Winter construction reinforced concrete electromagnetic curing method and system based on layered progressive mathematical control architecture

Abstract

The embodiment of the invention provides a winter construction reinforced concrete electromagnetic curing method and system based on a layered progressive mathematical control framework, and belongs to the technical field of crossing of intelligent construction, advanced process control and computational mathematics. The method comprises the following steps: acquiring temperature, hydration degree and environmental parameters in real time, performing space-time fusion with electromagnetic heating input, constructing a maintenance state feature vector, establishing a parameterized grey box prediction model based on a multi-physics coupling mechanism, and realizing online joint estimation of state and model parameters by combining Bayesian inference and unscented Kalman filtering. On this basis, distributed robust model predictive control is introduced, modeling is carried out on external disturbance uncertainty, an optimal heating strategy is generated, finally, execution of all heating subareas is coordinated through a distributed cooperative control algorithm, global optimization of the maintenance process is achieved, the temperature difference risk in winter construction is effectively reduced, and the maintenance efficiency is improved. The maintenance energy efficiency and the system operation stability are improved.

Description

Technical Field

[0001] The embodiments in this specification relate to the interdisciplinary technical fields of intelligent construction, advanced process control, and computational mathematics, and particularly to a method and system for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture. Background Technology

[0002] In the field of winter concrete curing, existing technologies mainly acquire internal temperature data by deploying embedded sensor networks and combining them with traditional proportional-integral-derivative (PID) control or basic model predictive control (MPC) to adjust the power output of heating equipment. The system achieves basic closed-loop regulation based on feedback data and utilizes zoned temperature control technology to improve the consistency of the temperature field of large components to a certain extent.

[0003] However, when dealing with the strong coupling characteristics of multiple physical fields such as "electricity-heat-chemistry", existing technologies often suffer from insufficient control precision and high energy consumption due to oversimplification of physical parameters, lack of effective uncertainty quantification mechanisms (such as environmental disturbance immunity), and weak logic in large-scale distributed regional coordination. This makes it difficult to cope with maintenance challenges in complex and extreme environments. Summary of the Invention

[0004] In view of this, embodiments of this specification provide a method for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture; it also relates to an electromagnetic curing system for reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture, a mathematical control architecture embodying the logic of the system, and a hardware deployment architecture of the system; a computing device, a computer-readable storage medium, and a computer program, in order to solve the technical defects existing in the prior art.

[0005] According to a first aspect of the embodiments of this specification, a method for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture is provided, comprising:

[0006] Offline modeling and initialization steps: Based on the design information, a multiphysics mechanism model is constructed and discretized into a parameterized state space structure to complete the basic model construction; then, the core parameters and prior knowledge base covering real-time estimation, robust optimization, distributed collaboration and meta-learning are initialized simultaneously to lay the foundation for subsequent control and evolution.

[0007] Multi-source data acquisition steps: Real-time acquisition of internal temperature, hydration degree, maturity parameters of reinforced concrete beams or slabs in zones under winter construction conditions, as well as environmental disturbance data including air temperature and wind speed.

[0008] Real-time status and parameter synchronization estimation steps: A "double-ring nested" online joint estimation mechanism is adopted. First, the parameters and status are predicted, and then the status is updated and the parameters are corrected by combining the measured feedback data, so as to capture the non-stationary characteristics of concrete during the curing process.

[0009] Multi-objective split-bar optimization steps: Based on the estimated real-time model parameters, construct an uncertainty set, and solve the split-bar model predictive control (DR-MPC) proposition under the premise of satisfying multi-objective constraints (temperature control accuracy, optimal energy efficiency, and intensity development standards).

[0010] Distributed collaborative tracking steps: The temperature control target output by global optimization is decomposed into tasks, and a distributed collaborative algorithm is used to achieve unified trajectory tracking and execution pace among heating zones.

[0011] Control command implementation steps: Output and execute electromagnetic heating control commands for multi-zone power adjustment, drive the electromagnetic induction heating device to precisely heat the components.

[0012] In one possible implementation, the real-time state and parameter synchronization estimation step employs Bayesian parameter estimation based on unscented Kalman filtering (UKF) or extended Kalman filtering (EKF) in the outer layer. By real-time correction of model parameters (such as equivalent heat release rate and activation energy), model drift caused by extreme environmental temperature differences is eliminated.

[0013] In one possible implementation, the uncertainty set modeling in the multi-objective sub-Bruker optimization step covers wind speed fluctuations and random changes in ambient temperature in the future prediction time domain, ensuring that the control command can still guarantee that the temperature difference between the inside and outside of the concrete does not exceed the safety threshold even under the "worst-case" conditions.

[0014] In one possible implementation, the distributed cooperative tracking step utilizes the Alternating Directional Multiplier Method (ADMM) to establish a communication consensus among the partitioned executors, transforming complex global optimization propositions into local low-dimensional computations to improve the real-time response of large-scale component partitioned maintenance.

[0015] According to a second aspect of the embodiments of this specification, a winter construction electromagnetic curing system for reinforced concrete based on a hierarchical progressive mathematical control architecture is provided, comprising:

[0016] Concrete components for winter construction: The underlying objects of the system are zoned reinforced concrete beams / slabs in frigid environments. By logically dividing large-scale components into multiple temperature-controlled zones, refined spatial management is achieved.

[0017] Multi-source sensor module: Deployed in the partitioned reinforced concrete components, it is used to collect raw sensing signals throughout the entire maintenance cycle in real time, including internal temperature, hydration degree, maturity, and environmental disturbance data.

[0018] The physics modeling and order reduction prediction module is used to store the zeroth layer multiphysics coupling mathematical model and to project the high-dimensional physical equations onto the parameterized state space using the intrinsic orthogonal decomposition (POD) technique, thereby constructing a first-layer high-fidelity "grey box" prediction model with real-time computing capabilities.

[0019] Online joint sensing and estimation module: It adopts a double-ring nested Bayesian inference mechanism to perform a second-level online joint estimation and synchronous correction of the state and parameters of the "gray box" prediction model using the original sensing signals.

[0020] Distributed Cooperative Optimization and Evolution Module: As the decision center, it integrates the distributed robust optimization unit, the distributed cooperative tracking unit, and the meta-learning unit. It is used to solve the third-layer robust control strategy and the fourth-layer distributed cooperative trajectory under multi-objective constraints, and to achieve lifelong optimization of the fifth-layer system performance based on historical data.

[0021] Electromagnetic induction heating device: As an actuator, it includes a multi-zone power adjustment unit for receiving control commands output by the distributed collaborative optimization and evolution module and performing zoned induction heating on the reinforced concrete component.

[0022] In one possible implementation, the physical modeling and order reduction prediction module, targeting the characteristics of winter construction, introduces a transient heat transfer equation containing phase change correction terms and hydration heat release correction terms into the zeroth layer multiphysics coupling mathematical model, which can accurately describe the nonlinear thermal behavior of concrete near the freeze-thaw critical point.

[0023] In one possible implementation, the physical modeling and order reduction prediction module adopts a mechanism-data dual-driven residual compensation strategy when constructing the first-layer high-fidelity "gray box" prediction model. It uses historical maintenance data to perform neural network compensation for random boundary disturbances not covered in the physical equation, thereby improving the prediction accuracy of the model under extreme cold wave conditions.

[0024] In one possible implementation, the electromagnetic induction heating device supports independent frequency and power adjustment for different zones, and can achieve differentiated and precise control of the temperature rise rate of beam and slab components with different thicknesses or locations based on the collaborative control trajectory output by the distributed collaborative optimization and evolution module.

[0025] In one possible implementation, the distributed collaborative optimization and evolution module has cross-scenario knowledge transfer capabilities. By extracting the optimal control weights from historical projects for different component scales or extreme cold conditions, it enables rapid deployment and cold start optimization of control laws in new construction scenarios.

