Bridge vehicle fire under the main cable internal temperature inversion method

Abstract

The application discloses a kind of bridge vehicle fire under main cable internal temperature inversion method, belong to bridge fire damage assessment field, specifically include the following steps: output data and working condition parameter are constructed into dataset, data set is preprocessed, constructs the main cable heat transfer inversion model based on LSTM conditional generative adversarial network, the generator of LSTM conditional generative adversarial network model is trained, obtains the main cable heat transfer inversion model after training;Finally take the working condition parameter of bridge vehicle fire scene, it is input to the main cable heat transfer inversion model after training, generates bridge main cable internal time-temperature variation curve.The application not only can complete the inversion of main cable internal time-temperature variation curve, solve the inversion problem that main cable internal temperature is not directly measurable, but also realize the order of magnitude breakthrough of evaluation efficiency, applicable to bridge fire post-disaster assessment scene, provides key time window for bridge safety condition rapid judgment and repair decision.

Description

Technical Field

[0001] This invention relates to fire assessment technology, specifically to a method for inverting the internal temperature of the main cable under a bridge vehicle fire. Background Technology

[0002] When a vehicle traveling on a bridge catches fire, the high-temperature flames and hot smoke produced not only directly damage the bridge deck structure but also cause long-term thermal damage to critical load-bearing components such as the main cable. As the main load-bearing element of the bridge, the internal temperature changes of the main cable directly determine the degree of degradation of the material's mechanical properties, thus affecting post-disaster load-bearing capacity assessment and repair decisions.

[0003] In post-fire damage assessment of bridges, accurately reconstructing the temperature and time-related changes within the main cable during the fire is crucial. Considering the enclosed structure of the main cable and the inability to directly measure its internal temperature, traditional assessment methods primarily rely on two technical approaches: one is to obtain surface temperature data based on thermocouples or infrared thermometry and extrapolate the internal temperature using empirical formulas; the other is to establish a sophisticated physical simulation model. However, these methods have significant limitations in post-disaster assessment scenarios. 1. Insufficient real-time capability hinders post-disaster assessment: Post-disaster assessment often requires determining the structural safety status as soon as possible. Traditional finite element simulation takes several hours to complete a single calculation, while the main cable temperature inversion needs to consider the entire process of "rapid heating-slow rise-gradual cooling," which is difficult to meet the timeliness requirements of post-disaster assessment. 2. Limited accuracy of internal damage assessment: Surface temperature measurement cannot capture the thermal inertia effect and heat transfer hysteresis characteristics inside the main cable. In physical simulation, the equivalent thermal conductivity changes complexly with temperature, and the variable physical parameters of fireproof materials such as aerogel felt further increase the uncertainty of the model, leading to bias in damage assessment. 3. Insufficient level of intelligence: Existing post-disaster assessments rely heavily on expert experience to interpret simulation results and lack automated internal temperature reconstruction capabilities. Especially when it is necessary to quickly analyze the damage impact of multiple fire scenarios on the main cable, the workload of traditional methods increases exponentially. 4. In bridge vehicle fires, the location of vehicle fires is random. Measuring the internal temperature of the main cable requires the deployment of temperature measurement equipment. First, it is difficult to deploy measurement equipment inside the main cable, and second, the cost of deploying measurement equipment throughout the entire cable is high. Therefore, developing an intelligent inversion method for the internal temperature of the main cable for post-disaster damage assessment has become a key technical bottleneck in improving the ability to assess bridge fire damage. Summary of the Invention

[0004] The purpose of this invention is to provide a method for inverting the internal temperature of the main cable under bridge vehicle fire conditions. This method can not only invert the time-temperature change curve inside the main cable, solving the problem that the internal temperature of the main cable cannot be directly measured, but also achieve a breakthrough in assessment efficiency by orders of magnitude. It is applicable to post-fire assessment scenarios of bridges and provides a critical time window for rapid judgment of bridge safety status and repair decisions.

