A method and system for calculating reliability of a post-fire steel-concrete composite beam

Abstract

The application discloses a kind of fire steel-concrete composite beam reliability calculation method and system, belong to structure fire safety and intelligent evaluation technical field, including: six steps of steel-concrete composite beam data acquisition after fire, material residual performance modeling, bending and interface slip coupling analysis, random variable modeling, reliability solution and intelligent evaluation output.By establishing the temperature-performance degradation model of steel, concrete and bolt connector, the residual bearing capacity and slip limit are obtained by coupling analysis;Using first-order reliability method and deep learning agent model to quickly calculate reliability index and failure probability, output safety level and load limit suggestion.The system is developed based on Python platform, integrates data input, model calculation and visualization module, and can realize real-time reliability intelligent evaluation of steel-concrete composite beam in laboratory fire test, with high precision and high efficiency.

Description

Technical Field

[0001] This invention relates to a method and system for calculating the reliability of steel-concrete composite beams after a fire, belonging to the field of structural fire safety and intelligent assessment technology. Background Technology

[0002] Steel-concrete composite beams are widely used in highway bridges, urban elevated roads, factory floors, and super high-rise structural systems due to their high strength, high stiffness, and excellent synergistic performance. In service environments, fire is one of the most destructive sudden loads. Fire causes a rapid temperature rise in steel beams, significantly reducing their elastic modulus and yield strength; the concrete interior experiences spalling and cracking due to dehydration, increased pore pressure, and concentrated thermal stress; and the interface welds and stud areas become zones of temperature gradient and shear concentration. These factors combined result in significant stiffness degradation and residual deformation in steel-concrete composite beams after a fire. If they are put back into service without proper evaluation, they are highly susceptible to secondary structural instability or brittle failure.

[0003] Current post-fire structural safety assessments mainly rely on: (1) the standard empirical method: calculating the residual bearing capacity based on the "temperature-strength reduction factor" specified in the standard. This method is simple but only applicable to standard components, and cannot consider complex structural boundaries, local spalling and non-uniform temperature distribution. The assessment results are conservative and have low accuracy; (2) the numerical simulation method: obtaining the residual stress and deformation distribution after the fire through thermo-mechanical coupled finite element analysis, which can reflect the structural behavior more realistically. However, the calculation process is complex, has many parameters, and takes a long time, making it difficult to meet the needs of emergency decision-making; (3) the detection regression method: estimating the bearing capacity by combining on-site detection data and empirical regression formulas. It lacks a unified model and probability theory basis, and the results are highly discrete and lack reliability significance.

[0004] With the development of artificial intelligence and deep learning, surrogate models have become an important direction for rapid prediction of structural performance. However, existing research mainly focuses on components at room temperature or with single materials, and research on intelligent reliability assessment of steel-concrete composite structures after fire is still in its early stages. Addressing the combined effects of coupled material degradation, spatial randomness, and multiple failure modes after fire, there is an urgent need to propose a unified assessment method and system that integrates physical mechanisms, probabilistic reasoning, and intelligent prediction capabilities to achieve quantitative, automated, and real-time determination of the reliability of composite beams after fire. Summary of the Invention

[0005] The purpose of this invention is to overcome the problems of low accuracy, long calculation time, and insufficient intelligence in the existing technology of post-fire structural safety assessment, and to provide a method and system for calculating the reliability of steel-concrete composite beams after a fire.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for calculating the reliability of steel-concrete composite beams after a fire includes:

[0008] Collect parameters of steel-concrete composite beams, standardize units, preprocess them, and export standardized input files;

[0009] A staged model of furnace gas temperature change was established. The temperature field was obtained by solving the unsteady heat conduction equation, and the boundary conditions were set by superimposing convection and radiation.

[0010] A material temperature-performance degradation model after a fire is constructed to calculate material parameters; a bending-slip coupling model is constructed and coupled analysis is performed to obtain the residual bearing capacity and slip limit.