[0026] According to a third aspect of the embodiments of this specification, a computing device is provided, comprising:

[0027] Memory and processor;

[0028] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the above-mentioned electromagnetic curing method for reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture.

[0029] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions, which, when executed by a processor, implement the steps of the above-described method for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture.

[0030] According to a fifth aspect of the embodiments of this specification, a computer program is provided, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the above-described method for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture.

[0031] This specification provides an embodiment of a method and system for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture. The core of its technical solution lies in: deploying multi-source sensors within partitioned components to perceive the microenvironment and hydration state in real time; combining the zero-layer electro-thermal-chemical multi-physics coupling mechanism with the first-layer POD high-fidelity order reduction technology to construct a parameterized gray-box digital twin model with real-time characteristic response capabilities; utilizing data assimilation algorithms such as unscented Kalman filtering (UKF) to perform the second-layer online joint estimation, real-time correction of non-stationary hydration dynamic parameter drift and maturity state caused by extreme temperature differences, achieving accurate understanding of complex working conditions; constructing an uncertainty set containing random environmental disturbances based on the third-layer distributed blue bar control (DR-MPC), solving the optimal temperature control strategy under the premise of satisfying internal and external temperature difference safety constraints, and employing a fourth-layer ADMM-based distributed collaborative algorithm to decompose the global task, achieving game-theoretic balance and accurate trajectory tracking of multi-partition power in large-scale complex components; further introducing a fifth-layer meta-learning evolution mechanism to automatically extract control features and optimize controller hyperparameters, realizing knowledge transfer across project working conditions and spontaneous evolution of the control law. One or more embodiments in this specification significantly improve the quality reliability, thermal energy utilization efficiency, system anti-disturbance robustness, and intelligent level of full-cycle decision-making in winter concrete curing, effectively solving the problems of poor uniformity, high energy consumption, and model mismatch in the curing of large-volume reinforced concrete components under extreme cold environments. Attached Figure Description

[0032] Figure 1 This is a flowchart of an embodiment of the electromagnetic curing method for reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture provided in this specification.

[0033] Figure 2 This is an overall architecture diagram of an electromagnetic curing system for reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture, provided by one embodiment of this specification.

[0034] Figure 3 This is a schematic diagram of a mathematical control architecture device for an electromagnetic curing system for reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture, provided in one embodiment of this specification.

[0035] Figure 4 This is a schematic diagram of the hardware deployment architecture of an electromagnetic curing system for reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture, provided by one embodiment of this specification. Detailed Implementation

[0036] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0037] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0038] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another.

[0039] This specification provides a method for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture. This specification also relates to an electromagnetic curing system for reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture, a mathematical control architecture embodying the system logic, and a hardware deployment architecture for the system. Furthermore, this specification also relates to a computing device, a computer-readable storage medium, and a computer program, which will be described in detail in the following embodiments.

[0040] Each of the following embodiments will be described in detail. See also Figure 1 , Figure 1 The flowchart illustrates a method for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture, according to an embodiment of this specification, specifically including the following steps:

[0041] Step 101: Perform offline modeling and initialization, including hardware deployment, constructing a zeroth layer PDE mechanism model containing electromagnetic-thermal-mechanical multiphysics fields, and using intrinsic orthogonal decomposition (POD) technology to construct a first layer parameterized reduced-order prediction model.

[0042] Step 102: Conduct multi-source data acquisition, and obtain the physical field parameters inside each temperature control zone and the environmental disturbance parameters of the construction site in real time through pre-embedded sensors and environmental monitoring stations.

[0043] Step 103: Perform real-time state and parameter synchronization estimation. Using a dual Bayesian inference architecture, the collected observation data is input into an unscented Kalman filter (UKF) to identify time-varying model parameters online and simultaneously correct the hidden state inside the concrete.

[0044] Step 104: Perform multi-objective partial bar optimization. Based on the meteorological scenario set and parameter uncertainty set, construct the partial bar model predictive control (DR-MPC) proposition and solve the optimal power planning for each region under the premise of satisfying safety constraints.

[0045] Step 105: Perform distributed collaborative tracking. Each partition edge computing unit uses a distributed algorithm to communicate and compete, achieving heat distribution consistency under the global total power budget constraint, so as to track the control target issued by the optimization layer.

[0046] Step 106: Implement control commands, convert the optimized digital commands into power control signals, drive the electromagnetic induction heating device to precisely adjust the heat generation of each zone component, and display maintenance monitoring information.

[0047] The zeroth-layer PDE mechanism model can refer to a full-physics field description model that includes Maxwell's equations, Krstulovic-Dabic hydration kinetic equations, and linear thermoelastic constitutive models. Offline modeling and initialization refers to establishing the PDE model based on the BIM model and multiphysics coupling theory before the start of maintenance, and using POD technology to reduce the state space from tens of thousands of dimensions to a lower dimension to meet the computational requirements of real-time control. The first-layer parameterized prediction model refers to a reduced-order state space model that can achieve second-level real-time calculations by using projection transformation while maintaining the physical characteristics of the PDE model. Multi-source data acquisition refers to temperature and strain data collected by fiber optic grating sensors, chemical reaction data collected by hydration degree sensors, and driving variables such as ambient temperature, wind speed, and solar radiation intensity collected by meteorological stations. Real-time state and parameter synchronous estimation refers to a nested real-time filter architecture that eliminates the distortion between the physical field model and the actual working conditions through a loop logic of "parameter prediction - state prediction - state update - parameter update". Unscented Kalman Filter (UKF) can refer to a class of algorithms used for state estimation of nonlinear systems. In this embodiment, it is used to identify key parameters such as thermal conductivity correction coefficient, hydration energy, and electromagnetic conversion efficiency online. Future scenario set can refer to a set of samples generated based on historical meteorological patterns using deep learning models such as Long Short-Term Memory (LSTM) networks, reflecting the probability of environmental temperature and wind speed fluctuations within the next 24 hours. Multi-objective distributed bar optimization can refer to a mathematical programming method that seeks the globally optimal control law by adjusting objective weight coefficients (such as temperature difference constraints and energy efficiency) while considering the worst-case meteorological disturbance distribution. Distributed cooperative tracking can refer to the iterative calculation performed by each partition unit using algorithms such as Alternating Direction Multiplier Method (ADMM) to collaboratively solve cross-partition thermal interaction and power allocation problems without relying on a global central processing unit. Control command implementation can refer to the process of adjusting the output power of a high-frequency induction power supply through techniques such as pulse width modulation (PWM), causing the electromagnetic coil to generate an alternating magnetic field inside the concrete structure and convert it into heat energy.

[0048] The embodiments of this specification are further described below through a detailed example:

[0049] This specification describes an embodiment of a method and system for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture, applied to the winter construction and curing process of a prestressed concrete box girder for a railway in a cold region. The box girder is 30m long, 2.2m high, and has a top slab width of 12m. Under extremely low ambient temperature conditions, this embodiment aims to achieve concrete strength of over 90% of the design strength within 72 hours through layered progressive mathematical control, and to ensure that the maximum internal and external temperature difference does not exceed [a certain value] throughout the curing process. The box girder is designed for precise temperature control. Physically, it is divided into 20 standard temperature control zones. Each zone is configured with different types of temperature control units, such as core zone temperature control units, web zone temperature control units, or top and bottom plate zone temperature control units, based on its geometric location and thermodynamic characteristics.

[0050] After the system starts running, offline modeling and initialization are performed first. In this step, the system first decouples the physical space based on the BIM model of the box girder, dividing the 30-meter-long box girder into 5 monitoring sections along its length, and further refining them into 20 temperature control zones, including the core area, web area, and top and bottom slab areas. At the same time, the system constructs a zero-layer refined three-dimensional physical mechanism model, including the steel reinforcement skeleton, formwork, and insulation layer. This model comprehensively considers electromagnetic field eddy current losses, dynamic thermophysical parameters of concrete, three-stage hydration kinetic evolution law, and thermodynamic constitutive relations with the development of hydration degree. In order to solve the real-time problem caused by the high computational freedom of large-scale components (such as up to 150,000 dimensions), the system adopts the intrinsic orthogonal decomposition technique to extract the dominant modes, project the high-dimensional physical field to the low-dimensional space, and construct a parameterized reduced-order prediction model with only 100-dimensional computation for the first layer, laying a mathematical foundation for subsequent real-time control.