[0005] To achieve the above objectives, the present invention provides a method for inverting the internal temperature of the main cable under a bridge vehicle fire, which specifically includes the following steps: S1, obtain the working condition parameters, which include fire source intensity, fire source distance and bridge deck wind speed. Input the working condition parameters into the bridge vehicle fire simulation model to calculate and output the ambient temperature field on the cable surface. S2, input the cable surface temperature field into the main cable heat transfer simulation model, and output the time-temperature data inside the main cable; S3, construct a dataset by combining the time-temperature data inside the main cable obtained from the experiment, the time-temperature data inside the main cable output from step S2, and the corresponding operating parameters, preprocess the dataset, and divide it into a training set, a validation set, and a test set; S4. Construct a main cable heat transfer inversion model based on LSTM conditional generative adversarial network. Use a weighted combination loss function that includes adversarial loss and physical constraints to train the generator of the LSTM conditional generative adversarial network model to obtain the trained main cable heat transfer inversion model. S5. Obtain the operating parameters of the bridge vehicle fire scenario, input them into the trained main cable heat transfer inversion model, calculate and output the predicted temperature value inside the main cable within a preset time range, and generate the time-temperature change curve inside the bridge main cable based on the predicted temperature value.

[0006] In some examples of the present invention, in step S4, the main cable heat transfer inversion model has a generator and a discriminator; The generator uses an LSTM autoregressive structure and includes the following steps: The input layer receives a conditional scalar c defined by operating parameters, and generates the initial hidden state and initial cell state of the LSTM through a fully connected layer. At each time step, the predicted temperature from the previous time step is concatenated with the conditional scalar c as input. The hidden state and cell state are updated through the LSTM gating mechanism. The output layer uses the Sigmoid activation function to generate normalized temperature values, thus obtaining the complete main cable temperature prediction sequence. The discriminator uses a conditional LSTM structure, specifically including the following steps: At each time step, the temperature value is concatenated with the conditional scalar c to form the input sequence. Temporal features are extracted by an LSTM encoder, and the final hidden state outputs the probability of authenticity through a fully connected layer and a Sigmoid activation function.

[0007] In some examples of the present invention, in the generator, the conditional scalar c is concatenated with the noise vector z to form a joint vector [z;c], which is then input into the input layer.

[0008] In some examples of this invention, the generator introduces a reconstruction term of the main cable temperature sequence on top of the adversarial loss, and the comprehensive loss can be written as: The discriminator loss uses the standard binary cross-entropy form: Where E represents the mathematical expectation operator; D(·) represents the probability function of the discriminator's output; G(·) represents the temperature sequence function of the main cable's internal temperature generated by the generator; y represents the physical constraint weight hyperparameter; y represents the actual internal temperature sequence of the main cable sampled from the real data distribution; Pdata represents the data distribution of the actual main cable temperature sequence.

[0009] In some examples of the present invention, the generator progressively generates a main cable temperature prediction sequence in an autoregressive manner. Specifically, it includes the following steps: The joint vector [z;c] generates the initial hidden state and the initial cell state through a fully connected layer: in, , For the LSTM hidden unit dimension, the split() function splits the fully connected layer output into hidden state h0 and cell state c0, W. init This is the initialization of the weight matrix, b init It is a bias term; At each time step The temperature value generated at the previous moment With conditional scalar c The current inputs that make up the gating mechanism: The current input x t Feeding into LSTM gating mechanism update ; Then, the normalized temperature prediction value for this moment is given by outputting a fully connected layer and sigmoid activation: Among them, among them, , It is the Sigmoid activation function. h is the transpose of the output weight matrix.t Hide the current state. For bias; During training, the global minimum and maximum values ​​are then used to... After inverse normalization to the physical temperature range, the main cable temperature prediction sequence is obtained through T autoregressive cycles: .

[0010] In some examples of this invention, the formula for calculating the "authenticity" probability of the output through two fully connected layers and Sigmoid activation in the discriminator is as follows: in, It is the Sigmoid activation function. Use the LeakyReLU activation function; It is considered as a high-dimensional feature representation of the entire sequence under a given working condition, and W1 is the weight matrix of the first fully connected layer. b1 and b2 are the transpose of the weight matrix of the second fully connected layer, and b1 and b2 are the biases of each fully connected layer.

