Nuclear power cooling tower structure safety analysis intelligent agent operation and maintenance method based on industrial physics AI engine

Abstract

The invention discloses a nuclear power cooling tower structure safety analysis intelligent agent operation and maintenance method based on an industrial physics AI engine, relates to the technical field of nuclear power, and provides a method for simulating concrete and double-layer two-way reinforcing steel bar layers by adopting a common-node multi-layer shell unit, and considering normal bias of each reinforcing steel bar layer relative to the middle surface of a tower wall; according to the method, three-dimensional solid modeling of a reinforced concrete structure is replaced with a traditional method, the number of degrees of freedom of the model is reduced on the premise that real structure composition is reflected, and in the aspect of material nonlinearity, a nonlinear damage model and a linear hardening model are adopted for simulating mechanical behaviors of concrete and steel bars respectively; the integration of the independent nonlinear constitutive model of the reinforced concrete, the superposition calculation of the total tangent stiffness matrix and the separation and extraction of the stress result of the reinforced concrete are realized, so that the solving efficiency and precision are improved, a more efficient finite element industrial engine is accessed to the AI intelligent operation and maintenance service flow of the nuclear power cooling tower, and the stability and damage condition of the cooling tower are intuitively predicted.

Description

Technical Field

[0001] This invention relates to the field of nuclear power technology, specifically to an intelligent agent operation and maintenance method for structural safety analysis of nuclear power cooling towers based on an industrial physics AI engine. Background Technology

[0002] Natural draft cooling towers are crucial equipment in nuclear power plants, typically hyperbolic cylindrical reinforced concrete structures. They are used to transfer waste heat from the nuclear power plant to the atmosphere in the form of hot air through heat exchange on the surface of the packing material. Compared to seawater direct-flow cooling systems, they offer significant advantages such as stable efficiency, simple maintenance, substantial reduction in sea area usage, and efficient waste heat recovery from wastewater discharge. With the evolution of nuclear power reactor types, the improvement of nuclear power plant power generation efficiency, and my country's increasing emphasis on environmental protection, cooling towers under construction and planned are gradually becoming ultra-large, with a water spray area exceeding 10,000 square meters. Since cooling towers are thin-shell structures, their overall stability under wind load and self-weight is a key evaluation indicator for cooling tower design and operation. Calculations are usually based on linear elastic assumptions and use empirical formulas from GB / T 50102 "Code for Design of Industrial Circulating Water Cooling". However, as the height and diameter of cooling towers continue to increase, empirical formulas may no longer be applicable. Therefore, industrial simulation engines for structural finite element stability analysis that can consider the nonlinearity of actual reinforced concrete structural materials are of great significance for the design and operation monitoring of ultra-large cooling towers. Nuclear power plant cooling towers are typically constructed of reinforced concrete, with double-layered, bidirectional steel reinforcement on both the inner and outer surfaces. This reinforcement layer enhances the ductility and tensile / compressive strength of the concrete by sharing and transferring external loads, preventing cracking. Patent CN118940563A, entitled "Optimization Calculation Method for Hyperbolic Natural Draft Cooling Tower Wall Thickness," discloses a method for optimizing the wall thickness of a hyperbolic natural draft cooling tower. This method includes: obtaining initial wall thickness data, shell curve equation, structural material parameters, and basic wind pressure of the cooling tower to be optimized; optimizing the initial wall thickness data based on the shell curve equation, structural material parameters, and basic wind pressure to obtain intermediate wall thickness data; performing finite element analysis on the intermediate wall thickness data to obtain shell internal force data; and optimizing the intermediate wall thickness data based on the shell internal force data to obtain the target wall thickness data. The patent's shortcomings lie in the fact that it uses a simplified shell model in the finite element calculation of the cooling tower wall and does not consider the nonlinear mechanical behavior of the steel reinforcement and concrete structure. The resulting shell internal forces may be overly conservative and unsuitable for the engineering design of ultra-large cooling towers. In the nonlinear simulation analysis of reinforced concrete structures of cooling towers, traditional simulation methods are mainly divided into two types: solid element modeling and single-shell modeling. Solid element modeling is a separate method, which uses solid elements to simulate concrete and uses shared-node surface elements or independently partitioned solid elements to simulate the reinforcement layer. It can more accurately calculate structural stress and local mechanical behaviors such as friction and slip at the reinforced concrete interface. However, it suffers from the drawbacks of difficult mesh generation and low computational efficiency. On the other hand, the single-shell modeling method usually combines and simplifies the reinforced concrete into a single layer of isotropic shell elements. It has simple mesh generation and high computational efficiency, but it cannot distinguish the stress state of the reinforcement or concrete separately, and it is difficult to simulate the real nonlinear mechanical behavior of reinforced concrete. Therefore, the traditional finite element simulation method has significant limitations in the design, operation and maintenance of ultra-large cooling towers or in intelligent overall stability monitoring systems based on digital twins. Summary of the Invention

[0003] This invention provides an intelligent operation and maintenance method for structural safety analysis of nuclear power cooling towers based on an industrial physics AI engine. This method can effectively solve the problem mentioned in the background that traditional finite element simulation methods have significant limitations in the design, operation and maintenance of ultra-large cooling towers or in intelligent overall stability monitoring systems based on digital twins.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine, specifically involving the following steps: Step S1: Based on the shape parameters or design drawings of the cooling tower, draw a geometric model including the ring base, support structure and tower wall, and divide it into first-order meshes. Copy the surface elements at the same position on the tower wall to simulate the steel reinforcement layer. Convert the original surface elements to second-order meshes and add center nodes to simulate concrete. Step S2: Select the most unfavorable operating condition of the cooling tower and directly calculate or use the finite element method to solve the temperature distribution on the inner and outer surfaces of the tower wall; Step S3: Considering the changes in cylinder wall thickness, reinforcement spacing and arrangement, and nonlinear material properties, constitutive model integration is performed on the nonlinear damage model of concrete and the linear hardening model of reinforcement respectively. Under the action of self-weight, wind pressure load and temperature, the nonlinear thermo-solid coupling equilibrium equation of the multi-layer shell model of reinforced concrete is solved by Newton's iteration method to calculate the stress field and concrete damage factor of the cooling tower reinforced concrete. Step S4: Calculate the local stability coefficient based on the stress results, gradually increase the wind pressure load, iteratively solve the problem in step S3 until the calculation diverges, obtain the critical wind pressure and wind speed, and calculate the overall stability coefficient of the cooling tower. Step S5: Encapsulate the physical simulation engine as a server in the form of Model Context Protocol (MCP) and integrate it into the AI ​​structural safety early warning module of the cooling tower digital twin system.