[0051] Subsequently, multi-source data acquisition was conducted. At the maintenance site, the system acquired multi-dimensional parameters in real time through a pre-deployed sensing network. These real-time sensing parameters were collected by high-interference-resistant fiber optic temperature sensors embedded within each zone, spaced at 0.5m intervals. This data covered temperature, strain, and hydration degree data reflecting the chemical reaction process at key measuring points inside the concrete. Environmental disturbance parameters were synchronously acquired through environmental monitoring stations installed around the beam and in internal ventilation areas, including real-time ambient temperature, wind speed, and solar radiation intensity. All sensing data underwent time synchronization and spatial alignment processing via an IoT gateway, serving as the real-time input for the system's closed-loop control.

[0052] Next, real-time state and parameter synchronization estimation is performed. The platform is configured with an unscented Kalman filter based on a dual Bayesian inference architecture for each partition. This step utilizes the acquired real-time temperature and strain observations and identifies time-varying parameters (such as thermal conductivity correction coefficients, electromagnetic-thermal conversion efficiency, and hydration kinetic pre-exponential factors) that fluctuate with environmental and material evolution through a "prediction-update" loop logic. Through this synchronization estimation mechanism, the system can optimally correct the deviation between the physical model and actual operating conditions and calculate the hidden thermal field distribution state inside the component in real time.

[0053] Then, multi-objective sub-Bruker optimization is performed. The decision engine uses deep learning networks (such as LSTM) to predict historical meteorological data, generating multiple typical scenario sets reflecting meteorological fluctuations in the next 24 hours. Based on this, the system constructs a sub-Bruker model to predict control propositions. Within each control cycle, the system strictly satisfies the maximum internal and external temperature difference (not exceeding...). Under the constraints of safety probability, and with the multiple objectives of minimizing temperature fluctuation risk and minimizing heating energy consumption, the optimal power planning sequence for each zone over a future period is solved. This optimization method ensures that the control strategy maintains strong robustness and predictability in the face of uncertain disturbances such as extreme cold waves or strong winds.

[0054] Subsequently, distributed collaborative tracking is implemented. For the collaborative needs of the 20 temperature-controlled zones, the edge computing units of each zone define coupling weights based on the adjacency relationships of the finite element model. Since the total power of the on-site transformers is constrained by budget, the system employs a distributed algorithm (such as the alternating direction multiplier method) for communication game theory. Each unit continuously exchanges thermal interaction information and power multipliers at the boundaries via the local area network. Without requiring centralized processing, a consistent consensus on global heat distribution is reached through multiple iterations, thereby accurately tracking the temperature control targets for each zone issued by the optimization layer.

[0055] Finally, control commands are implemented. The system converts the digital power commands for each zone obtained through collaborative solving into high-frequency pulse width modulation (PWM) signals, driving the electromagnetic induction power supply to precisely regulate the heat generation of the induction coils in each zone. Under extreme operating conditions, the system can complete initialization through a learning phase of approximately one hour, subsequently entering a stable control state and raising the core temperature of the box girder to the preset range. Actual measurements show that the internal and external temperature difference was successfully suppressed throughout the entire process. Within this range, not only is structural construction safety ensured, but energy savings of approximately 35% are also achieved compared to traditional fixed-power solutions.

[0056] After the task is completed, the system also performs knowledge accumulation. Following the 72-hour maintenance period, the system automatically summarizes nearly a thousand sets of data generated throughout the process, updates the hydration model parameters and controller weights through offline analysis and Bayesian optimization, and stores the construction experience in a digital knowledge base. This self-evolutionary mechanism provides a referable intelligent control solution for future railway box girder projects in similar frigid regions.

[0057] Through this hierarchical and progressive control architecture, one or more embodiments in this specification achieve a seamless connection from microscopic physical mechanisms to macroscopic collaborative scheduling. The real-time challenges in the curing of large-scale complex components are solved using order reduction techniques and distributed computing, and distributed bar optimization effectively addresses uncertain environmental disturbances during winter construction. Ultimately, while ensuring structural construction safety and precise temperature control, curing energy consumption is significantly reduced, promoting a deep transformation of concrete construction in frigid regions towards intelligence and low-carbon practices.

[0058] In one possible implementation, the offline modeling and initialization includes constructing a zeroth-layer refined three-dimensional physical mechanism model based on design information, and performing global initialization settings for the algorithm parameters, uncertainty sets, and meta-prior knowledge bases involved in the system at each layer. The zeroth-layer refined three-dimensional physical mechanism model includes a set of geometric properties composed of steel strands, ordinary reinforcing bars, formwork, and insulation layers, as well as a set of mechanism parameters including electromagnetic eddy current losses, dynamic thermophysical parameters of concrete, three-stage hydration kinetics, and thermodynamic constitutive relations.

[0059] Among them, electromagnetic field eddy current loss This refers to solving the frequency domain. The spatial distribution of heat generation obtained from the equation is used to quantitatively describe the heat source intensity generated by the induction coil around the prestressed steel strand; concrete thermal conductivity and specific heat capacity Using dynamic correction formula initial value The hydration kinetics evolution was modeled using the Krstulovic-Dabic three-stage model, with each stage employing... This describes the crystallization nucleation, phase boundary reaction, and diffusion stages of concrete during the curing process. The thermal stress analysis constitutive model uses a linear elastic constitutive model, with the elastic modulus evolving with the degree of hydration. ,in This is the final elastic modulus. System parameter initialization is completed synchronously, and the initial distribution of the parameters to be identified is set. and its covariance Filter (EKF / UKF) noise covariance Set the initial objective function weights for the third-level distributed bar optimization control (DR-MPC). Opportunity-constrained risk level and uncertainty set It also completes the loading of the fifth-layer prior knowledge base, providing a basic mathematical environment and prior parameter system for subsequent online adaptive control.

[0060] The first-level parameterized order reduction prediction model refers to the real-time computational order reduction model (ROM) formed by projecting the original high-dimensional (e.g., 150,000-dimensional) physical field state vector to a low-dimensional (e.g., 100-dimensional) subspace through intrinsic orthogonal decomposition (POD).

[0061] In practical applications, this step first involves constructing a detailed 3D geometric model based on the box girder's BIM model, including steel strands, ordinary reinforcing bars, formwork, and insulation layers, and then implementing physical space decoupling. The specific hardware deployment scheme is as follows: the 30-meter-long box girder is divided into five sections along its length. Each section is further divided into a core area, a web area, and top and bottom slab areas along its height and width, forming a total of 20 independent temperature-controlled zones. High-interference-resistant fiber optic grating (FBG) temperature sensors and a set of electromagnetic induction coils coupled to the prestressed steel strands are pre-embedded within each zone. Simultaneously, environmental sensors are placed around the girder and in internal ventilation areas, thus transforming the large-scale component into a modular sensing and control unit, providing a static parameter foundation for subsequent real-time calculations.

[0062] In practical implementation, the system first performs finite element mesh generation on the three-dimensional geometry of the box girder, generating approximately 50,000 tetrahedral elements. A state vector is defined, containing the temperature of all nodes, element hydration degree, and saturation degree, with an initial dimension of approximately 150,000. The system then calls upon a meta-prior knowledge base, combining engineering experience and historical data, to... and the opportunity constraint risk level of DR-MPC Initial values ​​are assigned based on preliminary predictions. These initial configurations, encompassing first-principles mechanistic models, probability distributions, and engineering constraints, not only provide a reliable physical foundation for subsequent real-time computations but also guide the convergence direction of all subsequent online adaptive iterations.