[0011] Construct a set of random variables as independent variables and the limit state value as dependent variable to construct the limit state function of the composite beam;

[0012] The failure probability and reliability index are calculated using the first-order reliability method. High-fidelity samples are generated and added to the training sample pool. The process is iterated until the preset convergence requirement is met.

[0013] A deep learning proxy model is constructed. After the model is fully trained, the parameters of the steel-concrete composite beam are input, the reliability prediction index is output, the structural safety is classified, and engineering treatment suggestions are automatically generated, and an assessment report is generated.

[0014] Furthermore, the phased establishment of the furnace gas temperature change model includes:

[0015] When the fire duration is less than or equal to the preset standard duration, construct the curve of furnace gas temperature change with fire duration during the heating stage in accordance with ISO 834 standard.

[0016] When the fire duration exceeds the preset standard duration, the exponent of the exponential decay model is constructed by the difference between the fire duration and the peak temperature occurrence time and the preset cooling rate coefficient. The difference between the peak temperature and the ambient temperature is calculated, multiplied by the exponent, and added to the ambient temperature to construct the curve of furnace gas temperature changing with the fire duration during the cooling stage.

[0017] Furthermore, the temperature field is obtained by solving the unsteady heat conduction equation, and the boundary conditions are set using a superposition of convection and radiation, including:

[0018] The first equation is obtained by multiplying the partial derivatives of the material density, specific heat, and temperature field distribution with respect to time. The second equation is obtained by multiplying the thermal conductivity by the divergence of the temperature gradient and the volume heat source term. An equation relationship is established between the first and second equations to form the unsteady-state heat conduction equation.

[0019] The heat flux density is obtained by multiplying the thermal conductivity and the normal temperature gradient. An equation is established by superimposing the heat flux density with the convective heat transfer heat flux density and the radiative heat transfer heat flux density, and the boundary conditions are established.

[0020] The heat conduction equation was solved using the finite element method. Mesh sizes were preset for key parts and normal areas. The complete temperature field was recorded based on a preset time step, and the peak temperature was extracted as the input for subsequent calculations of material degradation performance.

[0021] Furthermore, the calculation of material parameters includes:

[0022] By employing piecewise fitting or interpolation functions, a temperature-yield strength and elastic modulus degradation function is established for steel by calculating the product of a preset initial value and the corresponding degradation coefficient. The degradation coefficient is calculated using a preset piecewise function.

[0023] Similarly, degradation functions for compressive strength and modulus of elasticity are established for concrete, and degradation functions for studs / connectors and interface stiffness are established.

[0024] The original flange thickness and concrete spalling thickness are obtained from the parameters of the steel-concrete composite beam. The spalling correction coefficient is obtained by calculating the ratio of the difference to the original flange thickness. The effect of concrete spalling on the effective area of ​​the cross section is obtained by calculating the product of the spalling correction coefficient and the effective compressive area of ​​the original concrete or the difference between the effective compressive area of ​​the original concrete and the spalling area.

[0025] Furthermore, a bending-slip coupled model is constructed, and coupling analysis is performed, including:

[0026] Considering the interface behavior caused by the difference in thermal expansion between steel and concrete after a fire, a bilinear hysteresis or frictional hysteresis model is used to establish an interface shear force-slip relationship model to calculate the interface shear force.

[0027] The contribution of the steel member is obtained by calculating the stress at each point on the steel section, taking the moment about the moment center of each point, and then integrating over the entire steel area; the contribution of the concrete is obtained by calculating the stress at each point on the concrete section, taking the moment about the moment center of each point, and then integrating over the entire concrete area; the interface shear moment is obtained by calculating the product of the shear force at each interface connection point and the lever arm from each point to the moment center, and then summing them.

[0028] The total bending moment balance of the structure is obtained by summing the contributions of steel members, concrete members, and interface shear moments.

