joint spacing values ​​generated in step S2 to evaluate the merits of each candidate value. This step is the core of this method and specifically includes the following sub-steps:

[0011] S3.1 Perform finite element simulation: Based on the currently acquired material property parameters and the candidate values ​​of the construction joint spacing, perform a thermo-mechanical coupling simulation of the large-volume concrete. This simulation is used to predict the evolution of the physical field inside the large-volume concrete under a specific construction joint spacing, and finally obtain the predicted temperature field data and stress-strain field data of the large-volume concrete.

[0012] S3.2, Perform model parameter self-calibration: Acquire real-time monitoring data from the construction site, and based on the deviation between the real-time monitoring data and the predicted temperature and stress-strain fields obtained in step S3.1, iteratively adjust the material property parameters used in the finite element simulation to minimize the deviation. This step establishes a feedback loop from the actual working conditions to the simulation model, aiming to adjust the parameter set within the finite element model. To minimize the deviation between predicted and monitored data. The optimization objective can be expressed as:

[0013] ;

[0014] in, It is the optimized set of material property parameters; The operator for minimizing parameters.

[0015] This is the deviation function, used to quantify the difference between predicted and monitored data. The deviation function can be specifically defined as:

[0016] ;

[0017] in, A vector representing real-time monitoring data; Indicates the current parameter set The predicted data vector output by the finite element model; This represents the norm used to quantify vector differences.

[0018] S3.3 Calculate the multi-objective reward: Based on the finite element simulation results after model parameter self-calibration in step S3.2, calculate the multi-objective reward corresponding to the candidate construction joint spacing value. The multi-objective reward is a set of indicators used to quantitatively evaluate the performance of the current candidate construction joint spacing value across multiple engineering objective dimensions.

[0019] S4. Update the strategy generator: Based on the candidate values ​​for construction joint spacing and the multi-objective reward calculated in step S3.3, update the optimization model within the strategy generator. This step enables the strategy generator to learn from the evaluation results, thereby generating better decisions in subsequent iterations.

[0020] S5. Output optimized control decision: Based on the updated strategy generator, output the optimal construction joint spacing optimization control decision to guide on-site construction.

[0021] In one specific embodiment, the step of obtaining the material property parameters in step S1 specifically includes: obtaining the hydration heat release curve parameters of the current batch of concrete, the development law of elastic modulus with age, the tensile strength development curve, and the compensation shrinkage rate and temperature rise suppression effect parameters of the high-performance expansion agent as the material property parameters.

[0022] Preferably, in step S3.3, the step of calculating the multi-objective reward specifically includes: calculating the crack resistance performance target reward based on the maximum tensile stress or maximum tensile strain predicted by the finite element simulation; calculating the construction efficiency target reward based on the size of the candidate value of the construction joint spacing; and calculating the economic cost target reward based on the changes in labor and material costs caused by the change in the number of construction joints.

[0023] In one specific embodiment, the strategy generator in step S2 uses a Gaussian process or a Bayesian optimization algorithm to generate candidate values ​​for the construction joint spacing based on the environmental state information and the optimization model already built inside the strategy generator.

[0024] Furthermore, the thermo-mechanical coupling simulation step of the large-volume concrete in step S3.1 specifically includes: simulating the evolution of the temperature field, stress field and strain field of the large-volume concrete from the start of pouring to the temperature peak subsidence and shrinkage stabilization. The physical model of the evolution process considers the heat of hydration, temperature gradient, drying shrinkage, autogenous shrinkage, creep and the compensating shrinkage effect of the expansion agent.

[0025] Preferably, the deviation in step S3.2 is calculated as the error norm between the real-time monitoring data and the predicted temperature field data and the predicted stress-strain field data; the iterative adjustment uses gradient descent or Bayesian inversion methods to optimize and adjust the material property parameters used in the finite element simulation so as to minimize the error norm.

[0026] In one specific embodiment, the material property parameters to be iteratively adjusted in step S3.2 include the hydration heat release coefficient, the elastic modulus development curve parameter, and the actual expansion rate coefficient of the high-performance expansion agent.

[0027] In an optional implementation, the method further includes adding a surrogate model acceleration step between step S2 and step S3, specifically: constructing a surrogate model based on Gaussian process regression to quickly predict the multi-objective reward corresponding to the candidate construction joint spacing; the policy generator interacts with the surrogate model first to filter out potential candidate construction joint spacing values; the potential candidate construction joint spacing values ​​are then accurately evaluated by calling the finite element self-calibration verification module, and the evaluation results are used to update the surrogate model.

[0028] Furthermore, the step of updating the optimization model inside the strategy generator in step S4 specifically includes: the strategy generator continuously learns and improves its internal optimization model based on the candidate values ​​of the construction joint spacing and the historical data of the multi-objective reward.

[0029] Preferably, the step S5 of outputting the optimal construction joint spacing optimization control decision specifically includes: outputting one or a set of optimal construction joint spacings, and simultaneously providing key environmental state parameters and their weight contributions that support the optimization control decision as auxiliary decision information.