[0005] According to the above technical solution, step S1 specifically includes the following steps: S101, Draw the geometric model of the tower wall, support structure and ring base according to the shape parameters or design drawings of the cooling tower; S102, finite element mesh generation for cooling tower. Specifically, firstly, first-order mesh is generated for all geometric structures. Four layers of surface elements that overlap with the original elements and share nodes are copied at the same position on the tower wall to simulate the steel reinforcement layer. The original surface elements are converted into second-order elements and center nodes are added to simulate concrete.

[0006] According to the above technical solution, step S2 specifically includes the following steps: S201, selecting the most unfavorable operating condition of the cooling tower, calculates the temperature distribution on the inner and outer surfaces of the tower wall. Based on the actual engineering situation, and according to the simulation accuracy from low to high, there are three methods to obtain the temperature distribution on the inner and outer surfaces of the tower, as detailed below: Method 1: Calculate based on the design air temperature distribution inside the ventilation duct of a natural draft cooling tower and the formula for calculating the temperature difference between the inner and outer surfaces in the current design specifications; Method 2: Based on the actual monitoring data of temperature measuring points arranged on the inner and outer surfaces of the cooling tower, the temperature distribution of the inner and outer surfaces is fitted using the least squares method; Method 3: Based on the air temperature inside and outside the cooling tower and the convective heat transfer coefficient, the temperature field of the concrete structure of the cooling tower is solved using the finite element method based on a three-dimensional shell element steady-state heat transfer model.

[0007] According to the above technical solution, step S3 specifically includes the following steps: S301, calculate the right-hand term of the equilibrium equation for the reinforced concrete tower wall of a cooling tower under its own weight and wind pressure; S302, Based on the concrete temperature field obtained in step S1, calculate the right-hand term of the temperature load for subsequent displacement prediction calculation. S303, based on the elasticity assumption, displacement prediction is performed to obtain the initial displacement of the Newton iteration; S304 uses shell elements to simulate concrete and biased membrane elements to simulate double-layer bidirectional steel reinforcement. Considering the effect of temperature, the nonlinear damage model of concrete and the linear hardening model of steel reinforcement are constitutive integrals performed using analytical methods. S305, superimpose the stress fields of the concrete and steel reinforcement layers and update the displacement field. Calculate until convergence based on the Newton iteration method to obtain the displacement field of the cooling tower wall, the concrete stress field, and the damage distribution results.

[0008] According to the above technical solution, step S4 specifically includes the following steps: S401. Calculate the local stability coefficient of the cooling tower according to the current design specifications. When calculating the local stability coefficient, the circumferential stress of the cooling tower is the sum of the circumferential stress of the concrete and the circumferential stress of the inner and outer surface steel reinforcement layers. Similarly, the meridional stress of the cooling tower is the sum of the meridional stress of the concrete and the meridional stress of the inner and outer surface steel reinforcement layers. S402, amplify the wind pressure function corresponding to the equivalent wind load standard value in step S3. Then, perform nonlinear finite element solution according to the method in step S3; S403 increases the step size of the amplification factor. Halve the amount, return to step S401, until... Less than the preset threshold, the final result is This is the overall stability coefficient of the cooling tower.

[0009] According to the above technical solution, step S5 specifically includes the following steps: S501, the cooling tower stability and damage analysis physical simulation engine described in steps S2 to S4 is encapsulated as a server in the form of MCP, exposing callable interfaces for inputting parameters, performing nonlinear finite element simulation analysis, and returning result data. S502, in the nuclear power plant cooling tower twin system, uses the LangChain application development framework to build a structural safety analysis and early warning submodule composed of multiple large language model intelligent agents, forming an industrial AI engine for cooling tower stability and damage analysis.

[0010] According to the above technical solution, step S5 includes the following sub-module: Agent 1: Collects and analyzes cooling tower environmental and load monitoring data, triggered by questions or periodic commands from the power plant administrator; Agent 2: Nonlinear stability and damage analysis of cooling tower, receives ambient temperature and load information transmitted by Agent 1 and converts it into input parameters for the physics simulation engine; Agent 3: Performs structural safety evaluation and early warning for cooling towers, and receives the results of nonlinear stability and damage analysis of cooling towers transmitted by Agent 2.

[0011] According to the above technical solution, the intelligent agent 1 further includes a self-learning multi-source perception optimization module, which is used to perform real-time fusion and dynamic weight adjustment of multi-source monitoring data such as the temperature, wind pressure, humidity and meteorological parameters of the inner and outer surfaces of the cooling tower. This module compares historical data and current measurement results from each sensor over time to identify measurement point drift, abnormal fluctuations, and data loss, and automatically corrects the confidence weights of the corresponding sensors.

[0012] According to the above technical solution, the intelligent agent 2 further includes an AI-physics collaborative solution module, which combines deep learning-based approximate analysis methods with traditional finite element nonlinear calculation methods to form a hybrid solution mechanism with physical constraints and adaptive optimization capabilities.