[0029] The theoretical value of the residual bearing capacity is obtained by calculating the product of the preset bending reduction factor, the effective section residual yield strength, and the effective plastic modulus of the section; the actual residual bearing capacity, maximum interface slip, and residual deflection are obtained by running the nonlinear structural solver.

[0030] Furthermore, constructing the limit state function of the composite beam includes: constructing a function in the form of bearing capacity-demand difference or an empirical power function to represent the limit state of the composite beam.

[0031] Furthermore, the calculation of failure probability and reliability index using the first-order reliability method includes:

[0032] The first-order reliability method is used to solve for the design point in the standard normal space, and Monte Carlo importance sampling is performed in the neighborhood of the design point to obtain the failure probability.

[0033] The relationship between reliability index and failure probability is established using the standard normal distribution function, thus obtaining the reliability index.

[0034] Furthermore, constructing a deep learning agent model includes:

[0035] A deep learning proxy model framework is built based on one or a combination of long short-term memory networks, convolutional neural networks and long short-term memory network hybrid models, or deep feedforward neural networks; the number of training samples, the ratio of validation set to test set, hyperparameters and error target are set.

[0036] The training objective function is constructed by calculating the mean square error between the reliability index and the predicted reliability index, and combining it with a regularization term.

[0037] Furthermore, structural safety is classified and engineering treatment recommendations are automatically generated, resulting in an assessment report including:

[0038] The reliability prediction index is judged based on the preset safety classification rules. When the reliability prediction index is greater than or equal to the preset threshold P1, the structure is judged to be safe. When the reliability prediction index is between the preset threshold P1 and the preset threshold P2, it is judged to be used under limited load and monitoring or local reinforcement is recommended. When the reliability prediction index is less than the preset threshold P3, reinforcement or replacement is recommended.

[0039] It automatically generates reports including temperature field, material degradation curve, reliability-failure probability scatter plot, parameter sensitivity ranking, and graded load limit recommendations, and simultaneously generates machine-readable output.

[0040] A reliability calculation system for steel-concrete composite beams after a fire includes:

[0041] The data input module is used to receive the parameters of the steel-concrete composite beam and output a standardized data dictionary;

[0042] The materials modeling module is used to generate a temperature-dependent materials properties database based on the input data and the material parameters.

[0043] The sequential coupling analysis module is used to calculate thermal-structural coupling and output temperature field, residual bearing capacity and interface slip.

[0044] The random modeling and sampling module is used to generate training and validation samples;

[0045] The reliability calculation module is used to calculate and output failure probability and reliability indicators;

[0046] The intelligent agent module is used to train and deploy deep learning agent models and provide a real-time prediction interface;

[0047] The Reporting and Visualization module is used to generate reports and visualizations.

[0048] The beneficial effects of this invention are: first, it realizes a complete closed-loop calculation link from fire input to reliability output; second, it enhances the physical interpretability of the evaluation results through a clear material degradation library, interface model, and reliability process; third, it significantly improves online prediction efficiency by training a proxy model, facilitating rapid engineering decision-making; and fourth, the systematic module and JSON interface design ensures scalability and ease of use in field applications. Attached Figure Description

[0049] Figure 1 A flowchart of a method for calculating the reliability of steel-concrete composite beams after a fire;

[0050] Figure 2 This is a schematic diagram of the finite element model of the present invention;

[0051] Figure 3 This is a schematic diagram of the temperature rise curve according to ISO 834 standard.

[0052] Figure 4 This invention relates to an intelligent reliability assessment system for steel-concrete composite beams after a fire. Detailed Implementation

[0053] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0054] Example 1

[0055] refer to Figures 1 to 3 As shown in the figure, this embodiment introduces a method for calculating the reliability of steel-concrete composite beams after a fire, including:

[0056] The parameters of the steel-concrete composite beam were collected according to the post-fire inspection specifications, including material parameters, geometric parameters, fire parameters and load information; the temperature, length and time parameters were standardized to the same unit, and outliers were removed by median filtering combined with the 3σ criterion; and standardized input files were exported by time synchronization and interpolation processing.