[0030] By introducing a parameter self-calibration mechanism for the finite element model, a closed-loop feedback between real-time on-site monitoring data and the simulation physical model is established. This allows for dynamic adjustment of the material property parameters upon which the simulation model relies based on actual working conditions, improving the accuracy of the finite element simulation results and its ability to reflect actual working conditions. Furthermore, it enables the strategy generator to learn and make decisions based on a more reliable physical model, thereby achieving dynamic adaptive optimization of construction joint spacing decisions. This method can output optimal control decisions while balancing multiple objectives such as crack resistance, construction efficiency, and economic cost, providing a new technical approach for the scientific and refined construction control of large-volume concrete.

[0031] This invention provides a method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation. It offers the following advantages:

[0032] 1. This invention improves the accuracy of simulation results and its ability to reflect actual working conditions by introducing a parameter self-calibration mechanism for the finite element model. By comparing real-time on-site monitoring data with the predicted data from the finite element simulation, and iteratively adjusting the material property parameters within the model based on the deviation between the two, the simulation model can dynamically and continuously approximate the actual engineering state. This overcomes the technical shortcomings of traditional simulation methods that rely on fixed theoretical parameters and are disconnected from actual working conditions, providing a more reliable physical basis for subsequent optimization decisions.

[0033] 2. This invention combines a strategy generator with a self-calibrating finite element verification environment to form an intelligent learning closed loop. The strategy generator not only makes decisions based on changes in environmental conditions but also learns from an increasingly accurate simulation environment that continuously self-corrects using field data. This allows its generated optimized control decisions to be highly adaptable to uncertainties such as material batch fluctuations and environmental changes, improving decision quality and achieving dynamic adaptive optimization of construction joint spacing decisions, thus enhancing the scientific rigor and robustness of the decisions.

[0034] 3. This invention calculates multi-objective rewards encompassing multiple dimensions such as crack resistance, construction efficiency, and economic cost. This method can quantify and handle these conflicting engineering objectives. This ensures that the final optimized control decision no longer focuses solely on a locally optimal solution for a single performance indicator, but rather on a comprehensive, balanced solution that considers multiple aspects such as safety, schedule, and cost, providing scientific and refined decision support for engineering projects. It enables multi-objective collaborative optimization of construction joint spacing, achieving a comprehensive balance of engineering benefits. Attached Figure Description

[0035] Figure 1 This is an overall flowchart of the method of the present invention;

[0036] Figure 2This is a schematic diagram of the strategy generation workflow of the present invention;

[0037] Figure 3 This is a schematic diagram of the model parameter self-calibration process of the present invention;

[0038] Figure 4 This is a schematic diagram illustrating the multi-objective reward structure of an embodiment of the present invention;

[0039] Figure 5 This is a schematic diagram of the optimization process based on the agent model acceleration of the present invention. Detailed Implementation

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Please see the appendix Figure 1 This invention provides a method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation. This method can be executed by a computer device. The computer device includes a processor and a memory. The memory stores a computer program, and when the processor executes the computer program, it implements the method of this invention. The method may include the following steps:

[0042] First, in step S1, the information acquisition module acquires environmental status information during the construction of large-volume concrete. Environmental status information It is a dataset that includes: material property parameters, environmental parameters, construction progress parameters, real-time monitoring data collected from on-site sensors, and structural geometric parameters of the concrete structure to be poured.

[0043] Subsequently, in step S2, the strategy generation module receives the environmental status information provided by the information acquisition module. Based on environmental status information In addition, the optimization model within the strategy generation module generates one or more candidate values ​​for construction joint spacing to be evaluated. .

[0044] Next, in step S3, the finite element verification module verifies each candidate value of the construction joint spacing generated by the strategy generation module. Perform finite element self-calibration verification. This step is a core part of the method and is specifically broken down as follows:

[0045] In sub-step S3.1, the finite element verification module verifies the material properties based on the currently acquired parameters. Candidate values ​​for construction joint spacing A thermo-mechanical coupled finite element simulation of a large-volume concrete structure was performed. The simulation outputs predicted temperature and stress-strain fields of the large-volume concrete under given conditions. This dataset is denoted as [data set name missing]. .

[0046] In sub-step S3.2, the finite element verification module performs model parameter self-calibration. This module acquires real-time monitoring data collected by the information acquisition module. And calculate real-time monitoring data. With the prediction dataset Deviation between Subsequently, through a reverse optimization process, the material property parameters used in the finite element simulation were adjusted. Perform iterative adjustments to minimize the deviation. The optimization objective can be described by the following formula:

[0047] ;

[0048] in: It is the optimized set of material property parameters; The operator for minimizing parameters.

[0049] This is the deviation function, used to quantify the difference between predicted and monitored data. The deviation function can be specifically defined as:

[0050] ;

[0051] in, A vector representing real-time monitoring data; Indicates the current parameter set The predicted data vector output by the finite element model; This represents the norm used to quantify vector differences.