[0013] According to the above technical solution, the intelligent agent 3 further includes a swarm intelligence reinforcement learning collaborative optimization module, which is used to establish a distributed decision-making and collaborative control mechanism among multiple cooling towers in a nuclear power plant to achieve global dynamic optimization of structural safety.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention is applied to the reinforced concrete structure of nuclear power cooling towers. Addressing the challenge of traditional finite element simulation methods for rapid stability analysis of ultra-large nuclear power cooling towers based on separate modeling, this invention proposes using multi-layer shell elements with shared nodes to simulate both concrete and double-layer bidirectional reinforcement. It also considers the normal offset of each reinforcement layer relative to the mid-surface of the tower wall, replacing the traditional method of three-dimensional solid modeling of reinforced concrete structures. This reduces the number of degrees of freedom in the model while reflecting the true structural composition. Regarding material nonlinearity, nonlinear damage and linear hardening models are used to simulate the mechanical behavior of concrete and reinforcement respectively. This enables the independent integration of the nonlinear constitutive model of reinforced concrete, the superposition calculation of the total tangent stiffness matrix, and the separate extraction of stress results for reinforced concrete, thereby improving the solution efficiency and accuracy of the finite element industrial engine. Furthermore, in practical industrial digital twin applications built with AI, the industrial AI engine is connected to the intelligent operation and maintenance business flow of nuclear power cooling towers to quickly assess the overall and local stability of in-service cooling towers using limited computing resources, thereby improving the reliability of the results, reducing conservatism, and intuitively predicting the area and degree of concrete damage. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0016] In the attached diagram: Figure 1 This is a flowchart illustrating the analysis engine of this invention; Figure 2 This is a schematic diagram of the multi-layer shell mesh model of the cooling tower of the present invention; Figure 3 This is a schematic diagram of the maximum damage factor of each height layer of the concrete in the cooling tower of the present invention. Figure 4 This is a schematic diagram showing the maximum stress of the circumferential steel reinforcement layer on the outer surface of each height layer of the cooling tower of the present invention. Figure 5This is a schematic diagram of the minimum local stability coefficient at each height level of the cooling tower of the present invention. Detailed Implementation

[0017] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0018] Example: This invention provides a technical solution, an intelligent operation and maintenance method for structural safety analysis of nuclear power cooling towers based on an industrial physics AI engine. A super-large high-level seawater cooling tower for water collection at a nuclear power plant in a southern coastal region is selected. A multi-layer shell model is used for nonlinear stability and damage analysis. The cooling tower is 218.7 meters high, has a zero-meter diameter of 174.5 meters, a throat height of 164 meters, a throat diameter of 89.6 meters, and a single-tower water spray area of ​​20,000 square meters. The thickness of the cooling tower decreases linearly from 2 meters at the bottom to 0.39 meters at the throat, and then increases to 0.8 meters at the top. Circumferential and meridional reinforcing steel layers are arranged near the inner and outer surfaces, with a spacing of 0.2 meters and a diameter of 20 millimeters. The cooling tower cylinder is constructed of C50 concrete. The cooling tower support structure consists of 48 pairs of X-columns. Based on GB / T 50102 "Code for Design of Industrial Circulating Water Cooling", the basic wind pressure at the nuclear power plant site is calculated to be 0.8 kPa. like Figure 1 As shown, the specific steps involved are as follows: Step S1: Based on the shape parameters or design drawings of the cooling tower, draw a geometric model including the ring base, support structure and tower wall, and divide it into first-order meshes. Copy the surface elements at the same position on the tower wall to simulate the steel reinforcement layer. Convert the original surface elements to second-order meshes and add center nodes to simulate concrete. Step S2: Select the most unfavorable operating condition of the cooling tower, and directly calculate or use the finite element method to solve the temperature distribution on the inner and outer surfaces of the tower wall, which will be used for nonlinear thermo-mechanical coupling analysis in step S3; Step S3: Considering the changes in cylinder wall thickness, reinforcement spacing and arrangement, and nonlinear material properties, constitutive model integration is performed on the nonlinear damage model of concrete and the linear hardening model of reinforcement respectively. Under the action of self-weight, wind pressure load and temperature, the nonlinear thermo-solid coupling equilibrium equation of the multi-layer shell model of reinforced concrete is solved by Newton's iteration method to calculate the stress field and concrete damage factor of the cooling tower reinforced concrete. Step S4: Calculate the local stability coefficient based on the stress results, gradually increase the wind pressure load, iteratively solve the problem in step S3 until the calculation diverges, obtain the critical wind pressure and wind speed, and calculate the overall stability coefficient of the cooling tower. Step S5: Encapsulate the physics simulation engine as a server in the form of Model Context Protocol (MCP) and integrate it into the AI ​​structural safety early warning module of the cooling tower digital twin system. This provides simulation tool services for the large language model and enables physics-based intelligent structural safety analysis, early warning, and emergency response for cooling towers.

[0019] Based on the above technical solution, step S1 specifically includes the following steps: S101. Based on the shape parameters or design drawings of the cooling tower, draw the geometric model of the tower wall, supporting structure and annular base. Specifically, obtain the mid-surface radius and elevation of the hyperbolic cooling tower wall at the top, throat and bottom. Use B-spline curves to generate the tower wall contour curve passing through three feature nodes. Rotate it around the axis of symmetry to generate the mid-surface of the tower wall. Then use line segments to draw the geometry of the supporting structure and use surfaces to draw the geometry of the annular base. Finally, merge the annular base, supporting structure and tower wall, and eliminate overlapping parts to form the geometric model of the cooling tower. S102, Generate the finite element mesh for the cooling tower. Specifically, first, generate a first-order mesh for the entire geometric structure, selecting an appropriate mesh size so that the finite element model can accurately calculate the stress distribution in key areas such as the throat, tower base, and supporting structure. Figure 2 As shown, the cooling tower mesh model obtained by mesh generation contains a total of 5085 nodes and 9685 mesh elements. The tower wall is an unstructured mesh. The application of the multi-layer shell model significantly reduces the mesh size compared to the three-dimensional model. To simulate the double-layer bidirectional reinforcement on the inner and outer surfaces, four surface elements were replicated at the same location on the tower wall. The new surface elements overlapped with the original surface elements and shared nodes, thus enabling the superposition of the stiffness matrices of each layer of elements without increasing the number of degrees of freedom in the model. The original surface elements were converted into second-order elements and a central node was added. Subsequently, shell elements were used for concrete structural mechanics calculations. To improve the calculation accuracy, in addition to the translational and rotational degrees of freedom of the nodes on each side, the central node of the concrete shell element had a rotational degree of freedom, thus enabling a more accurate simulation of the curvature changes of the shell element.