[0057] A staged model of furnace gas temperature change was established. The temperature field distribution was obtained by solving the unsteady heat conduction equation, and the boundary conditions were set by superimposing convection and radiation.

[0058] Based on the temperature distribution of the temperature field, a material temperature-performance degradation model after a fire is constructed to calculate material parameters and form a temperature-dependent material performance database; a bending-slip coupling model is constructed and coupled analysis is performed to obtain the residual bearing capacity and slip limit.

[0059] Based on the material parameters in the material property database, a set of random variables is constructed; using the set of random variables as independent variables and the limit state value as dependent variable, the limit state function of the composite beam is constructed.

[0060] Based on the limit state function of the composite beam, the first-order reliability method is used to solve the design point in the standard normal space. The importance sampling density is constructed in the neighborhood, and Monte Carlo importance sampling is implemented to calculate the failure probability and reliability index. High-fidelity samples are generated and added to the training sample pool, and the process is iterated until the preset convergence requirement is reached.

[0061] A deep learning proxy model is constructed, with a set of random variables as input and reliability index or failure probability as output, and the model is fully trained. After training, the parameters of steel-concrete composite beams obtained online or detected in near real-time are input into the deep learning proxy model, and the reliability prediction index of composite beams after fire is output. The structural safety is classified and engineering treatment suggestions are automatically generated, and an assessment report is generated.

[0062] Furthermore, the phased establishment of the furnace gas temperature change model includes:

[0063] When the fire duration is less than or equal to 60 minutes, the furnace gas temperature change curve during the heating stage is constructed according to ISO 834 standard, and the expression is as follows:

[0064]

[0065] in, t represents the furnace gas temperature, and t represents the fire duration.

[0066] When the fire duration exceeds 60 minutes, an exponential decay model is established to construct the furnace gas temperature change curve during the cooling stage, as shown in the following expression:

[0067]

[0068] in, Here, t represents the furnace gas temperature, and t represents the fire duration. For ambient temperature, Peak temperature The time when the peak temperature occurs. This is the cooling rate coefficient, which is fitted from field data or given a recommended range in the appendix of the instruction manual.

[0069] Furthermore, the temperature field distribution is obtained by solving the unsteady heat conduction equation, and the boundary conditions are set using a superposition of convection and radiation, including:

[0070] The first equation is obtained by multiplying the partial derivatives of the material density, specific heat, and temperature field distribution with respect to time. The second equation is obtained by multiplying the thermal conductivity by the divergence of the temperature gradient and adding the volumetric heat source term. By establishing an equation relationship between the first and second equations, the unsteady-state heat conduction equation is constructed, as shown in the following expression:

[0071]

[0072] Where T(x, t) represents the temperature distribution of the temperature field. For material density, For specific heat, Thermal conductivity, For the heat source term;

[0073] The convective heat flux density is obtained by multiplying the convective heat transfer coefficient by the difference between the surface temperature and the ambient temperature; the radiative heat flux density is obtained by multiplying the surface emissivity and the Stefan-Boltzmann constant by the difference between the fourth power of the surface temperature and the fourth power of the ambient temperature, respectively; the heat flux density is obtained by multiplying the thermal conductivity and the normal temperature gradient. An equation is established by superimposing the radiative heat flux density on the convective heat flux density, and the boundary conditions are established as follows:

[0074]

[0075] in, Let be the derivative of temperature with respect to the normal direction of the material surface. The surface temperature of the material. For ambient temperature, The convective heat transfer coefficient is... This represents the surface emissivity, typically ranging from 0.65 to 0.95. It is the Stefan-Boltzmann constant;

[0076] The heat conduction equation was solved using the finite element method. For critical areas, such as interfaces, welds, and stress concentration zones, a mesh size of 5-15 mm was used, while for general areas, a mesh size of 10-30 mm was used. A time step of 0.1–5 s was set to record the complete temperature field, and the peak temperature was extracted as the input for subsequent calculations of material degradation performance.