[0052] In sub-step S3.3, the finite element verification module calculates candidate values ​​for construction joint spacing based on the finite element simulation results after self-calibration of the model parameters in sub-step S3.2. The corresponding multi-objective reward. This multi-objective reward is a set of indicators used to quantitatively evaluate the performance of the current candidate construction joint spacing value across multiple engineering objectives.

[0053] Then, in step S4, the candidate values ​​for construction joint spacing are... The data pair, consisting of the multi-objective reward calculated by the finite element verification module, is fed back to the strategy generation module. The strategy generation module uses this data pair to update its internal optimization model.

[0054] Finally, in step S5, the decision output module outputs the optimal construction joint spacing control decision based on the updated optimization model from the strategy generation module. This decision can be used to guide large-volume concrete construction on site.

[0055] The information acquisition module, strategy generation module, finite element verification module, and decision output module are functionally divided modules. They can be implemented as program instructions stored in the computer device's memory, called and executed by the processor. By repeatedly executing the iterative process of steps S2 to S4, the optimization model of the strategy generation module is continuously improved until the preset convergence condition is met or the maximum number of iterations is reached, finally outputting the optimization result.

[0056] In step S1, the environmental state information is obtained. It forms the data foundation for performing subsequent optimization steps, and its specific composition and acquisition method are as follows.

[0057] Environmental status information includes material property parameters. In one specific embodiment, obtaining the material property parameters includes the following:

[0058] Obtain the hydration heat release curve parameters for the current batch of concrete. These parameters define the generation rate and total amount of heat sources in the finite element simulation. They can be expressed as a function describing the change in the hydration heat release rate over time or equivalent age, such as a double exponential function. Key parameters include the total adiabatic temperature rise and the reaction rate constant, which are obtained through adiabatic or semi-adiabatic thermal tests on the same batch of cement used in the current project.

[0059] The parameters for obtaining the elastic modulus of concrete over time are determined. These parameters define the evolution of the mechanical properties of concrete over time. This evolution can be described by a functional model, such as a function based on the CEB-FIP model, whose input parameters include the elastic modulus of concrete at 28 days and coefficients related to cement type. These parameters are calibrated by conducting compressive strength and elastic modulus tests on specimens made from the current batch of concrete at different ages.

[0060] The parameters for obtaining the tensile strength development curve are obtained. Similar to the development law of the elastic modulus, these parameters define the functional relationship between the tensile strength of concrete and its age, serving as a direct basis for crack risk assessment. These parameters are determined by conducting splitting tensile tests on concrete specimens at different ages.

[0061] The parameters for the compensation shrinkage rate and temperature rise inhibition effect of the high-performance expansive agent were obtained. The compensation shrinkage rate parameter defines the development of the expansion strain generated by the expansive agent under constrained conditions over time, and this parameter was obtained through constrained expansion rate tests on concrete. The temperature rise inhibition effect parameter was used to correct the hydration heat release curve and quantify the retarding effect of the expansive agent on the early hydration reaction rate; this parameter was also determined by calorimetry.

[0062] Environmental status information also includes environmental parameters. These parameters include real-time and predicted atmospheric temperature, relative humidity, wind speed, and solar radiation intensity at the construction site. These parameters are collected in real-time by meteorological stations deployed at the construction site, and forecast data for the future period is obtained through access to public meteorological data services. These parameters are used as thermal boundary conditions in subsequent finite element thermo-mechanical coupling analysis to calculate convective and radiative heat transfer between the concrete structure and the external environment.

[0063] Environmental status information also includes construction progress parameters. These parameters include the geometric model of the completed pouring sections, the actual pouring completion time of each block, and the specific locations of the construction joints. This information can be directly read from Building Information Modeling (BIM) or construction management information systems to determine the initial state and computational domain for subsequent finite element simulations.

[0064] Environmental status information also includes real-time monitoring data. This real-time monitoring data is time-series data collected through a sensor network pre-embedded at specific locations within the already cast large-volume concrete structure. The sensor network includes several temperature sensors (e.g., thermocouples or fiber Bragg grating temperature sensors) and strain sensors (e.g., vibrating wire strain gauges). The sensor locations are selected in areas with significant temperature gradients and stress concentrations, such as the structural geometric center, surfaces, and corners. The collected temperature and strain history data serve as the baseline true values ​​for the model parameter self-calibration performed in step S3.2.

[0065] Environmental status information also includes structural geometric parameters. These parameters are the detailed three-dimensional geometric dimensions of the concrete section to be evaluated and poured, including its length, width, thickness, and the reinforcement ratio and distribution of internal steel bars. These parameters are extracted from design drawings or BIM models and used to generate the computational mesh model for subsequent finite element simulations. The information acquisition module integrates all these parameters into a standardized data structure: environmental status information. It is then output to the policy generation module.

[0066] See attached document Figure 2 In step S2, the policy generation module generates the policy based on the received environmental state information. The process generates one or more candidate values ​​for construction joint spacing to be evaluated. The purpose of this step is to identify, in a computationally efficient manner, candidate values ​​with the greatest optimization potential within a broad decision space, and submit them to the subsequent computationally expensive finite element verification module for evaluation.