[0020] Based on the above technical solution, step S2 specifically includes the following steps: S201, Selecting the most unfavorable operating condition of the cooling tower, calculate the temperature distribution on the inner and outer surfaces of the tower wall. In this specific embodiment, based on the air temperature inside and outside the cooling tower and the convective heat transfer coefficient, the finite element method is used to solve the temperature field of the concrete structure of the cooling tower. Specifically, assuming the concrete age is higher than 28 days, i.e., ignoring the heat release from hydration reaction, and considering only the convective heat transfer of the inner and outer walls, the weak form of the steady-state heat transfer equation of the cooling tower can be written as: ; in, Indicates the thermal conductivity of concrete. Represents the temperature field. Represents a virtual temperature field. This represents the computational domain, specifically the concrete of the tower wall. Indicates the inner and outer walls. Indicates the convective heat transfer coefficients of the inner and outer walls. The temperature represents the ambient temperature of the inner and outer walls. Based on actual monitoring and meteorological data, the internal temperature of the cooling tower in winter is 15 degrees Celsius, and the external temperature is 5 degrees Celsius. A shell element model considering the curvature of the mesh elements is used to solve the heat transfer equation. To improve the accuracy of the temperature field calculation, a third-order asymptotic approximation is used for the temperature field in the thickness direction, i.e., the shell element normal. Therefore, the temperature field to be solved can be expressed as: ; in, For normal coordinates The quadratic polynomial, The coordinates of the cooling tower's inner, outer, and inner surfaces relative to the shell element plane are respectively. Temperature field distribution; Take as: ; This allows the weak form of the heat transfer equation to be transformed into: ; in, , , These are three matrices characterizing the temperature distribution along the thickness direction of the shell elements. The matrix characterizes the in-plane heat transfer performance of the unit cell. The matrix characterizes the stiffness component of the shell element's heat transfer performance in the thickness direction and the convective heat transfer conditions of its inner and outer surfaces. The matrix represents the ambient temperature portion of the convective heat transfer conditions on the inner and outer surfaces. The specific calculation method is as follows: ; ; ; in, These represent the two radii of curvature in the local coordinate direction within the shell element plane. This represents the average curvature of the shell element. This represents half of the local thickness of the cooling tower wall. The thickness of the cooling tower wall is a function of elevation and can be selected based on actual design parameters or linearly interpolated between the thicknesses at the bottom, throat, and top of the tower. S201, the above processing of the temperature field and the steady-state heat transfer equation of concrete enables the spatial discretization and solution of the control equations based on the traditional finite element theory in the finite element industrial engine, based on the three-dimensional shell element model in the tangential plane of the tower wall, to simulate the three-dimensional steady-state heat transfer process and obtain the temperature field of the middle surface and inner and outer surfaces of the cooling tower wall for subsequent thermo-solid coupling calculations.