[0077] Furthermore, the calculation of material parameters includes:

[0078] For steel, a piecewise fitting or interpolation function is used to establish the temperature-yield strength versus elastic modulus degradation function, as shown in the following expression:

[0079]

[0080] in, This represents the residual yield strength of steel at temperature T. This represents the initial yield strength of the steel. This represents the residual elastic modulus of steel at temperature T. The initial elastic modulus of the steel. Peak temperature at key locations;

[0081] Similarly, the degradation function of compressive strength and elastic modulus of concrete is expressed as follows:

[0082]

[0083] in, This represents the residual compressive strength of concrete at temperature T. This represents the initial compressive strength of the concrete. This represents the residual elastic modulus of concrete at temperature T. This represents the initial elastic modulus of the concrete.

[0084] Similarly, a degradation function is established for the stud / connector and interface stiffness, expressed as follows:

[0085]

[0086] in, The residual mechanical properties of the stud / connector at temperature T. For the initial mechanical properties of the studs / connectors, Interface stiffness at temperature T This represents the initial interface stiffness;

[0087] The original flange thickness and concrete spalling thickness are obtained from the parameters of the steel-concrete composite beam. The spalling correction coefficient is obtained by calculating the ratio of the difference between the original flange thickness and the concrete spalling thickness to the original flange thickness. The effect of concrete spalling on the effective area of ​​the cross section is obtained by calculating the product of the spalling correction coefficient and the effective compressive area of ​​the original concrete or the difference between the effective compressive area of ​​the original concrete and the spalling area.

[0088] Furthermore, a bending-slip coupled model is constructed, and coupling analysis is performed, including:

[0089] Considering the interfacial behavior caused by the difference in thermal expansion between steel and concrete after a fire, a bilinear hysteresis or frictional hysteresis model is used to establish an interfacial shear force-slip relationship model to calculate the interfacial shear force, as expressed below:

[0090]

[0091] Where V is the interfacial shear force and s is the slip. For initial stiffness, This is the slip yield value. For corresponding shear force, To soften stiffness, For the limit slip, Residual shear force;

[0092] The contribution of the steel member is obtained by calculating the stress at each point on the steel section, taking the moment about the moment center of each point, and then integrating over the entire steel area; the contribution of the concrete is obtained by calculating the stress at each point on the concrete section, taking the moment about the moment center of each point, and then integrating over the entire concrete area; the interface shear moment is obtained by calculating the product of the shear force at each interface connection point and the lever arm from each point to the moment center, and then summing them.

[0093] The total bending moment balance of the structure is obtained by summing the contributions of steel members, concrete members, and interface shear moments.

[0094] The theoretical value of the residual bearing capacity is obtained by calculating the product of the preset bending reduction factor, the effective section residual yield strength, and the effective plastic modulus of the section; the actual residual bearing capacity, maximum interface slip, and residual deflection are obtained by running the nonlinear structural solver.

[0095] Furthermore, constructing the limit state function of the composite beam includes:

[0096] Construct a function in the form of load-bearing capacity-demand difference or an empirical power function to calculate the ultimate state of the composite beam;

[0097] When the limit state function value of the composite beam is less than or equal to 0, it is determined to be a failure. Based on the limit state function of the composite beam, the total failure probability of the system is calculated as follows:

[0098]

[0099] in, Pr[·] represents the failure probability, and Pr[·] is the probability operator. Let the limit state function of the composite beam be... Let X be the joint probability density function of the basic random variable vector X.

[0100] Furthermore, implementing Monte Carlo importance sampling to calculate failure probability and reliability metrics includes:

[0101] The Monte Carlo importance sampling estimation formula is established as follows:

[0102]

[0103] in, It is an estimate of the failure probability. For indicator functions, For the original joint density, For importance sampling density (e.g., multivariate normal recursion transformation centered on the design point), N is the sample size, which should be adaptively selected according to the target confidence requirement. The instruction manual recommends 10³ ≤ N ≤ 10⁵.