[0063] The beneficial effect of this embodiment lies in that, by combining the modular partitioning of BIM space with parametric order reduction technology based on first principles, it fundamentally solves the problem of balancing "accuracy of physical field simulation" and "real-time performance of on-site control" in the curing of large-scale complex components in frigid regions. While ensuring accurate capture of complex mechanisms such as dynamic thermal conductivity, three-stage hydration kinetics, and variable modulus thermal stress, it achieves a cross-order-of-magnitude compression of computational dimensions, laying a reliable mathematical and physical foundation for adaptive and precise curing of concrete components during winter construction.

[0064] In one possible implementation, the first-layer model discretization, parameterized order reduction, and data acquisition refer to reducing the order of the physical field through intrinsic orthogonal decomposition (POD) technology and simultaneously constructing a physical sensing network to obtain the real-time data stream driving the digital twin model. The first-layer parameterized order reduction prediction model is a low-dimensional state-space model constructed based on the dominant modes extracted by POD.

[0065] To eliminate the real-time control computation bottleneck caused by high-dimensional models, the system employs POD technology to extract the top 100 dominant modes from large-scale historical simulation data. Through projection transformation within this mode subspace, the original partial differential equations (PDEs) are transformed into a first-level parameterized reduced-order prediction model (ROM) with significantly reduced dimensionality. The projection relationship satisfies... ,in The basis matrix of POD is... This is a reduced-order state vector. The real-time sensing parameters (measurement vector) This includes temperature, strain, and hydration data obtained from over 500 measuring points pre-embedded within 20 zones; the environmental disturbance parameters (interference vectors) This includes real-time ambient temperature, wind speed, and solar radiation intensity.

[0066] In practical applications, this step constructs the system's real-time sensing layer through a "physical sensing network." After system startup, high-density sensors distributed across 20 zones and external environmental monitoring stations enter continuous monitoring mode. The heterogeneous signals collected in a dispersed manner are aggregated to the edge computing unit deployed on-site. Utilizing edge computing capabilities, the sensor coordinates are mapped in real-time to the 20 modular temperature control zone matrices established in step 201, ensuring that each physical zone receives real-time internal and external field observation feedback matching its geometric location. Environmental disturbance parameters (such as real-time wind speed) are used to correct the convective heat dissipation coefficient of the beam surface, and solar radiation intensity is used to adjust the thermal flow boundary conditions of the roof slab. These data collectively constitute the real-time driving data stream for the Bayesian state estimator and the distributed bar controller.

[0067] In practice, data acquisition follows a preset sampling period. The edge computing unit not only performs data pass-through but, more importantly, implements strict time synchronization processing to ensure that the sensing parameters of each partition are maintained. Environmental disturbance parameters Complete alignment on the time scale eliminates phase errors caused by sensor response lag or network latency. The acquired multidimensional parameters are vectorized and injected into the system memory pool, serving as the original evidence for state updates and parameter corrections in the second-layer algorithm using the unscented Kalman filter (UKF), thereby driving the online adaptive iteration of the entire closed-loop system.

[0068] The beneficial effect of this embodiment lies in clarifying the classification, collection, and spatiotemporal alignment mechanism of multi-source sensing data. By synchronously collecting data related to internal state perception and external environmental disturbances, and combining this with a high-density arrangement of 0.5m spacing, a physical sensing network, and a distributed edge computing architecture, "full-field, real-time, and blind-spot-free" monitoring of the entire process of large-scale component maintenance is achieved. This not only solves the "information blind spot" problem caused by the scarcity of monitoring points in traditional maintenance, but also ensures the anti-interference capability and control accuracy of the subsequent multi-bar control strategy through high-quality real-time input.

[0069] In one possible implementation, the real-time state and parameter synchronization estimation refers to the platform configuring an unscented Kalman filter (UKF) based on a dual Bayesian inference architecture for each partition, and using the collected real-time temperature and strain observations to implement a "prediction-update" loop logic, thereby identifying time-varying parameters that fluctuate with the evolution of the environment and materials online, and calculating the hidden thermal field distribution state inside the component in real time.

[0070] The parameter vector to be identified online is defined as follows: This includes thermal conductivity correction coefficients, pre-exponential factors and activation energies for the three stages of hydration, and electromagnetic thermal efficiency coefficients for 20 zones. The filter design employs UKF for state estimation and a nested recursive maximum a posteriori (MAP) estimator for parameter updates. The process noise covariance is also defined. Corresponding temperature state, unit K², measurement noise covariance The synchronous estimation loop includes: (1) parameter prediction, which infers the prior of the current parameters based on the posterior distribution of the previous time step. (2) State prediction: Based on the expected value of parameter prediction, UKF prediction is performed using the first-level reduced-order model to obtain the reduced-order state. Covariance (3) State update, using measured vectors Correcting prediction bias to obtain the current optimal state estimate (4) Parameter update, based on measurement innovation items Recursively update the posterior distribution of parameters .

[0071] In practical applications, this step achieves dynamic alignment between the physical model and actual working conditions. Due to the extreme and complex environment at construction sites in winter, the thermophysical properties of materials (such as heat conduction efficiency after initial setting) and the heat generation efficiency of induction coils fluctuate drastically over time. The system employs a dual Bayesian inference architecture. On one hand, it uses local observation data from over 500 measuring points to calculate the temperature distribution throughout the 30-meter box girder (i.e., calculates the 100-dimensional reduced-order state after projecting the 150,000-dimensional state vector). On the other hand, it senses and corrects fluctuations in the energy conversion efficiency of 20 zoned electromagnetic coils and changes in the rate of heat release during concrete hydration in real time. This mechanism ensures that the digital twin model can "self-correct" based on the actual site conditions, providing a precise initial state for subsequent power optimization control.

[0072] In practice, the edge computing unit performs a synchronous estimation once per sampling period (e.g., 5 minutes). The system first generates parameter priors using a parameter evolution model (assuming parameters change slowly with material hardening). Then, the predicted parameters are substituted into the first-level reduced-order prediction model. The nonlinear physical evolution is captured by the Sigma point set generated through unscented transformation. When feedback from the fiber optic sensor is obtained... Then, the system automatically calculates the residual between the predicted temperature field and the measured temperature, and adjusts the filter gain accordingly. Finally, the system feeds back the updated error to the parameter identification layer to correct the pre-exponential factor A or the thermal efficiency. This completes closed-loop adaptive learning. These optimal states are identified in real time. and parameter vector It is pushed to the third-level sub-bar optimization control module in real time.

[0073] The beneficial effect of this embodiment lies in achieving a dual breakthrough in overcoming "model uncertainty" and "state unmeasurability" during the curing process of large-scale components through deep coupling of dual Bayesian inference and UKF. This mechanism can not only automatically identify the microscopic evolution of the material's thermophysical properties, but also optimally correct the deviation between the physical model and the actual working conditions. Even under extreme conditions of drastic environmental fluctuations or sensor damage, the system can still use its powerful statistical inference capabilities to calculate the hidden temperature field, strain field, and hydration degree distribution inside the component in real time, ensuring that the control logic always operates on the "physical reality" track, providing solid data determinism for achieving millimeter-level precise temperature difference control of concrete components.

[0074] In one possible implementation, the multi-objective partial-bar optimization control refers to the decision engine constructing a partial-bar model predictive control (DR-MPC) proposition based on the real-time state estimation, parameter identification results, and uncertainty set provided by the preceding steps. Within each control cycle, the system strictly satisfies the maximum internal and external temperature difference (not exceeding...). Under the premise of safety probability constraints, with the multiple objectives of minimizing temperature fluctuation risk and minimizing heating energy consumption, the optimal power planning sequence for each zone in the future period is solved.