[0021] Based on the above technical solution, step S3 specifically includes the following steps: S301, calculate the right-hand side of the equilibrium equation load for the reinforced concrete tower wall of the cooling tower under its own weight and wind pressure. Specifically, for the calculation of the self-weight load, neglecting the rotational degree of freedom of the tower wall, calculate the virtual work of gravity for each finite element as follows: ; in, These represent the concrete unit, the four-layer steel reinforcement unit, and the virtual work done by gravity, and their sum, respectively. These represent the densities of the concrete and the reinforcing steel layer, respectively, and are each set to 2500. and 7850 , Represents gravitational acceleration. Number the nodes. These represent the interpolation shape functions for the translational degrees of freedom of the concrete and steel reinforcement layers, respectively. This represents the vertical virtual displacement of the node; For wind pressure load, the equivalent wind load standard value is calculated according to GB / T 50102 "Code for Design of Industrial Circulating Water Cooling". Assuming that the wind pressure only acts on the concrete element and neglecting rotational degrees of freedom, the virtual work of the wind pressure load on each finite element is calculated as follows: ; in, This represents the virtual work done by the wind pressure in the concrete unit. Represents the element calculation surface. This represents the equivalent wind load and wind pressure function. Represents the element normal vector. The shape function matrix represents the interpolation shape function matrix for the translational degrees of freedom of the concrete element. Indicates the virtual displacement of the node. The area element is represented by the equivalent wind load and wind pressure function, which is calculated as follows: ; in, For wind vibration coefficient, The inter-tower interference coefficient, The wind pressure distribution coefficient is calculated using the formula for ribless hyperbolic cooling towers. For vertical and circumferential coordinates, This is the wind pressure height variation coefficient. This is the basic wind pressure; Finally, the virtual work done by the self-weight load and the wind pressure load in each element is superimposed to obtain the right-hand side vector of the element load: ; in, The vector represents the right-hand side of the element load, which is assembled according to the element nodes to obtain the right-hand side of the load in the system equilibrium equation. S302, Based on the concrete temperature field obtained in step S1, calculate the right-hand term of the temperature load for subsequent displacement prediction calculations. Specifically, for concrete shell elements, the calculation method for the right-hand term of the temperature load is as follows: ; in, The compression modulus of concrete. The coefficient of thermal expansion is Celsius is a reference temperature. Let be the strain matrix, satisfying , Let U be the strain tensor and U be the displacement vector. As a unit array, The Gaussian point temperature interpolation is calculated as follows: ; in, For the local coordinates of the Gaussian point normal of the shell element, For half the thickness of the shell unit, Representing the average temperature field and temperature gradient field, which are independent of the normal coordinates, respectively, calculated by integrating the temperature field along the thickness direction: ; ; S303, based on the elasticity assumption, displacement prediction is performed to obtain the initial displacement of the Newton iteration. Specifically, the initial displacement is calculated as follows: ; in, Here is the elastic stiffness matrix. For the right-hand term of the element load The load right-hand side terms of the equilibrium equations of the assembled system; S304 uses shell elements to simulate concrete and biased membrane elements to simulate a double-layer bidirectional steel reinforcement layer. Considering the effect of temperature, constitutive integration is performed on the nonlinear damage model of concrete and the linear hardening model of steel reinforcement to obtain the Gaussian point stress results. Specifically, for the three-dimensional shell element of concrete, the nonlinear damage model is used to simulate the stiffness loss caused by tensile cracking of concrete and the mechanical behavior of stiffness recovery after compression. In the constitutive integration process, the mechanical strain input of the current iteration step is used, excluding thermal strain. The Gaussian point stress of the shell element is calculated as follows: ; in, The Lamé coefficient of the material. The trace of the strain tensor This is a step function; it takes the value 1 when the independent variable is greater than 0, and 0 otherwise. These represent the eigenvalues ​​of the stress and strain tensors, respectively. Damage factor, and variable within the model. Here are the model parameters, representing the ratio of concrete stiffness before and after damage. The damage factor is calculated as follows: ; in, The damage threshold represents the upper limit of the rate of decrease in the material's strain energy density during damage evolution. Let the damage factor at the current Gaussian point in the previous iteration be... The damage factor for the current iteration step is taken as The damage factor threshold is calculated based on the tensile strength of concrete: ; in, For compressive strength, For elastic modulus, Poisson's ratio; For the steel reinforcement layer, a linear hardening elastoplastic model is used to simulate the plastic yielding and hardening behavior of the steel. Two layers of orthogonal steel reinforcement with circumferential and meridional directions are laid on the inner and outer surfaces of the cooling tower, respectively. By considering the simulation using plate elements offset by half the wall thickness relative to the mid-surface of the tower wall, and based on the membrane element assumption and the assumption of mid-surface non-deformation, the normal strain and transverse shear strain are ignored, thus simplifying the mechanical simulation of the steel reinforcement layer into a plane strain problem. At the same time, considering that the arrangement direction of each layer of steel reinforcement is consistent, the constitutive model within each layer of steel reinforcement is simplified to a one-dimensional elastoplastic model, and its yield surface is expressed as: ; in, These represent the elastic modulus and the hardening modulus, respectively. , Indicates accumulated plastic strain, This represents the yield stress of the reinforcing steel. Based on the strain increment input of the current iteration step, excluding thermal strain, the stress of the reinforcing steel is first predicted based on the elastic assumption as follows: ; in, This represents the stress and elastic modulus of the previous iteration step. To represent the mechanical strain increment, let... To accumulate plastic strain for the previous iteration, if Then the incremental plastic strain accumulates. ,otherwise: ; To analyze whether the initial displacement or the current iteration step displacement satisfies the nonlinear convergence condition, i.e., the residual error is less than a given threshold, the relative residual is used for calculation: ; in, The system strain matrix, For the Gaussian point stress of the system, This represents the displacement result of the initial or nth Newton iteration. The residual threshold is used. If step S305 is not performed and the convergence condition is met, it indicates that neither the concrete nor the steel reinforcement layer has been damaged or has undergone plastic yielding. Step S4 is then performed directly. Otherwise, step S305 is performed to iteratively correct the predicted initial displacement. S305, superimposing the stress fields of the concrete and reinforcement layers and updating the displacement field, calculates to convergence based on the Newton-Raphson iteration method, obtaining the displacement field, concrete stress field, and damage distribution results of the cooling tower wall. Specifically, the stress fields of the concrete shell element and the four reinforcement layer elements are superimposed, and the displacement increment is solved: ; in, Indicates the first The displacement increment of the second Newton iteration. This represents the total tangent stiffness matrix after the concrete and steel reinforcement layers are stacked. Represents the strain matrix of the concrete layer. Indicates the first The Gaussian stress vector of the concrete layer in the next iteration. This represents the strain matrix of the i-th reinforcement layer, which is related to the reinforcement arrangement direction and the relative mid-plane offset. Indicates the first The Gaussian stress vector of the next iteration of the reinforcement layer Indicates the first The internal force vector of the support structure in the next iteration; Update the total displacement to Repeat steps S304 and S305 until the residual error is less than the set threshold. Extract the stress field of the intermediate concrete layer and each steel reinforcement layer, as well as the concrete damage factor, as the final output. The stress concentration area and high damage factor area of ​​the cooling tower cylinder are the weak areas of the cooling tower under the current working conditions. Under long-term operation conditions, there is a high risk of aging and cracking of reinforced concrete materials, which should be given special attention during operation and maintenance. The maximum values ​​of the damage factor of the cooling tower concrete at each height level are as follows: Figure 3 As shown, under the action of self-weight, standard wind load and temperature load, concrete begins to show tensile damage at a height of 125 meters, reaching the maximum value of 0.037 at the top of the tower, which is much less than 1. This indicates that even if some areas of concrete experience a small loss of stiffness under normal operating conditions, the risk of structural failure due to concrete tensile cracking is low. The maximum stress distribution along the elevation of the circumferential reinforcement layer on the outer surface of the cooling tower is as follows: Figure 4 As shown, under the external load mainly composed of wind pressure in the area from the throat to the top of the tower, the circumferential reinforcement exhibits a tensile stress of 1 MPa. This demonstrates the role of the reinforcement layer in helping the concrete to share and transfer the external load, thereby improving the overall tensile strength of the reinforced concrete material and effectively reducing concrete damage. Based on the above technical solution, step S4 specifically includes the following steps: S401, according to GB / T 50102 "Code for Design of Industrial Circulating Water Cooling", the local stability coefficient of the cooling tower is calculated by solving the quadratic polynomial equation: ; in, The local stability coefficient is... For circumferential and meridional stresses, For the critical stresses in the circumferential and meridional directions, when calculating the local stability coefficient, the circumferential stress of the cooling tower is the sum of the circumferential stress of the concrete and the circumferential stress of the inner and outer surface steel reinforcement layers. Similarly, the meridional stress of the cooling tower is the sum of the meridional stress of the concrete and the meridional stress of the inner and outer surface steel reinforcement layers. Minimum local stability coefficients at each height level of the cooling tower, as follows: Figure 5 As shown, due to the significant stress concentration between the tower bottom and throat, the stability coefficient is low, making it prone to local buckling instability. Since this case does not consider internal suction and the rigid ring at the top of the tower, the calculated value of the local stability coefficient is too low. The abnormal local stability coefficient at the top of the tower caused by the lack of a rigid ring is not considered. The local stability coefficients of the main body of the cooling tower are all higher than the threshold of 5.0 specified in GB / T 50102 "Code for Design of Industrial Circulating Water Cooling". S402, amplify the wind pressure function corresponding to the equivalent wind load standard value in step S3. If the Newton iteration converges or does not enter the Newton iteration stage, then proceed with the nonlinear finite element solution using the method described in step S3. Accumulation Repeat step S3 until the Newton iteration diverges; S403 increases the step size of the amplification factor. Halve the amount, return to step S401, until... Less than the preset threshold of 0.02, the final result is , which is the overall stability coefficient of the cooling tower; Nonlinear finite element calculation of cooling towers The divergence indicates that the overall stability coefficient of the cooling tower is 11.47, which is significantly higher than the safety threshold of 5.0 specified in GB / T 50102 "Code for Design of Industrial Circulating Water Cooling", and the cooling tower has no risk of overall buckling instability. Based on the above technical solution, step S5 specifically includes the following steps: S501, the physical simulation engine for cooling tower stability and damage analysis in steps S2 to S4 is encapsulated as a server in the form of MCP, exposing callable interfaces for inputting parameters, performing nonlinear finite element simulation analysis, and returning result data. The input parameters include ambient air temperature or inner and outer surface temperature of the tower, wind speed, and wind direction. The returned results include: the stress field of the intermediate concrete layer and each steel reinforcement layer, returned in the form of a data table containing grid node coordinates and stress components; the concrete damage factor, returned in the form of a data table containing grid node coordinates and damage factor; the local stability coefficient distribution of the cooling tower, returned in the form of a data table containing grid node coordinates and stability coefficient; and the overall stability coefficient of the cooling tower, returned in floating-point form. S502, in the nuclear power plant cooling tower twin system, uses the LangChain application development framework to build a structural safety analysis and early warning submodule composed of multiple large language model intelligent agents, forming an industrial AI engine for cooling tower stability and damage analysis.