[0104] The relationship between the reliability index and the failure probability is established as follows:

[0105]

[0106] in, It is the standard normal distribution function. As a reliability indicator, This represents the probability of failure.

[0107] Furthermore, constructing a deep learning agent model includes:

[0108] A deep learning proxy model framework is built using one or a combination of LSTM, CNN+LSTM, or DNN to adapt to temporal or static features; the number of training samples is set to 300-800, and cross-validation is used, with the validation set and test set each accounting for 10%-20% of the samples; hyperparameters are set, with the learning rate set to 1×10⁻⁶. -3 The optimizer is set to Adam, the batch size is set to 16–64, the early stopping patience is set to 30–100, and the number of hidden layer units is set to 32–512.

[0109] By calculating the mean squared error between the reliability index and the predicted reliability index, and combining it with a regularization term, a training objective function is constructed, expressed as follows:

[0110]

[0111] in, For loss function, For parameters of deep learning proxy models, For a set of random variables, For sample size, To predict reliability indicators, Let i be the reliability index corresponding to the i-th sample. The regularization coefficient is used to control model complexity. It is the squared L2 norm of the parameters of the deep learning agent model;

[0112] The embodiments in the specification demonstrate that, under the premise of meeting the error target (e.g., average relative error < 5%), the surrogate prediction time can be 1 / 10 to 1 / 100 of FORM+IS-MCS, thus making it suitable for online / near real-time evaluation.

[0113] Furthermore, structural safety is classified and engineering treatment recommendations are automatically generated, resulting in an assessment report including:

[0114] Based on preset safety classification rules, the reliability prediction index is judged. When the reliability prediction index is ≥3.0, the structure is judged to be safe; when 2.0≤reliability prediction index<3.0, it is judged to be used under limited load and monitoring or local reinforcement is recommended; when the reliability prediction index<2.0, reinforcement or replacement is recommended.

[0115] It automatically generates PDF / A reports that include temperature field, material degradation curve, reliability-failure probability scatter plot, parameter sensitivity ranking (e.g., SHAP value), and graded load limit recommendations, and simultaneously generates machine-readable output in JSON format for easy integration with subsequent maintenance systems or databases.

[0116] Example 2

[0117] refer to Figure 4 As shown in the figure, this embodiment introduces a reliability calculation system for steel-concrete composite beams after a fire, including:

[0118] The data input module is used to receive geometric, material, fire, and monitoring data and output a standardized data dictionary;

[0119] The materials modeling module is used to generate a temperature-dependent materials properties database based on input data and degraded materials parameters;

[0120] The sequential coupling analysis module is used to call the finite element solver to calculate the thermo-mechanical coupling and output the temperature field, residual bearing capacity and interface slip.

[0121] The random modeling and sampling module is used to generate training and validation samples using Latin hypercube or Monte Carlo sampling methods; the number of training samples is preferably 300 to 800, and the validation and test samples each account for 10%–20%;

[0122] The reliability calculation module is used to perform first-order reliability calculation and Monte Carlo importance sampling on each group of samples, and calculate and output failure probability and reliability index.

[0123] The intelligent agent module is used to train and deploy deep learning agent models and provide a real-time prediction interface;

[0124] The Reporting and Visualization module is used to generate PDF / JSON reports and visualization charts.

[0125] The modules are interconnected via a JSON-formatted data stream interface. The main control software is implemented on the Python platform, the graphical interface is implemented using PyQt, the proxy model is built using TensorFlow or PyTorch, and numerical computation is implemented using NumPy and SciPy.