[0067] In one specific embodiment, the strategy generation module employs a Bayesian optimization algorithm. The core of this algorithm is maintaining an internal optimization model, which in this embodiment is a probabilistic surrogate model based on a Gaussian process. This probabilistic surrogate model is used to approximate the construction joint spacing. With the final multi-objective reward The complex nonlinear relationship between them.

[0068] The probabilistic surrogate model based on Gaussian processes is constructed using historical evaluation data points (i.e., data pairs consisting of candidate construction joint spacing values ​​evaluated in each iteration and their corresponding multi-objective rewards). For any construction joint spacing that has not been evaluated... This model can not only output a predicted mean of multi-objective rewards, but also... It can also output a measure of uncertainty about the prediction, namely the prediction variance. .

[0069] The policy generation module uses a sampling function to utilize the predicted mean output by the probabilistic surrogate model. and prediction variance This process determines the next most valuable candidate value for evaluating construction joint spacing. The acquisition function balances utilization and exploration. Utilization refers to searching in areas where the current model considers the reward high, while exploration refers to searching in areas where the model has high uncertainty.

[0070] In a specific implementation, the data acquisition function can be an upper confidence interval function. This function has the following form:

[0071] ;

[0072] in: Construction joint spacing The corresponding acquisition function value; The Gaussian process model predicts the construction joint spacing as... Average reward at that time; The Gaussian process model predicts the construction joint spacing as... Standard deviation of rewards at that time; It is an adjustable hyperparameter used to control the balance between exploitation and exploration.

[0073] The specific workflow of the strategy generation module is as follows:

[0074] First, at the start of each iteration, the policy generation module uses the new data pairs fed back from the previous step S4 to update the posterior distribution of its internal probabilistic surrogate model based on Gaussian processes.

[0075] Secondly, the strategy generation module, within a preset range of construction joint spacing values, performs data acquisition on the data acquisition function. Perform optimization to find the solution.

[0076] Finally, the strategy generation module will enable the acquisition function The construction joint spacing that achieves the maximum value is used as the optimal candidate value generated in this iteration. and the candidate value Output to the finite element verification module.

[0077] In step S3, sub-step S3.1, the finite element verification module uses the material property parameters, structural geometric parameters, and environmental parameters obtained from step S1, as well as the candidate values ​​of construction joint spacing received from step S2, to perform the verification. A thermo-mechanical coupled finite element simulation of a large-volume concrete structure was constructed and executed. The purpose of the simulation was to predict the evolution of the internal physical field of the concrete structure from the start of pouring until shrinkage stabilization under the given conditions.

[0078] First, based on structural geometric parameters and candidate values ​​for construction joint spacing. A three-dimensional finite element mesh model is generated. Subsequently, a transient thermo-mechanical coupling analysis is performed. This analysis includes a temperature field analysis and a stress field analysis, which are coupled through the time correlation between temperature and material properties.

[0079] In temperature field analysis, unsteady-state heat conduction differential equations are solved to determine the temperature at each point within the structure as a function of time. The change in . The equation can be expressed as:

[0080] ;

[0081] in, This refers to the density of concrete. The specific heat of concrete; The sign of the partial derivative; It is a temperature field function; For time; For gradient operators; It is a divergence operator; The thermal conductivity of concrete; This refers to the heat source for hydration within the concrete. The generation rate is determined by the hydration heat release curve parameters obtained in step S1, which define the variation of the hydration reaction rate with the equivalent age. The rate of increase in internal energy per unit volume of concrete; The rate of heat flowing into this unit volume via heat conduction.

[0082] The boundary conditions for the temperature field analysis are set based on the environmental parameters obtained in step S1. On the surface of concrete in contact with air, the combined effects of convective and radiative heat transfer are considered. On the surface in contact with the formwork or existing concrete, the corresponding heat flux boundary is set according to the thermal conductivity of the contact material.

[0083] In stress field analysis, the nodal temperatures calculated at each time step in temperature field analysis are applied as thermal loads to the model. Total strain The algebraic sum of the components is decomposed as follows:

[0084] ;

[0085] in, The elastic strain is related to the stress and follows Hooke's law. The elastic modulus used in the calculation is updated in real time according to the elastic modulus obtained in step S1 as it develops with age. Temperature strain, caused by temperature changes, is calculated using the following formula: ,in, The coefficient of thermal expansion is... This is the reference temperature for stress calculation; As defined in claim 5, the shrinkage strain is further decomposed into self-shrinkage strain caused by the cement hydration reaction itself and drying shrinkage strain caused by moisture evaporation due to internal and external humidity gradients. The development models for both shrinkage strains are determined by the corresponding parameters obtained in step S1. Creep strain is the strain component that evolves over time under sustained stress. Its calculation employs a creep model based on age and loading history, with the specific parameters of the model calibrated using the experimental data obtained in step S1. For expansion strain, this component is used to simulate the compensating shrinkage effect of the high-performance expansion agent, and its development law over time is determined by the restricted expansion rate test data obtained in step S1.