[0022] Based on the above technical solution, step S5 includes the following sub-module: Agent 1: Cooling tower environment and load monitoring data acquisition and analysis. Triggered by questions or periodic instructions from the power plant administrator, it first performs semantic analysis and intent recognition to capture relevant instructions for cooling tower structural safety assessment. It then extracts sensor data from the internal and external environmental temperature measurement points, concrete temperature measurement points, and wind pressure measurement points of the cooling tower through the sensor data acquisition system interface, converts it into text information, and outputs it in a specific data structure. Agent 2: Nonlinear stability and damage analysis of cooling tower. It receives the ambient temperature and load information transmitted by Agent 1, converts it into input parameters for the physical simulation engine, calls the MCP simulation analysis service tool to obtain the cooling tower damage factor, local stability coefficient distribution and overall stability coefficient results and outputs them. Agent 3: Cooling tower structural safety evaluation and early warning. It receives the nonlinear stability and damage analysis results of the cooling tower transmitted by Agent 2, and combines the relevant information provided by the RAG component generated by the enhanced retrieval, which includes nuclear power design and operation and maintenance specifications and historical safety early warnings and operation and maintenance records of the cooling tower in the nuclear power plant. It summarizes and analyzes the simulation analysis results, forms the cooling tower safety risk evaluation conclusion, and provides risk warnings and emergency response plans based on specifications and historical information. In this case, the industrial AI engine for stability and damage analysis of the cooling tower digital twin system received the following user question: "Based on the power plant's ambient temperature and wind pressure data in February 2025, please provide an assessment of the structural damage and stability safety status of the cooling tower of Unit 1, and whether there is a risk of short-term structural failure?" The key conclusion returned by the intelligent agent is: "Based on the calculation and evaluation by the finite element physical simulation engine, in February 2025, under the action of its own weight, standard wind load and temperature load, the concrete of Unit 1 cooling tower began to show slight tensile damage at a height of 125 meters, reaching a maximum value of about 0.037 at the top of the tower. There is no risk of cracking or structural failure. The minimum local stability coefficient of the main body of the cooling tower is 6.0, and the overall stability coefficient is 11.47, both of which are higher than the threshold of 5.0 specified in GB / T 50102 'Code for Design of Industrial Circulating Water Cooling'. The cooling tower has no risk of buckling instability. In summary, there is no risk of structural failure of Unit 1 cooling tower in the short term."