[0075] The configuration parameters for the multi-objective sub-Bruker optimization problem include: objective function weights, which refer to setting performance weights. To emphasize the safety of temperature difference constraints; uncertainty set construction refers to using 10 years of winter meteorological data from the project site to train an LSTM deep learning network as a weather prediction model, and generating 100 sets (S=100) of typical scenarios reflecting meteorological fluctuations in the next 24 hours. Simultaneously, the covariance is estimated based on the second-level parameters. Construct a 95% confidence ellipsoid As a set of parameter uncertainties; constraint setting refers to setting the maximum internal and external temperature difference. And set the allowed probability of violation (opportunity constraint risk level). Time domain parameters: set the sampling time Predicting the time domain (Corresponding to 2 hours), control time domain (Corresponds to 0.5 hours).

[0076] In practical applications, this step, through a distributed blob optimization architecture, ensures that the control strategy maintains strong robustness and predictability when facing uncertain disturbances such as extreme cold waves, strong winds, or drastic fluctuations in sunlight. The system no longer optimizes for a single "predicted value," but instead makes forward-looking decisions based on multiple meteorological scenario sets generated by the LSTM. This optimization method automatically balances heating costs with cracking risks: when a cold wave is predicted to occur within the next two hours, the decision engine increases the power reserve in corner zones in advance, thereby achieving system-level Pareto optimal maintenance while maintaining millimeter-level temperature difference control accuracy.

[0077] In practice, the system utilizes the nonlinear programming solver Ipopt and the code generation tool CasaADi to solve the DR-MPC problem once every 5-minute sampling period. First, the decision engine injects the optimal state at the current moment. and parameter vector The system then retrieves the future weather disturbance scenarios predicted by LSTM. Subsequently, the system satisfies... Under the premise of physical evolution laws and power upper and lower limits, Conditional Value at Risk (CVaR) is used to measure the degree of constraint violation in the worst-case scenario. The solver efficiently calculates the optimal power sequence in the future prediction time domain through sequential quadratic programming or interior-point methods. The system only issues the first step of the control input for the sequence to each partition, and continuously solves the problem over time to achieve closed-loop adaptive control.

[0078] The beneficial effect of this embodiment is that, by introducing bibliometric optimization and LSTM weather prediction, it solves the problem of the failure of traditional maintenance control methods that rely on guesswork or deterministic model prediction when faced with complex environmental disturbances. The system can quantitatively handle the uncertainty of material parameters and the randomness of the environment. Even when encountering unexpected sudden strong winds, it can still respond in advance based on its predictive mechanism, strictly locking the internal and external temperature difference within a certain range. Within the safety limit. Compared with traditional fixed-power solutions, this step significantly reduces energy loss while substantially improving the molding quality of large-scale concrete components in frigid regions.

[0079] In one possible implementation, the distributed collaborative tracking refers to addressing the collaborative needs of the 20 independent temperature control zones of the box girder. Each zone's edge computing units define coupling weights based on the adjacency relationships of the finite element model, and engage in communication and game theory among the units using the Alternating Direction Multiplier Method (ADMM). This step aims to achieve accurate tracking of the temperature control targets for each zone issued by the optimization layer, without requiring centralized processing, through information exchange within the local area networks of each zone, while meeting the global total power budget constraint.

[0080] The coupling weight of the interval thermal interaction is defined as follows: ,in Adjacent partitions Contact area This is the physical distance between the center points of the two partitions. The relevant configuration parameters for the distributed algorithm are set as follows: ADMM penalty parameter. The convergence tolerance is set to Furthermore, the maximum number of iterations within each control cycle is set to 20. Local cost function. It includes local tracking bias, power constraint multipliers, and thermal consistency terms with neighboring partitions.

[0081] In practical applications, this step solves the problems of "curse of dimensionality" and "power overload" in the zoning control of large and complex structures. This is because the total power of the transformers at the construction site is subject to strict budget constraints. The system no longer relies on a single central server, but instead uses a global reference trajectory. The system is decomposed to its edge units. Through a distributed game theory mechanism, each partition can perceive the heat penetration of adjacent areas in real time and automatically negotiate power allocation. This mechanism ensures that even under extreme cold waves, with limited total on-site power resources, the system can prioritize the temperature safety of critical components and achieve optimal balance in overall heat distribution.

[0082] In practical implementation, distributed collaborative tracking is employed. For the collaborative needs of 20 temperature-controlled zones, the edge computing units of each zone define coupling weights based on the adjacency relationships of the finite element model. Due to budget constraints on the total power of the on-site transformers, the system uses a distributed algorithm (such as the alternating direction multiplier method) for communication game theory. Each unit continuously exchanges thermal interaction information and power multipliers at the boundaries via the local area network. Without centralized processing, a consensus on global heat distribution is reached through multiple iterations (typically converging after 10-20 iterations), thereby accurately tracking the temperature control targets for each zone issued by the optimization layer. Each zone then uses the converged first-step control quantity... The signal is sent to the local actuator (high-frequency power supply) to drive the induction coil to perform heating.

[0083] The beneficial effects of this embodiment are that by introducing a distributed cooperative tracking architecture, the computational pressure is decentralized, significantly improving the system's real-time response speed and redundancy. Even if short-term communication fluctuations occur in the edge computing units of individual partitions, the system can still maintain the smoothness of the thermal field distribution by relying on neighborhood consistency weights. This mechanism successfully transforms the complex 20-partition power coordination problem into efficient local iterative calculations, ensuring that the 30-meter large-scale component can strictly follow the preset temperature rise curve in frigid environments, providing a final, precise execution guarantee for eliminating concrete temperature cracks.

[0084] In one possible implementation, the execution of control commands refers to the system applying the optimal power commands for each partition obtained through distributed collaborative solution. The signal is converted into a high-frequency pulse width modulation (PWM) signal in real time. This signal, driven by a power electronic module, adjusts the output duty cycle of the high-frequency induction power supply, thereby precisely regulating the heat generation of the electromagnetic induction coils embedded in 20 temperature-controlled zones. This step, through closed-loop feedback control, ensures that the actual temperature rise trajectory inside the concrete strictly follows the reference curve issued by the optimization layer.

[0085] Regarding the setting of the implementation environment and performance indicators, the system is designed for an ambient temperature of... to In extreme operating conditions of strong winds and severe cold, an adaptive "learning phase" of approximately one hour is set after system startup to enable online estimation algorithms and quickly correct initial material parameter values, thereby stabilizing the core temperature control range during the stable control phase. High-precision control target.

[0086] In practical applications, this step demonstrates the system's superior control performance under extreme weather conditions. At the construction site, the system, through the decision-making of the preceding steps, can predict in advance the impact of environmental disturbances on the corners and edges. Even in the event of an unexpected instantaneous level 8 gale, the DR-MPC controller can ensure the uniformity of the internal thermal field of large-scale components by increasing the compensation power of the windward side zones, thus eliminating the risk of cracking caused by temperature stress from a physical level.

[0087] In practical implementation, the final control command is executed. The system converts the digital power commands for each zone obtained through collaborative solving into high-frequency pulse width modulation (PWM) signals, driving the electromagnetic induction power supply to precisely regulate the heat generation of the induction coils in each zone. Under extreme operating conditions, the system can complete initialization through a learning phase of approximately one hour, subsequently entering a stable control state and uniformly raising the core temperature of the box girder to the preset range within eight hours. Actual measurements show that the internal and external temperature difference was successfully suppressed throughout the entire process. Within, better Safety constraints and limitations.

[0088] The beneficial effect of this embodiment lies in achieving millimeter-level temperature difference control accuracy for a 30-meter-long box girder through deep coupling of high-frequency PWM precision control and a distributed architecture. Compared to traditional fixed-power maintenance schemes, this system achieves energy savings of approximately 35% by allocating heat energy on demand while ensuring structural safety. This closed-loop mechanism based on "mathematical optimization-physical execution" provides highly reliable equipment support for the construction of complex structures in frigid regions.