[0086] The solution process for thermo-mechanical coupling simulation employs the incremental time step method. Within each time increment step, the temperature field equation is first solved to obtain the temperature distribution at that moment; then, the temperature distribution is used as a heat load, and the material mechanical parameters (such as elastic modulus and tensile strength) at the current age are updated, thereby solving the mechanical equilibrium equations to obtain the stress and strain fields at that moment. This process begins with the simulated casting ( From that moment, iterative calculations continue until the preset analysis termination time is reached, such as the age when the temperature peak completely subsides and the contraction tends to stabilize.

[0087] The final output of the simulation is a set of predicted temperature, stress, and strain fields for all nodes and elements in the structural model, covering the entire analysis timeline. This complete set of time-series data constitutes the prediction dataset. It is then passed to sub-step S3.2 for performing subsequent model parameter self-calibration.

[0088] See attached document Figure 3 In sub-step S3.2, the finite element verification module performs model parameter self-calibration. The purpose of this step is to utilize real-time monitoring data collected from the construction site through a closed-loop feedback mechanism. For the uncertain material property parameters upon which the finite element simulation model depends Inverse solving and optimization are performed to eliminate the deviation between the simulation model and the actual physical conditions.

[0089] First, calculate the deviation. Deviation It is the set of predicted data output by the quantization sub-step S3.1 With real-time monitoring data The scalar value representing the difference between them. Specifically, from real-time monitoring data. In the process, the monitoring values ​​of each sensor in the time series are extracted, for example, the first... A temperature sensor in Temperature value at time Meanwhile, from the prediction dataset In the process, extract the elements from the finite element model that are related to the first element. Predicted values ​​of nodes corresponding to the physical locations of each sensor in the same time series .

[0090] The deviation is calculated using the error norm. In a specific embodiment, this error norm can be calculated using the root mean square error. For example, for the deviation of the temperature field... The calculation formula is as follows:

[0091] ;

[0092] in, This is the total number of sensors used for calibration. This represents the total number of data points in the time series. Similarly, the deviation of the strain field can be calculated. Total deviation It is a weighted sum of the deviations of each physical field:

[0093] ;

[0094] in, and These are preset weighting coefficients.

[0095] Secondly, iterative parameter adjustments are performed. This process employs an optimization algorithm to minimize the total deviation. The objective is to achieve this. In one specific embodiment, the gradient descent method is employed. This method iteratively updates the set of material property parameters. To achieve optimization, the update rules are as follows:

[0096] ;

[0097] in, It is the parameter set for the next iteration. It is the parameter set of the current iteration. It is the learning rate, used to control the step size of each update. It is a deviation function Regarding the parameter set in parameter space The gradient. This gradient points to the deviation. The direction of fastest growth is the direction in which the parameter is updated, so updating the parameter in its negative direction can reduce the deviation.

[0098] gradient It is a vector whose components are deviation functions. For each material parameter ( partial derivatives of ) .because and The relationship between them is established through a complete finite element simulation, and since their analytical expression is difficult to obtain, the partial derivative is approximated using numerical methods (e.g., the finite difference method). At that time, for parameters Apply a small perturbation Re-running the finite element simulation yields new deviations. ,in Except Become external and Given the same set of parameters, the partial derivatives can be approximated as:

[0099] ;

[0100] By analyzing the parameter set By performing the above operation on each parameter, the complete gradient vector can be obtained. .

[0101] This iterative adjustment process continues until a preset convergence condition is met, such as the gradient magnitude being less than a threshold, or the deviation being... The decrease is less than a threshold, or the preset maximum number of iterations has been reached.

[0102] In one specific embodiment, the material property parameters selected for iterative adjustment The key parameters include: the hydration heat release coefficient, which directly determines the peak value and shape of the temperature field; the elastic modulus development curve parameter, which is a crucial basis for calculating thermal stress and shrinkage stress; and the actual expansion coefficient of the high-performance expansive agent, which directly affects the compensation effect on shrinkage strain. These parameters were chosen because they have the most significant impact on the simulation results of the temperature and stress fields, and their values ​​are most easily affected by the batch of raw materials and the on-site construction environment, leading to uncertainties.

[0103] The final output of sub-step S3.2 is the optimized set of optimal material property parameters. This parameter set This set of parameters is considered to be the most accurate description of the actual physical and mechanical behavior of concrete under current working conditions. The finite element verification module will use this parameter set. Perform a final finite element simulation again, and the output will be used to calculate the multi-objective reward in sub-step S3.3.

[0104] See attached document Figure 4 In sub-step S3.3, the finite element verification module verifies the current candidate values ​​for construction joint spacing based on the final finite element simulation results output after calibration in sub-step S3.2. Conduct multi-dimensional quantitative evaluation and calculate the corresponding multi-objective rewards. This multi-objective reward It is a vector whose purpose is to provide a clearly directional feedback signal for the update of the strategy generation module in step S4.