[0126] In summary, by integrating physical mechanism models and deep learning proxy models, this invention constructs a complete thermo-mechanical sequential coupling analysis chain from heat conduction and material degradation to interface slip, ensuring the accuracy of the physical process and transforming complex reliability calculations into intelligent predictions at the second level. It not only quantifies the impact of multiple uncertainties such as material degradation and interface damage, but also automatically generates an assessment report containing safety classification and disposal recommendations, significantly improving the scientific rigor, automation level, and engineering practical value of post-fire structural safety assessment.

[0127] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for calculating the reliability of steel-concrete composite beams after a fire, characterized in that, include: Collect parameters of the steel-concrete composite beam, standardize the units, and perform preprocessing; A staged model of furnace gas temperature change was established. The temperature field was obtained by solving the unsteady heat conduction equation, and the boundary conditions were set by superimposing convection and radiation. A material temperature-performance degradation model after a fire is constructed to calculate material parameters; a bending-slip coupling model is constructed and coupled analysis is performed to obtain the residual bearing capacity and slip limit. Construct a set of random variables as independent variables and the limit state value as dependent variable to construct the limit state function of the composite beam; The failure probability and reliability index are calculated using the first-order reliability method. High-fidelity samples are generated and added to the training sample pool. The process is iterated until the preset convergence requirement is met. A deep learning proxy model is constructed. After the model is fully trained, the parameters of the steel-concrete composite beam are input, the reliability prediction index is output, the structural safety is classified, and engineering treatment suggestions are automatically generated, and an assessment report is generated.

2. The method for calculating the reliability of a steel-concrete composite beam after a fire, as described in claim 1, is characterized in that... The phased establishment of the furnace gas temperature change model includes: When the fire duration is less than or equal to the preset standard duration, construct the curve of furnace gas temperature change with fire duration during the heating stage in accordance with ISO 834 standard. When the fire duration exceeds the preset standard duration, the exponent of the exponential decay model is constructed by the difference between the fire duration and the peak temperature occurrence time and the preset cooling rate coefficient. The difference between the peak temperature and the ambient temperature is calculated, multiplied by the exponent, and added to the ambient temperature to construct the curve of furnace gas temperature changing with the fire duration during the cooling stage.

3. The method for calculating the reliability of a steel-concrete composite beam after a fire, as described in claim 1, is characterized in that... The temperature field is obtained by solving the unsteady heat conduction equation, and the boundary conditions are set by superimposing convection and radiation, including: The first equation is obtained by multiplying the partial derivatives of the material density, specific heat, and temperature field distribution with respect to time. The second equation is obtained by multiplying the thermal conductivity by the divergence of the temperature gradient and the volume heat source term. An equation relationship is established between the first and second equations to form the unsteady-state heat conduction equation. The heat flux density is obtained by multiplying the thermal conductivity and the normal temperature gradient. An equation is established by superimposing the heat flux density with the convective heat transfer heat flux density and the radiative heat transfer heat flux density, and the boundary conditions are established. The heat conduction equation was solved using the finite element method. Mesh sizes were preset for key parts and normal areas. The complete temperature field was recorded based on a preset time step, and the peak temperature was extracted as the input for subsequent calculations of material degradation performance.

4. The method for calculating the reliability of a steel-concrete composite beam after a fire, as described in claim 1, is characterized in that... The calculation of material parameters includes: By employing piecewise fitting or interpolation functions, a temperature-yield strength and elastic modulus degradation function is established for steel by calculating the product of a preset initial value and the corresponding degradation coefficient. The degradation coefficient is calculated using a preset piecewise function. Similarly, degradation functions for compressive strength and modulus of elasticity are established for concrete, and degradation functions for studs / connectors and interface stiffness are established. The original flange thickness and concrete spalling thickness are obtained from the parameters of the steel-concrete composite beam. The spalling correction coefficient is obtained by calculating the ratio of the difference to the original flange thickness. The effect of concrete spalling on the effective area of ​​the cross section is obtained by calculating the product of the spalling correction coefficient and the effective compressive area of ​​the original concrete or the difference between the effective compressive area of ​​the original concrete and the spalling area.