[0011] In some examples of the present invention, in step S2, the main cable structure is regarded as a "steel wire-void" porous medium in the heat transfer simulation model of the main cable. By fitting the equivalent thermal conductivity and correcting the equivalent specific heat capacity, a simplified equivalent model of the steel wire-void main cable structure is established. Finally, the ambient temperature field of the cable surface in step S1 is input into the simplified equivalent model, and the time-temperature data inside the main cable is output.

[0012] In some examples of this invention, the simplified equivalent model is derived using a rectangular finite element model. The rectangular finite element model consists of the main cable wires and the air gaps between the wires. The internal wire dimensions and porosity are consistent with the experimental model. There is surface-to-surface thermal radiation between the wire gaps. The radiating surface is defined as a diffuse reflective surface and its emissivity is set. The radiation direction is set to be controlled by opacity. The opacities of the wires and air are set to opaque and transparent, respectively. The air is set to be non-flowing, and the thermal convection of the air is ignored. The equivalent thermal conductivity of the simplified equivalent model is calculated using the following formula: in, The average heat flux density is denoted as H; H is the thickness of the main cable protective layer; T1 and T2 are the upper and lower boundary temperatures, respectively. Numerical fitting was performed on the equivalent thermal conductivity, and a fourth-order polynomial was used to obtain the formula for calculating the equivalent thermal conductivity as a function of temperature: in, , , , For temperature coefficient, It is a constant. Ambient temperature; The equivalent specific heat capacity is corrected to 60%-80% of the original specific heat capacity based on the porosity of the model.

[0013] In some examples of the present invention, the dataset is preprocessed in step S3. Specifically, the time-temperature data calculated and output by the main cable heat transfer simulation model is uniformly mapped to 91 standard time points with 1-minute intervals within 0-90 minutes using linear interpolation, and the temperature data is normalized.

[0014] In some examples of the present invention, in step S4, the training rounds are 300-500 rounds, and the minimum mean absolute error of the validation set is used as the early stopping criterion. In step S5, the evaluation index of the main cable heat transfer inversion model on the test set is: mean square error (MSE) < 35, mean absolute error (MAE) < 5, and coefficient of determination (R²). 2 >0.9, and the calculation time for a single inversion is controlled within 3 seconds.

[0015] Compared with existing technologies, the present invention provides a method for inverting the internal temperature of the main cable under bridge vehicle fire conditions. This method constructs a main cable heat transfer inversion model based on LSTM conditional generative adversarial network, which can not only invert the time-temperature change curve inside the main cable and solve the problem that the internal temperature of the main cable cannot be directly measured, but also achieves a breakthrough in evaluation efficiency by orders of magnitude. It is applicable to post-fire assessment scenarios of bridges and provides a critical time window for rapid judgment of bridge safety status and repair decisions. In addition, only the operating parameters of the vehicle fire need to be input to directly generate the internal temperature field of the main cable, solving the problem that it is not feasible to deploy temperature measurement equipment on the entire main cable, and avoiding the need for additional temperature measurement and the difficulty in deploying measurement equipment. In this model, the generator adopts an autoregressive generation method where the output of the previous time step is used as the input of the next time step. This allows the generator to naturally satisfy the time smoothness and continuity constraints numerically, and can better reproduce the gradual temperature rise and slow temperature drop characteristics of the main cable in the finite element simulation of solid heat transfer. The discriminator learns the physical correspondence between the "operating parameters and the internal temperature rise curve of the main cable" by explicitly introducing conditional scalars in the input layer. By using a weighted combination loss function that includes adversarial loss and physical constraints, it ensures a high degree of consistency between the generated sequence and the real data, which greatly improves the evaluation efficiency and achieves high evaluation accuracy. The gating mechanism can automatically "remember" the key stages related to the temperature rise of the main cable and "forget" the disturbances that are less related to the final result during a long period of temperature evolution. This enables time-series modeling of the entire process of "rapid heating, slow climbing, and gradual cooling" of the main cable temperature, especially the modeling of thermal inertia effect and heat transfer hysteresis characteristics, providing a more reliable data foundation for damage assessment. A simplified equivalent model is used to replace the high-precision numerical model. The main cable structure is regarded as a porous medium of "steel wire-void". The heat transfer process is equivalently replaced by the equivalent heat transfer coefficient and the equivalent specific heat capacity, which can improve the calculation efficiency while ensuring the calculation accuracy. Attached Figure Description