[0105] This multi-objective reward It consists of multiple components, and in one specific embodiment, it includes a crack resistance performance target bonus. Construction efficiency target bonus and economic cost target rewards .

[0106] Crack resistance performance target bonus The calculation is based on the maximum principal tensile stress of the entire structure throughout the entire analysis time period, output by the finite element simulation. First, the principal tensile stresses of all elements at all time steps are extracted from the simulation results, and their global maximum values ​​are determined. and the age at which this value appears. Simultaneously, based on the tensile strength development curve calibrated in step S1, the concrete's strength at various ages is obtained. Design value of tensile strength at time Safety factor for crack resistance It can be defined as:

[0107] ;

[0108] Crack resistance performance target bonus Defined as this safety factor The function is designed to... When the value is much greater than 1.0 (i.e., the safety margin is sufficient), the reward value is a relatively high positive value; when When the value is close to or less than 1.0 (i.e., there is a risk of cracking), the reward value drops sharply and becomes a negative value with a large absolute value to penalize unsafe candidate values.

[0109] Construction efficiency target bonus The calculation is based on candidate values ​​for construction joint spacing. The size of the construction joint itself. Without compromising structural safety, a larger construction joint spacing means a longer continuous pouring length, reducing the number of interruptions in formwork installation and removal, rebar connection, and concrete pouring caused by construction joints. Therefore, construction efficiency is related to the construction joint spacing. The magnitude of the reward is positively correlated with the value of the reward. This reward can be directly defined as... A function that is directly proportional to:

[0110] ;

[0111] in, It is a preset weighting coefficient used to convert length units into dimensionless reward values.

[0112] Economic cost target reward The calculation is based on the cost changes caused by variations in the number of construction joints. For a total length of... The number of construction joints in the structure to be poured Approximately Each construction joint incurs additional costs, including the cost of special treatment and installation of the formwork, the cost of waterproofing materials such as waterstops, and the labor cost of roughening and cleaning the joint surface. The cost per unit construction joint is denoted as... Therefore, the total cost associated with construction joints and Inversely proportional. Economic cost target reward Defined as a measure of cost savings, therefore... Proportional. This reward can be expressed as:

[0113] ;

[0114] in, It is a baseline constant to ensure the range of reward values.

[0115] After calculating the individual rewards, since each reward has different dimensions and numerical ranges, they need to be normalized, for example, mapped to the interval [-1, 1]. The normalized individual rewards together constitute the final multi-objective reward vector:

[0116] ;

[0117] in, This represents the normalized reward value. This reward vector... It is output to step S4 as the basis for updating the optimization model inside the strategy generation module; This represents the normalized crack resistance performance target bonus. This represents the normalized construction efficiency target bonus. This represents the normalized economic cost target reward.

[0118] In step S4, the candidate values ​​for construction joint spacing generated in step S2 are... Compared with the multi-objective reward vector calculated in step S3.3 This forms a data pair. This data pair is fed back to the policy generation module 102 to update its internal probabilistic proxy model based on Gaussian processes.

[0119] First, for multi-objective reward vectors Perform scalarization to convert it into a single utility value. This scalarization can be achieved using a predefined utility function, such as a linear weighted sum:

[0120] ;

[0121] in, , and These are preset weighting coefficients corresponding to crack resistance, construction efficiency, and economic cost, respectively, and they add up to 1. These weighting coefficients are configured according to specific engineering needs to reflect the relative importance of different engineering objectives.

[0122] The newly acquired scalarized data points are then added to the historical evaluation database maintained within the strategy generation module. This database stores all previously evaluated construction joint spacings and their corresponding utility values.

[0123] Next, the policy generation module uses the updated, complete historical evaluation database containing new data points to perform a posterior update on its internal Gaussian process-based probabilistic surrogate model. This update process is the core step of Bayesian optimization, which recalculates the posterior distribution of the Gaussian process based on Bayes' theorem, combining the model's prior distribution and the likelihood of all observed data.

[0124] Specifically, this posterior update recalculates the mean function of the Gaussian process. The covariance function (or kernel function). This update will affect the decision space for the entire construction joint spacing: at this evaluation point At this point, the prediction uncertainty (i.e., prediction variance) of the updated model. It will significantly reduce its predicted mean. It will approach the observed utility value Meanwhile, in Within the neighborhood of the model, the predicted mean and variance will also be affected and adjusted accordingly, thereby improving the fitting accuracy of the probabilistic surrogate model to the real utility function.

[0125] By repeatedly executing steps S2, S3, and S4, the strategy generation module continuously updates and improves its internal optimization model based on an increasing amount of historical data (construction joint spacing candidate values, multi-objective rewards). The output of this process is an updated strategy generation module with a more accurate description of the decision space. This updated module will be used in step S2 of the next iteration to generate a new, more promising construction joint spacing candidate value.

[0126] In step S5, after the iterative optimization process from steps S2 to S4 meets the preset termination condition, the decision output module generates and outputs the optimal construction joint spacing optimization control decision based on the final updated optimization model in the strategy generation module. .