[0089] In one possible implementation, the knowledge accumulation and meta-learning evolution refer to the system automatically initiating the fifth layer of "self-evolution and learning" logic after the 72-hour maintenance task concludes. The system aggregates the high-dimensional time-series dataset generated throughout the entire process. Using offline analysis and Bayesian optimization techniques, the meta-prior distribution of the underlying physical model was analyzed. and the controller's hyperparameters Update the system to enable lifelong learning and cross-scenario migration of control strategies.

[0090] In the mathematical correction process of the aforementioned knowledge accumulation, the system uses the measured data to update the pre-exponential factor of hydration kinetics. With activation energy The meta-prior distribution is used, and the energy consumption weight coefficients of the third-layer DR-MPC are combined with the Bayesian optimization algorithm. (For example, fine-tuning from 0.01 to 0.008) to pursue higher energy efficiency. Finally, the updated model parameters and hyperparameters are labeled as "Railway Prestressed Box Girder C50 Concrete - Severe Cold Environment" and stored in the digital knowledge base.

[0091] In practical applications, this step endows the maintenance system with the ability to "self-evolve," solving the pain point of traditional control algorithms requiring repeated manual parameter adjustments when facing different working conditions. As construction tasks accumulate, the digital knowledge base can automatically identify the similarities between new and historical projects, thereby achieving "second-level startup" and "rapid migration" of the controller. This mechanism transforms single construction experience into reusable mathematical assets, significantly reducing the debugging costs of similar projects in the future.

[0092] In practice, after the task is completed, the system also performs knowledge accumulation. After the 72-hour maintenance task is finished, the system automatically summarizes nearly a thousand sets (specifically 864 sets) of data generated throughout the process, updates the hydration model parameters and controller weights through offline analysis and Bayesian optimization, and stores the construction experience in a digital knowledge base. This self-evolutionary mechanism provides a referable intelligent control solution for future railway box girder projects in similar frigid regions.

[0093] The beneficial effect of this embodiment lies in breaking the limitations of traditional "one-time control" in maintenance by introducing a meta-learning framework, achieving a step-by-step improvement in algorithm performance. The system can learn from each construction error, continuously compressing the uncertainty range of the model, making subsequent projects more robust and energy-efficient. This mechanism marks a leap from "automation" to "true intelligence" in winter construction maintenance, laying the core algorithmic foundation for building a digital twin of a smart construction site.

[0094] In one overall embodiment, see Figure 2 This solution is used in a winter construction electromagnetic curing system for reinforced concrete based on a hierarchical progressive mathematical control architecture. This system deeply integrates mathematical algorithms with physical equipment through a closed-loop logic of "perception-decision-execution." The specific architecture is described below:

[0095] 1. Concrete components for winter construction

[0096] The concrete components used in winter construction are the controlled physical entities of the system, specifically zoned reinforced concrete beams / slabs deployed in frigid environments. These components are logically divided into multiple temperature-controlled zones, serving both as the physical carrier of thermal field evolution and intensity growth throughout the curing cycle, and as the source of sensing signals and the ultimate target of control commands.

[0097] 2. Multi-source sensor module

[0098] The multi-source sensor module is deployed in the component and its construction environment to collect raw sensing signals throughout the entire maintenance cycle, including internal temperature, hydration level, maturity, and environmental disturbance data. This module converts the physical state into a high-frequency digital signal stream, providing high-fidelity input support for the subsequent mathematical control architecture layer, ensuring that the system can perceive the multidimensional state evolution of the physical entity in real time and accurately.

[0099] 3. Hierarchical and progressive mathematical control architecture

[0100] As the system's "computing hub," this architecture consists of three nested core functional modules, implementing five layers of logic from mechanism simulation to intelligent evolution:

[0101] The physics modeling and order reduction prediction module is responsible for storing the zeroth layer multiphysics coupling mathematical model and using intrinsic orthogonal decomposition (POD) technology to construct a first-layer high-fidelity "grey box" prediction model with real-time computing capabilities.

[0102] Online joint sensing and estimation module: It adopts a double-ring nested Bayesian inference mechanism to perform a second-level online joint estimation and synchronous correction of the state and parameters of the "gray box" prediction model using the original sensing signals.

[0103] Distributed Cooperative Optimization and Evolution Module: Integrates distributed robust optimization unit, distributed cooperative tracking unit and meta-learning unit, used to solve the third-layer robust control strategy and the fourth-layer distributed cooperative trajectory under multi-objective constraints, and realize the lifelong optimization of the fifth-layer system performance based on historical data.

[0104] 4. Electromagnetic induction heating device

[0105] The electromagnetic induction heating device, as the system's actuator, includes a multi-zone power regulation unit responsible for receiving digital commands output by the distributed collaborative optimization and evolution module. This device drives induction coils embedded in the component using high-frequency pulse width modulation (PWM) technology, converting digital power commands into zoned induced heat energy to achieve differentiated thermal compensation for each temperature-controlled zone, thereby ensuring the uniformity of the concrete's thermal field distribution.

[0106] 5. System Collaborative Operation Characteristics and Technical Effects

[0107] This solution constructs a closed loop of "perception-decision-execution" through the aforementioned modules. The system utilizes a mathematical architecture layer for real-time simulation and dynamic calibration of physical entities, guiding the execution mechanism to accurately allocate spatial power. This architecture achieves closed-loop dynamic constraints on the internal and external temperature differences of concrete components under extreme cold conditions, significantly reducing curing energy consumption while ensuring construction quality and safety, and enabling the digital accumulation of curing experience.

[0108] Corresponding to the above method embodiments, this specification also provides an embodiment of a mathematical control architecture device for a winter construction reinforced concrete electromagnetic curing system based on a hierarchical progressive mathematical control architecture. Figure 3 This specification illustrates a schematic diagram of a mathematical control architecture device for a winter construction reinforced concrete electromagnetic curing system based on a hierarchical progressive mathematical control architecture, according to an embodiment of this specification. Its core logic consists of the following three nested functional levels:

[0109] I. Physical modeling and order reduction prediction layer, including:

[0110] Physics Modeling Module 201 provides the first-principles foundation for the entire control system. The state of the controlled object is described by the following strongly coupled nonlinear partial differential equations (PDEs):

[0111] 1. Electromagnetic field-thermal field coupling equations (considering eddy current skin effect and temperature-dependent conductivity):

[0112]

[0113]

[0114] in, The electric field intensity is a complex vector. Permeability, The electrical conductivity of concrete is given by temperature. and hydration degree The function, Angular frequency, This represents the spatial distribution of the electromagnetic heat source term.

[0115] 2. Transient heat transfer equations involving phase change and heat release during hydration:

[0116]

[0117]

[0118] in, For density, This is the equivalent specific heat capacity (considering the latent heat of the ice-water phase transition). Thermal conductivity, For the final heat of hydration, This is the latent heat of phase transition of pore water ice, which is the saturation. The function.

[0119] 3. Hydration kinetic equations based on activation energy theory:

[0120]

[0121] in, Pre-exponential factor, This is a reaction mechanism function (such as the Krstulovic-Dabic model). The apparent activation energy is a function of the degree of hydration. is the gas constant.

[0122] 4. Thermal stress field equations (used for constraint analysis):

[0123]

[0124]

[0125]

[0126] in, For stress tensor, The elasticity matrix is ​​the elastic modulus. (Related to hydration degree) and Poisson's ratio The function, For strain tensor, For thermal strain, is the coefficient of thermal expansion.

[0127] The high-fidelity downgraded "gray box" module 202 transforms the continuous PDEs of the first-layer module into a parameterized state-space model suitable for real-time control calculations.

[0128] Spatial discretization: The control volume is meshed using the finite element method (FEM) or the finite volume method (FVM). State vectors are defined. , containing all nodes / units Value. Control input. This represents the power of each electromagnetic heating zone. Interference vector. .