[0023] According to the above technical solution, the intelligent agent 1 further includes a self-learning multi-source perception optimization module, which is used to perform real-time fusion and dynamic weight adjustment of multi-source monitoring data such as the temperature, wind pressure, humidity and meteorological parameters of the inner and outer surfaces of the cooling tower. This module compares historical data and current measurement results from each sensor over time to identify measurement point drift, abnormal fluctuations, and data loss, and automatically corrects the confidence weights of the corresponding sensors to achieve long-term stability and adaptive accuracy improvement of the monitoring input. In practical applications, the system first establishes a correlation matrix of the cooling tower's thermal-air field based on the spatial distribution characteristics of the sensors, and then dynamically updates its correlation coefficient by calculating the covariance relationship between different measuring points. Subsequently, based on the changing trend within the sliding time window, the coupling strength between changes in external meteorological conditions and internal response is determined, thereby automatically adjusting the sampling frequency and data filtering parameters. The self-learning capability of this module enables the monitoring system to continuously accumulate experience during the operating cycle. When new climate patterns or atypical load conditions appear, it automatically updates the sensing strategy and data fusion parameters, thereby ensuring the matching of the input model with the actual operating conditions. Through the above design, the cooling tower environment and load monitoring module can not only maintain stable output even when some sensors fail or there is noise interference, but also form an adaptive feature map during long-term operation, realizing intelligent evolution and reliability enhancement of input data, and providing higher quality physical boundary conditions for subsequent nonlinear finite element solutions.

[0024] According to the above technical solution, the intelligent agent 2 further includes an AI-physics collaborative solution module, which combines the deep learning-based approximate analysis method with the traditional finite element nonlinear calculation method to form a hybrid solution mechanism with physical constraints and adaptive optimization capabilities. During execution, the system first calls the finite element physical simulation engine to perform standard calculations on the stress state of the reinforced concrete structure of the cooling tower, and obtains the initial stress distribution and damage prediction results; at the same time, the AI ​​sub-engine establishes a nonlinear stress-strain mapping model based on the historical simulation sample library to quickly estimate the structural response under complex working conditions. When there is a discrepancy between the AI ​​sub-engine and the physical simulation results, the system automatically determines which type of result is more credible by calculating confidence weights and dynamically balances the contributions of the two, thereby maintaining simulation accuracy and significantly improving computational efficiency. In addition, during operation, the module can automatically trigger local mesh refinement based on the solved residuals, and recalculate the stress in the tower throat, the connection area of ​​the support structure and the high gradient region to reduce the local errors that may exist in traditional modeling. Through this collaborative approach of AI and physical simulation, the nonlinear stability analysis of cooling tower structures can be performed in near real-time without sacrificing accuracy, making it particularly suitable for continuous monitoring and dynamic risk assessment in digital twin platforms. This technical solution overcomes the limitations of traditional finite element methods, such as long computation time and sensitivity to input noise, enabling AI engines to have the dual advantages of physical interpretability and data-driven prediction capabilities in engineering applications.

[0025] According to the above technical solution, the intelligent agent 3 further includes a swarm intelligence reinforcement learning collaborative optimization module, which is used to establish a distributed decision-making and collaborative control mechanism among multiple cooling towers in a nuclear power plant to achieve global dynamic optimization of structural safety. This module treats each cooling tower as an independent intelligent agent node. Each node shares operating status parameters, local stability coefficients, concrete damage factors, and wind pressure and temperature change data through a communication network. The module defines the state, actions, and reward functions of each node through reinforcement learning algorithms and adaptively adjusts the strategies of each intelligent agent based on the global reward signal to achieve joint optimization of the cooling tower group. When the structural response of any tower exceeds the preset safety threshold, the corresponding intelligent agent automatically broadcasts the early warning information to the group and triggers the linkage control strategy at the group level, including wind pressure distribution redistribution, local tower load limiting, or maintenance scheduling optimization. During continuous operation, each agent achieves self-organized balance of the plant's safety parameters through game learning and gradient updates, thereby maximizing the overall stability margin of the group. This module can also combine historical operating data of nuclear power plants, design specifications and meteorological models for iterative learning, so that cooling tower groups can continuously optimize risk response capabilities and maintenance strategies during long-term operation. Through this mechanism, nuclear power plants can maintain the coordinated stability of multi-tower structures under extreme weather conditions, significantly improving system resilience and safety; In practice, the goal of group collaborative optimization can be expressed as: ; in, , Let be the local stability coefficient of the i-th cooling tower. , Its concrete damage factor, and These are weighting coefficients, which comprehensively consider stability and damage level to achieve optimal allocation and dynamic balance for group-level safety.

[0026] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for the operation and maintenance of nuclear power plant cooling tower structures based on an industrial physics AI engine, characterized in that: The specific steps involved are as follows: Step S1: Based on the shape parameters or design drawings of the cooling tower, draw a geometric model including the ring base, support structure and tower wall, and divide it into first-order meshes. Copy the surface elements at the same position on the tower wall to simulate the steel reinforcement layer. Convert the original surface elements to second-order meshes and add center nodes to simulate concrete. Step S2: Select the most unfavorable operating condition of the cooling tower and directly calculate or use the finite element method to solve the temperature distribution on the inner and outer surfaces of the tower wall; Step S3: Considering the changes in cylinder wall thickness, reinforcement spacing and arrangement, and nonlinear material properties, constitutive model integration is performed on the nonlinear damage model of concrete and the linear hardening model of reinforcement respectively. Under the action of self-weight, wind pressure load and temperature, the nonlinear thermo-solid coupling equilibrium equation of the multi-layer shell model of reinforced concrete is solved by Newton's iteration method to calculate the stress field and concrete damage factor of the cooling tower reinforced concrete. Step S4: Calculate the local stability coefficient based on the stress results, gradually increase the wind pressure load, iteratively solve the problem in step S3 until the calculation diverges, obtain the critical wind pressure and wind speed, and calculate the overall stability coefficient of the cooling tower. Step S5: Encapsulate the physical simulation engine as a server in the form of Model Context Protocol (MCP) and integrate it into the AI ​​structural safety early warning module of the cooling tower digital twin system.