[0127] The termination condition for the iterative optimization process can be set as one or a combination of the following:

[0128] The total number of iterations executed has reached a preset maximum value.

[0129] Candidate values ​​for construction joint spacing generated by the strategy generation module The change is less than a preset threshold in several consecutive iterations.

[0130] Scalar utility value of multi-objective reward calculated by the finite element verification module In a series of iterations, its gain is less than a preset threshold.

[0131] When the termination condition is met, the decision output module retrieves its final state from the policy generation module, using a Gaussian process-based probabilistic surrogate model. This model contains information about all evaluated data points up to the current point. The decision output module then calculates the predicted mean function of this model. A global optimization is performed to determine the construction joint spacing that maximizes the predicted utility value; this value is then determined as the optimal construction joint spacing. :

[0132] ;

[0133] in, The final updated Gaussian process model is used for construction joint spacing. The predicted mean of utility.

[0134] In addition to outputting the optimal construction joint spacing In addition, the decision output module also outputs auxiliary decision-making information. In a specific embodiment, the auxiliary decision-making information includes:

[0135] With the optimal construction joint spacing The corresponding predictive performance metric. This metric includes the predictive utility value under the optimal decision. And the predicted values ​​of each normalized component reward that constitutes this utility value, i.e. These individual predicted values ​​clearly demonstrate the specific performance of the optimal decision in terms of crack resistance, construction efficiency, and economic cost.

[0136] With the optimal construction joint spacing The corresponding prediction uncertainty. This information is derived from the Gaussian process model. Predicted standard deviation at This indicates that it quantifies the confidence level of the current model in predicting the optimal decision.

[0137] In an optional embodiment, the scalarization process in step S4 can be omitted to handle multi-objective optimization problems. In this case, the policy generation module internally maintains a Gaussian process model capable of directly processing multi-objective vectors. The output of step S5 is no longer a single optimal value, but a Pareto front solution set. This solution set consists of multiple non-dominated solutions, where any improvement in the performance of one objective in any solution necessarily leads to a decrease in the performance of at least one other objective. This Pareto front solution set provides engineering decision-makers with a set of alternative optimal solutions that make different trade-offs among different objectives.

[0138] Ultimately, the optimized control decision is output by the decision output module. The auxiliary decision-making information is transmitted to a user terminal interface or directly integrated into the construction management information system to generate specific construction instructions to guide on-site construction personnel in the positioning and installation of construction joint formwork.

[0139] See attached document Figure 5 To reduce the computational time cost of the entire optimization control method, this invention provides an optional embodiment. The main difference between this embodiment and the previous embodiment is that a proxy model is introduced between the strategy generation module and the finite element verification module to reduce the number of calls to high-fidelity finite element simulation.

[0140] The technical motivation for this optional embodiment is that the thermo-coupled finite element simulation in sub-step S3.1 requires a large amount of computational resources and time to execute each time. If this high-fidelity simulation is called in every iteration, the entire optimization process will have a very long cycle.

[0141] In this optional embodiment, the interaction flow of steps S2 to S4 is modified. First, a proxy model is constructed. In a specific implementation, this proxy model is a Gaussian process regression-based model. This proxy model is trained using all data pairs (construction joint spacing, multi-objective reward) that have been evaluated by the finite element verification module and have obtained real rewards. The function of this proxy model is to establish a model based on construction joint spacing... An approximate mapping relationship to its corresponding reward with extremely low computational cost.

[0142] This method forms a two-level interactive optimization loop. In the inner loop, the strategy generation module does not interact directly with the finite element verification module, but rather with the surrogate model. The strategy generation module uses its acquisition function to perform a large number of rapid virtual queries on the surrogate model within a preset decision space to find the next candidate point that maximizes the acquisition function value.

[0143] In the outer loop, once the inner loop determines an optimal candidate point, a high-fidelity finite element simulation is not executed immediately. Instead, a preset criterion is used to determine whether to call the finite element verification module. This criterion could be: a high-fidelity finite element simulation is only initiated when the expected gain or uncertainty value of the acquisition function at the optimal candidate point exceeds a preset threshold.

[0144] If the criterion is met, the finite element verification module is invoked to accurately evaluate the candidate points and obtain a true reward value. This newly obtained, high-fidelity data pair is then used to update the surrogate model, improving its prediction accuracy in subsequent iterations. If the criterion is not met, the evaluation can be performed on the suboptimal candidate points predicted by the surrogate model, or the current iteration can be terminated directly.

[0145] This approach creates a closed loop for active learning. The majority of exploratory evaluations are performed by computationally inexpensive surrogate models, while computationally expensive high-fidelity finite element simulations are used only to evaluate the most informative points that can maximally improve the surrogate model or are the most likely candidates for the global optimum. This configuration significantly reduces the total number of finite element simulation calls while ensuring the accuracy of the final optimization decision, thereby drastically shortening the overall optimization process time.