5. The method for calculating the reliability of a steel-concrete composite beam after a fire, as described in claim 1, is characterized in that... Constructing a bending-slip coupled model and performing coupling analysis includes: Considering the interface behavior caused by the difference in thermal expansion between steel and concrete after a fire, a bilinear hysteresis or frictional hysteresis model is used to establish an interface shear force-slip relationship model to calculate the interface shear force. The contribution of the steel member is obtained by calculating the stress at each point on the steel section, taking the moment about the moment center of each point, and then integrating over the entire steel area; the contribution of the concrete is obtained by calculating the stress at each point on the concrete section, taking the moment about the moment center of each point, and then integrating over the entire concrete area; the interface shear moment is obtained by calculating the product of the shear force at each interface connection point and the lever arm from each point to the moment center, and then summing them. The total bending moment balance of the structure is obtained by summing the contributions of steel members, concrete members, and interface shear moments. The theoretical value of the residual bearing capacity is obtained by calculating the product of the preset bending reduction factor, the effective section residual yield strength, and the effective plastic modulus of the section; the actual residual bearing capacity, maximum interface slip, and residual deflection are obtained by running the nonlinear structural solver.

6. The method for calculating the reliability of a steel-concrete composite beam after a fire, as described in claim 1, is characterized in that... Constructing the limit state function of a composite beam includes: constructing a function in the form of bearing capacity-demand difference or an empirical power function to represent the limit state of the composite beam.

7. The method for calculating the reliability of a steel-concrete composite beam after a fire, as described in claim 1, is characterized in that... The calculation of failure probability and reliability indices using the first-order reliability method includes: The first-order reliability method is used to solve for the design point in the standard normal space, and Monte Carlo importance sampling is performed in the neighborhood of the design point to obtain the failure probability. The relationship between reliability index and failure probability is established using the standard normal distribution function, thus obtaining the reliability index.

8. The method for calculating the reliability of a steel-concrete composite beam after a fire, as described in claim 1, is characterized in that... Building a deep learning agent model includes: A deep learning proxy model framework is built based on one or a combination of long short-term memory networks, convolutional neural networks and long short-term memory network hybrid models, or deep feedforward neural networks; the number of training samples, the ratio of validation set to test set, hyperparameters and error target are set. The training objective function is constructed by calculating the mean square error between the reliability index and the predicted reliability index, and combining it with a regularization term.

9. The method for calculating the reliability of a steel-concrete composite beam after a fire, as described in claim 1, is characterized in that... The system performs structural safety classification and automatically generates engineering treatment recommendations, producing an assessment report that includes: The reliability prediction index is judged based on the preset safety classification rules. When the reliability prediction index is greater than or equal to the preset threshold P1, the structure is judged to be safe. When the reliability prediction index is between the preset threshold P1 and the preset threshold P2, it is judged to be used under limited load and monitoring or local reinforcement is recommended. When the reliability prediction index is less than the preset threshold P3, reinforcement or replacement is recommended. It automatically generates reports including temperature field, material degradation curve, reliability-failure probability scatter plot, parameter sensitivity ranking, and graded load limit recommendations, and simultaneously generates machine-readable output.

10. A reliability calculation system for steel-concrete composite beams after a fire, characterized in that, include: The data input module is used to receive the parameters of the steel-concrete composite beam and output a standardized data dictionary; The materials modeling module is used to generate a temperature-dependent materials properties database based on the input data and the material parameters. The sequential coupling analysis module is used to calculate thermal-structural coupling and output temperature field, residual bearing capacity and interface slip. The random modeling and sampling module is used to generate training and validation samples; The reliability calculation module is used to calculate and output failure probability and reliability indicators; The intelligent agent module is used to train and deploy deep learning agent models and provide a real-time prediction interface; The Reporting and Visualization module is used to generate reports and visualizations.

Images