[0129] Parameterized state-space model:

[0130]

[0131]

[0132] in, It is a discretized nonlinear state transition function, the specific form of which is given by discretized PDEs. It is a key time-varying parameter vector, for example: These parameters are explicitly modeled as state-dependent functions, such as Its coefficients are updated online. The process noise and measurement noise are respectively, and both follow a Gaussian distribution. , For the observation matrix, The sensor output is measured.

[0133] II. Online Joint Sensing and Estimation Layer, including:

[0134] Online joint estimation module 203: This module dynamically updates the system status by fusing real-time measurement data. and key parameters This understanding addresses model uncertainty. A dual estimation framework is employed:

[0135] Outer layer: parameters The recursive Bayesian estimation is used. Assuming the parameters are slowly time-varying, recursive maximum a posteriori estimation or particle filtering is employed for tracking.

[0136]

[0137] in, It is a parametric evolution model (such as random walk). By continuously using new data... Update the posterior probability density function (PDF) of the parameters to obtain the optimal estimate of the parameters. and its uncertainty covariance .

[0138] Inner layer: State Estimate the current parameters. Under these conditions, extended Kalman filtering or unscented Kalman filtering is used for state estimation.

[0139]

[0140]

[0141]

[0142]

[0143]

[0144] This dual estimation loop ensures that the prediction model can adapt to changes in material properties and the environment, always keeping pace with the physical entity.

[0145] III. Distributed Cooperative Optimization and Evolutionary Layer, including:

[0146] The sub-Bruker optimization module 204 performs forward-looking optimization decisions based on the updated model from the high-fidelity reduced-order "grey box" module and the state and parameter estimates and their uncertainties provided by the online joint estimation module. To handle uncertainty, the sub-Bruker optimization framework is employed.

[0147] Uncertainty set modeling: incorporating disturbances and parameters The uncertainty is modeled as a fuzzy set The uncertainty set of the disturbance can be generated by a set of scenarios from a deep generative model (such as a conditional variational autoencoder CVAE) trained on historical meteorological data. The uncertainty of parameters is described by their estimated confidence intervals. To describe.

[0148] The multi-objective partial bar optimization problem involves solving the following finite-time optimization problem in each control period k:

[0149]

[0150] in, The control sequence to be optimized. Corresponding to performance target The worst interference scenario, Value at risk (VaR) is used to measure risk in the worst-case scenario. In probabilistic tail scenarios, the average degree of violation of the temperature difference constraint. This represents a probability constraint, a chance constraint, requiring the temperature difference constraint to at least... The probability satisfies, The acceptable risk level is used; this problem is a complex stochastic nonlinear programming problem. In solving it, scenario approximation can be used to approximate the continuous probability distribution as a discrete set of scenarios, or convex approximation and sequence programming methods can be used for efficient solution. The optimization result is the optimal power sequence in the future control time domain, but only the first step is implemented. .

[0151] The distributed collaborative tracking module 205 addresses the issue that centralized DR-MPC computation is too burdensome for large structures. This module decomposes the global optimization results of the distributed bar optimization module into local tracking problems for each partition, and achieves efficient collaboration through a distributed optimization algorithm.

[0152] Objective decomposition: Upper-level optimization provides a global reference trajectory and total power budget Definition of the first The local cost function for each partition is:

[0153]

[0154] in, Indicates control inputs for other partitions, It is a partition The neighborhood set, It is a coupling weight matrix, which reflects the intensity of thermal interaction.

[0155] Cooperative Game Theory and ADMM Solution: The control problem of each partition forms a coupled Nash game. A distributed solution is performed using the alternating direction multiplier method. Each partition... In the iteration step Parallel solution:

[0156]

[0157] in, It is a copy of the global variable. It is a dual variable (Lagrange multiplier). It is a penalty parameter. By exchanging a small amount of boundary temperature prediction information and multiplier information between adjacent partitions, after several iterations, all partitions can converge to a Nash equilibrium solution that satisfies the global budget constraint, achieving efficient collaboration.

[0158] Meta-learning evolution module 206: This module uses historical data to continuously optimize the models and strategies of each underlying layer, achieving lifelong performance improvement of the system.

[0159] Meta-learning of model parameters: Utilizing the posterior distribution of parameters obtained through online Bayesian inference. As a meta-prior. When facing a new project, utilize a small amount of new data. Perform meta-updates to quickly obtain parameter posteriors adapted to new scenarios. This enables rapid transfer learning.

[0160] Controller parameter hyperparameter optimization: optimizing the controller's weight matrix. Prediction Time Domain Risk level etc. as hyperparameters Define a long-term performance metric. (e.g., total energy consumption + final intensity uniformity), using Bayesian optimization or reinforcement learning frameworks, in a digital twin simulation environment... Perform offline or online optimization:

[0161]

[0162] in, These represent different construction scenarios (such as different environments and different mix proportions). The optimized hyperparameters can be updated online or used as initial values ​​for the next similar project.

[0163] The above is a schematic scheme of the mathematical control architecture device for a winter construction reinforced concrete electromagnetic curing system based on a hierarchical progressive mathematical control architecture, according to this embodiment. It should be noted that the technical solution of this mathematical control architecture device for a winter construction reinforced concrete electromagnetic curing system based on a hierarchical progressive mathematical control architecture belongs to the same concept as the technical solution of the aforementioned winter construction reinforced concrete electromagnetic curing method and system based on a hierarchical progressive mathematical control architecture. Details not described in detail in the technical solution of the mathematical control architecture device for a winter construction reinforced concrete electromagnetic curing system based on a hierarchical progressive mathematical control architecture can be found in the description of the aforementioned technical solution of the winter construction reinforced concrete electromagnetic curing method and system based on a hierarchical progressive mathematical control architecture.

[0164] Corresponding to the above method embodiments, this specification also provides an embodiment of a hardware deployment architecture for a winter construction reinforced concrete electromagnetic curing system based on a hierarchical progressive mathematical control architecture. Figure 4 This specification illustrates a schematic diagram of the hardware deployment architecture of an electromagnetic curing system for reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture, according to one embodiment of this specification. The architecture includes:

[0165] The central server 301 serves as the system's global strategic decision-making center and meta-learning evolution platform. Its key features include: an integrated global optimization DR-MPC module, a meta-learning engine, and a long-cycle data management unit; responsible for constructing a set of disturbance uncertainties based on a scenario generation model driven by historical meteorological big data; generating a globally optimal temperature control reference trajectory that balances energy efficiency and cracking risk by solving an objective function that includes conditional value at risk (CVaR); and utilizing meta-learning algorithms to iterate online on the model's meta-prior distribution under different construction scenarios, enabling cross-project self-evolution and rapid migration of the system's control strategy.

[0166] The distributed collaborative control layer 302 serves as the networked logical hub and resource scheduling layer of the system. Located between global decision-making and local execution, this layer is characterized by: utilizing a distributed consensus game theory algorithm (such as ADMM) to decompose the global total power budget and reference trajectory issued by the central server into tasks; and establishing peer-to-peer communication links between edge computing units to achieve real-time exchange of thermal interaction information at the boundaries of each sub-section and iterative game theory of power commands, while satisfying global transformer capacity constraints. This enables balanced scheduling of heat across the entire field without relying on large-dimensional centralized computing.

[0167] The physical temperature control zone 303 serves as the system's underlying physical execution array and multi-dimensional sensing terminal. This physical temperature control zone comprises component areas logically divided into independent temperature control modules. Its key feature is that each zone is pre-embedded with a high-density multi-source sensor array and an electromagnetic induction coil matching its geometric characteristics. The sensor array is responsible for high-fidelity acquisition of core state parameters such as temperature, hydration degree, and porosity saturation, which dynamically change within the concrete during hydration. The electromagnetic induction coil, acting as a physical energy conversion interface, achieves differentiated thermal compensation for the physical entity of the component within that zone through efficient electro-magnetic-thermal conversion driven by a high-frequency induced current.