Claims

1. A method for optimizing the spacing of construction joints in mass concrete based on finite element simulation, characterized in that, Includes the following steps: S1. Obtain environmental status information: Obtain environmental status information during the construction of large-volume concrete, including material property parameters, environmental parameters, construction progress parameters, real-time monitoring data, and structural geometric parameters. S2. Generate candidate values ​​for construction joint spacing: Based on the environmental state information, generate one or more candidate values ​​for construction joint spacing to be evaluated through a strategy generator; S3. Perform finite element self-calibration verification: Perform finite element self-calibration verification on the candidate values ​​of the construction joint spacing. The finite element self-calibration verification includes the following steps: S3.1 Perform finite element simulation: Based on the currently acquired material property parameters and the candidate values ​​of the construction joint spacing, perform a thermo-mechanical coupling simulation of the large-volume concrete to obtain the predicted temperature field data and stress-strain field data of the large-volume concrete. S3.2 Perform model parameter self-calibration: acquire the real-time monitoring data, and based on the deviation between the real-time monitoring data and the temperature field prediction data and the stress-strain field prediction data, iteratively adjust the material property parameters used in the finite element simulation to minimize the deviation; S3.3 Calculate the multi-objective reward: Based on the finite element simulation results after self-calibration of the model parameters, calculate the multi-objective reward corresponding to the candidate value of the construction joint spacing; S4. Update the strategy generator: Update the optimization model inside the strategy generator based on the candidate values ​​of the construction joint spacing and the multi-objective reward; S5. Output optimized control decision: Based on the updated strategy generator, output the optimal construction joint spacing optimization control decision.

2. The method for optimizing the construction joint spacing of mass concrete based on finite element simulation according to claim 1, characterized in that, The steps for obtaining the material property parameters in step S1 specifically include: The parameters of the hydration heat release curve, the development law of elastic modulus with age, the tensile strength development curve, the compensation shrinkage rate and temperature rise suppression effect of the high-performance expansion agent of the current batch of concrete are obtained as the material property parameters. 3.The method for optimizing and controlling the construction joint spacing of mass concrete based on finite element simulation according to claim 1, characterized in that, In step S3.3, the step of calculating multi-objective rewards specifically includes: Calculate the target bonus for crack resistance based on the maximum tensile stress or maximum tensile strain predicted by the finite element simulation. Calculate the construction efficiency target reward based on the candidate values ​​of the construction joint spacing; Calculate the economic cost target reward based on the impact of the number of construction joints on labor and material costs. 4.The method for optimizing and controlling the construction joint spacing of mass concrete based on finite element simulation according to claim 1, characterized in that, The S2 step specifically includes: The strategy generator uses a Gaussian process or a Bayesian optimization algorithm to generate candidate values ​​for the construction joint spacing based on the environmental state information and the optimization model inside the strategy generator.

5. The method for optimizing the construction joint spacing of mass concrete based on finite element simulation according to claim 1, characterized in that, The thermal coupling simulation step for large-volume concrete in step S3.1 specifically includes: The temperature, stress, and strain fields of the large-volume concrete were simulated from the start of pouring to the temperature peak subsidence and shrinkage stabilization. The evolution process takes into account the heat of hydration, temperature gradient, drying shrinkage, autogenous shrinkage, creep, and the compensating shrinkage effect of the expansion agent.

6. The finite element simulation-based mass concrete construction joint spacing optimization control method according to claim 1, characterized in that, The deviation in step S3.2 is calculated as the error norm between the real-time monitoring data and the temperature field prediction data and the stress-strain field prediction data. The iterative adjustment employs gradient descent or Bayesian inversion methods to optimize and adjust the material property parameters used in the finite element simulation, thereby minimizing the error norm, i.e., minimizing the deviation.

7. The finite element simulation-based mass concrete construction joint spacing optimization control method according to claim 6, characterized in that, The material property parameters include the hydration heat release coefficient, the elastic modulus development curve parameters, and the actual expansion rate coefficient of the high-performance expansion agent.

8. The method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation according to claim 1, characterized in that, The method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation further includes adding a surrogate model acceleration step between step S2 and step S3, specifically: A surrogate model based on Gaussian process regression is constructed to quickly predict the multi-objective reward corresponding to the candidate values ​​of the construction joint spacing. The strategy generator interacts with the agent model first to filter out potential candidate values ​​for the construction joint spacing. The potential candidate values ​​for construction joint spacing are then evaluated using the finite element self-calibration verification module, and the evaluation results are used to update the surrogate model.

9. The method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation according to claim 1, characterized in that, The step of updating the optimization model inside the strategy generator in step S4 specifically includes: the strategy generator continuously learns and improves its internal optimization model based on the candidate values ​​of the construction joint spacing and the historical data of the multi-objective reward.

10. The method for optimizing and controlling the spacing of construction joints in large-volume concrete based on finite element simulation according to claim 1, characterized in that, The step S5, which outputs the optimal construction joint spacing optimization control decision, specifically includes: outputting the optimal construction joint spacing, and using key environmental state parameters and their weight contributions to support the optimal construction joint spacing optimization control decision as auxiliary decision information.

Images