2. The method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine according to claim 1, characterized in that: Step S1 specifically includes the following steps: S101, Draw the geometric model of the tower wall, support structure and ring base according to the shape parameters or design drawings of the cooling tower; S102, finite element mesh generation for cooling tower. Specifically, firstly, first-order mesh is generated for all geometric structures. Four layers of surface elements that overlap with the original elements and share nodes are copied at the same position on the tower wall to simulate the steel reinforcement layer. The original surface elements are converted into second-order elements and center nodes are added to simulate concrete.

3. The method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine according to claim 1, characterized in that: Step S2 specifically includes the following steps: S201, selecting the most unfavorable operating condition of the cooling tower, calculates the temperature distribution on the inner and outer surfaces of the tower wall. Based on the actual engineering situation, and according to the simulation accuracy from low to high, there are three methods to obtain the temperature distribution on the inner and outer surfaces of the tower, as detailed below: Method 1: Calculate based on the design air temperature distribution inside the ventilation duct of a natural draft cooling tower and the formula for calculating the temperature difference between the inner and outer surfaces in the current design specifications; Method 2: Based on the actual monitoring data of temperature measuring points arranged on the inner and outer surfaces of the cooling tower, the temperature distribution of the inner and outer surfaces is fitted using the least squares method; Method 3: Based on the air temperature inside and outside the cooling tower and the convective heat transfer coefficient, the temperature field of the concrete structure of the cooling tower is solved using the finite element method based on a three-dimensional shell element steady-state heat transfer model.

4. The method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine according to claim 2, characterized in that: Step S3 specifically includes the following steps: S301, calculate the right-hand term of the load in the equilibrium equation of the reinforced concrete tower wall of the cooling tower under its own weight and wind pressure; S302, Based on the concrete temperature field obtained in step S1, calculate the right-hand term of the temperature load for subsequent displacement prediction calculation. S303, based on the elasticity assumption, displacement prediction is performed to obtain the initial displacement of the Newton iteration; S304 uses shell elements to simulate concrete and biased membrane elements to simulate double-layer bidirectional steel reinforcement. Considering the effect of temperature, the nonlinear damage model of concrete and the linear hardening model of steel reinforcement are constitutive integrals performed using analytical methods. S305, superimpose the stress fields of the concrete and steel reinforcement layers and update the displacement field. Calculate until convergence based on the Newton iteration method to obtain the displacement field of the cooling tower wall, the concrete stress field, and the damage distribution results.

5. The method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine according to claim 4, characterized in that: Step S4 specifically includes the following steps: S401. Calculate the local stability coefficient of the cooling tower according to the current design specifications. When calculating the local stability coefficient, the circumferential stress of the cooling tower is the sum of the circumferential stress of the concrete and the circumferential stress of the inner and outer surface steel reinforcement layers. Similarly, the meridional stress of the cooling tower is the sum of the meridional stress of the concrete and the meridional stress of the inner and outer surface steel reinforcement layers. S402, amplify the wind pressure function corresponding to the equivalent wind load standard value in step S3. Then, perform nonlinear finite element solution according to the method in step S3; S403 increases the step size of the amplification factor. Halve the amount, return to step S401, until... Less than the preset threshold, the final result is This is the overall stability coefficient of the cooling tower.

6. The method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine, as described in claim 5, is characterized in that: Step S5 specifically includes the following steps: S501, the cooling tower stability and damage analysis physical simulation engine described in steps S2 to S4 is encapsulated as a server in the form of MCP, exposing callable interfaces for inputting parameters, performing nonlinear finite element simulation analysis, and returning result data. S502, in the nuclear power plant cooling tower twin system, uses the LangChain application development framework to build a structural safety analysis and early warning submodule composed of multiple large language model intelligent agents, forming an industrial AI engine for cooling tower stability and damage analysis.

7. The method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine according to claim 6, characterized in that: Step S5, the submodule includes: Agent 1: Collects and analyzes cooling tower environmental and load monitoring data, triggered by questions or periodic commands from the power plant administrator; Agent 2: Nonlinear stability and damage analysis of cooling tower, receives ambient temperature and load information transmitted by Agent 1 and converts it into input parameters for the physics simulation engine; Agent 3: Performs structural safety evaluation and early warning for cooling towers, and receives the results of nonlinear stability and damage analysis of cooling towers transmitted by Agent 2.

8. The method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine according to claim 7, characterized in that: The intelligent agent 1 further includes a self-learning multi-source perception optimization module, which is used to perform real-time fusion and dynamic weight adjustment of multi-source monitoring data such as the temperature, wind pressure, humidity and meteorological parameters of the inner and outer surfaces of the cooling tower. This module compares historical data and current measurement results from each sensor over time to identify measurement point drift, abnormal fluctuations, and data loss, and automatically corrects the confidence weights of the corresponding sensors.

9. The method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine according to claim 7, characterized in that: The intelligent agent 2 further includes an AI-physics collaborative solution module, which combines deep learning-based approximate analysis methods with traditional finite element nonlinear calculation methods to form a hybrid solution mechanism with physical constraints and adaptive optimization capabilities.

10. The method for the operation and maintenance of a nuclear power plant cooling tower structure safety analysis intelligent agent based on an industrial physics AI engine according to claim 7, characterized in that: The intelligent agent 3 further includes a swarm intelligence reinforcement learning collaborative optimization module, which is used to establish a distributed decision-making and collaborative control mechanism among multiple cooling towers in a nuclear power plant to achieve global dynamic optimization of structural safety.

Images