[0168] Edge computing unit 304 serves as the system's real-time response tactical node and local digital twin maintenance unit. The edge computing unit is deployed one-to-one with the physical temperature control zones. Its key features include: online denoising and feature extraction of local high-frequency sensing signals; and second-level online identification of time-varying physical property parameters (such as thermal conductivity correction coefficients and electromagnetic conversion efficiency) in the reduced-order physical model using a recursive Bayesian inference framework (such as UKF), maintaining strong synchronization between the local digital twin model and the physical entity. Simultaneously, based on the game equilibrium solution provided by the cooperative control layer, it generates refined actuator control commands through high-frequency pulse width modulation (PWM) technology, ensuring the system's rapid real-time response under sudden disturbance conditions such as extreme cold waves.

[0169] The beneficial effects of this embodiment lie in the fact that by linking the central server, the distributed collaborative control layer, the physical temperature control zones, and the edge computing units, high-dimensional mathematical control algorithms are transformed into physical execution instructions with millisecond-level response capabilities. Through distributed game theory logic and edge-side online identification technology, the "curse of dimensionality" in the maintenance of large-scale components is effectively overcome, and the risk of state distortion caused by strong environmental disturbances and material parameter drift in frigid environments is eliminated. In summary, this architecture constructs a high-fidelity closed-loop digital twin temperature control system through a progressive logic of "cloud-based strategic prediction - network tactical collaboration - edge real-time maintenance - bottom-level precise execution." While ensuring optimal allocation of power resources, it eradicates the hidden danger of component cracking at the hardware level and significantly improves maintenance efficiency, serving as the core equipment foundation for realizing intelligent construction of reinforced concrete.

[0170] The above is a schematic scheme of the hardware deployment architecture of a winter construction reinforced concrete electromagnetic curing system based on a hierarchical progressive mathematical control architecture provided in this embodiment. It should be noted that the technical solution of this hardware deployment architecture of the winter construction reinforced concrete electromagnetic curing system based on a hierarchical progressive mathematical control architecture belongs to the same concept as the technical solution of the above-described winter construction reinforced concrete electromagnetic curing method and system based on a hierarchical progressive mathematical control architecture. Details not described in detail in the technical solution of the hardware deployment architecture of the winter construction reinforced concrete electromagnetic curing system based on a hierarchical progressive mathematical control architecture can be found in the description of the technical solution of the above-described method embodiment.

[0171] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described method and system for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture.

[0172] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the above-described method and system for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the above-described method embodiment.

[0173] An embodiment of this specification also provides a computer program, wherein when the computer program is executed in a computer, the computer executes the steps of the above-described method and system for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture.

[0174] The above is an illustrative scheme of a computer program according to this embodiment. It should be noted that the technical solution of this computer program belongs to the same concept as the above-described method and system for electromagnetic curing of reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture. For details not described in detail in the technical solution of the computer program, please refer to the description of the technical solution of the above-described method embodiment.

[0175] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0176] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0177] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0178] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0179] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture, characterized in that, Includes the following steps: Offline modeling and initialization steps: A multi-physics coupling model is constructed based on the physical characteristics of the components, and a high-fidelity "gray box" prediction model and initialization parameters are constructed using POD technology; Multi-source data acquisition steps: Internal multi-physics data and external environmental interference data of each partition are collected in real time through a sensor array to provide digital input for the mathematical architecture; Real-time state and parameter synchronization estimation steps: A recursive Bayesian inference framework is adopted to dynamically correct model uncertainties through cyclic prediction and updating of parameters and states; Multi-objective sub-Bruker optimization steps: Global optimization is performed using the sub-Bruker model predictive control algorithm to generate temperature reference trajectories and optimal power prediction sequences; Distributed cooperative tracking steps: A distributed consensus game algorithm is used to solve the real-time heating commands of partitions that meet global budget constraints through interactive cooperation between edge units; Control command implementation steps: The electromagnetic induction heating device receives commands and drives the induction coil using high-frequency pulse width modulation technology to achieve differentiated thermal compensation for each partition.

2. The method for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture according to claim 1, characterized in that, In the offline modeling and initialization steps, the intrinsic orthogonal decomposition (POD) technique is used to project the multiphysics coupled partial differential equations onto the parameterized state space to achieve model order reduction, and to initialize the prior distribution and covariance matrix of the filter parameter vectors.

3. The method for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture according to claim 1, characterized in that, The multiphysics state data includes the temperature, hydration degree, and pore saturation of each node inside the concrete, which are collected in real time by a cluster of sensors; the external environmental interference data includes ambient temperature, wind speed, and solar radiation intensity.

4. The method for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture according to claim 1, characterized in that, The real-time state and parameter synchronization estimation step uses unscented Kalman filtering or extended Kalman filtering as an online identification algorithm to synchronously and jointly estimate the time-varying parameter vector composed of thermal conductivity correction coefficient, specific heat capacity correction coefficient, hydration mechanism function pre-exponential factor, and electromagnetic-thermal conversion efficiency; wherein, the rate of change of temperature over time is positively correlated with the electromagnetic heat source term and the hydration heat release rate, and negatively correlated with the equivalent specific heat capacity.

5. The method for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture according to claim 1, characterized in that, The sub-Bruker model predictive control algorithm constructs a set of disturbance uncertainty scenarios through a generative model driven by historical meteorological data, and constructs a set of parameter uncertainty by combining the confidence intervals of parameter estimation. Its objective function is a weighted sum of temperature tracking deviation term, total energy consumption term, temperature uniformity penalty term, and temperature difference constraint risk term based on conditional value of risk (CVaR).

6. The method for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture according to claim 1, characterized in that, The specific logic of the distributed collaborative tracking is as follows: Each edge computing unit abstracts the local temperature control task into a non-cooperative game model, and uses the alternating direction multiplier method (ADMM) to exchange the temperature prediction value and Lagrange multiplier information at the boundary with the neighboring unit through the local area network. Under the condition of satisfying the global total power budget constraint, the real-time power command of each partition actuator is iteratively solved.

7. The method for electromagnetic curing of reinforced concrete during winter construction based on a hierarchical progressive mathematical control architecture according to claim 1, characterized in that, The system performance lifetime meta-learning optimization steps include: updating the meta-prior distribution of model parameters using the control dataset of the entire maintenance process; and adaptively fine-tuning the controller's weight matrix, prediction time domain, and risk level hyperparameters based on long-term performance indicators, so as to realize the self-evolution and rapid migration of the control strategy for different construction scenarios.

8. A winter construction electromagnetic curing system for reinforced concrete based on a hierarchical progressive mathematical control architecture, characterized in that, include: Multi-source sensor module: used to collect real-time maintenance status data and environmental interference data of the reinforced concrete component; Electromagnetic induction heating device: used to perform thermal compensation on the reinforced concrete component; Control unit: Used to run a hierarchical, progressive mathematical control architecture to generate thermal compensation instructions, the architecture including: (1) Physical modeling and order reduction prediction module: used to construct a continuous medium mathematical model containing electromagnetic-thermal-mechanical coupling, and to construct a parameterized state space "gray box" prediction model using the intrinsic orthogonal decomposition algorithm; (2) Online joint sensing and estimation module: used to perform synchronous joint estimation of time-varying physical property parameters and system state in the "gray box" prediction model based on the recursive Bayesian inference framework; (3) Distributed Cooperative Optimization and Evolution Module: It is used to solve the optimal heat compensation command by using the multi-objective split-bar model predictive control algorithm, realize distributed cooperative solution by using game theory and consensus algorithm, and realize lifelong self-optimization of system performance by using meta-learning algorithm.

9. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions; The processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the electromagnetic curing method for reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture as described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer-executable instructions stored thereon, characterized in that, When executed by a processor, the computer-executable instructions implement the steps of the electromagnetic curing method for reinforced concrete in winter construction based on a hierarchical progressive mathematical control architecture as described in any one of claims 1 to 7.

Images