[0016] Figure 1 This is an overall flowchart of the present invention; Figure 2 This is a flowchart illustrating the generator in the main cable heat transfer inversion model of this invention. Figure 3 This is a flowchart illustrating the discriminator in the main cable heat transfer inversion model of this invention. Figure 4 This is a comparison chart of predicted temperature and simulated temperature in Example 2 of this invention; (a) Test sample under the condition that the maximum ambient temperature of the main cable is 700℃; (b) Test sample under the condition that the maximum ambient temperature of the main cable is 900℃; (c) Test sample under the condition that the maximum ambient temperature of the main cable is 1000℃; (d) Test sample under the condition that the maximum ambient temperature of the main cable is 1100℃; Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The same reference numerals in the drawings represent the same components. It should be noted that the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms “first,” “second,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, “an” or “a” and similar terms do not necessarily indicate a quantity limitation. Terms such as “comprising” or “including” mean that the element or object preceding the word encompasses the element or object listed following the word and its equivalents, without excluding other elements or objects. Terms such as “connected” or “linked” are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as “upper,” “lower,” “left,” and “right” are used only to indicate relative positional relationships; these relative positional relationships may change accordingly when the absolute position of the described object changes.

[0019] like Figure 1 As shown, the present invention provides a method for inverting the internal temperature of the main cable under a bridge vehicle fire, which specifically includes the following steps: S1, obtain the working condition parameters, which include fire source intensity, fire source distance, and bridge deck wind speed. Input the working condition parameters into the bridge vehicle fire simulation model and output the ambient temperature field of the cable surface. S2, input the cable surface temperature field into the main cable heat transfer simulation model, and output the time-temperature data inside the main cable; S3, construct a dataset by combining the time-temperature data inside the main cable obtained from the experiment, the time-temperature data inside the main cable output from step S2, and the corresponding operating parameters, preprocess the dataset, and divide it into a training set, a validation set, and a test set; S4. Construct a main cable heat transfer inversion model based on LSTM conditional generative adversarial network. This model has a generator and a discriminator. A weighted combination loss function containing adversarial loss and physical constraints is used to train the generator of the LSTM conditional generative adversarial network model to obtain the trained main cable heat transfer inversion model. S5. Obtain the operating parameters of the bridge vehicle fire scenario, input them into the trained main cable heat transfer inversion model, output the predicted temperature value inside the main cable within a preset time range, and generate the time-temperature change curve inside the bridge main cable based on the predicted temperature value.

[0020] Specifically, in step S1, a large number of working condition parameters can be obtained through experimental simulation, historical fire data, etc. These working condition parameters include fire source intensity, bridge deck wind speed, and fire source distance. By inputting different working condition parameters into the bridge vehicle fire simulation model, the ambient temperature field of the cable surface is calculated and output, that is, the time-temperature data of the cable surface is output. In step S2, the heat transfer simulation model of the main cable can be a traditional high-precision numerical model, which outputs time-temperature data inside the main cable after a large amount of calculation. Considering the low calculation efficiency of this model, a simplified equivalent model can be used to replace the original high-precision numerical model in this step, which can improve the calculation efficiency while ensuring the calculation accuracy. The heat transfer simulation model of the main cable is a simplified equivalent model. The main cable structure is regarded as a porous medium of "steel wire-void". By fitting the equivalent thermal conductivity and correcting the equivalent specific heat capacity, a simplified equivalent model of the steel wire-void main cable structure is established. Finally, the ambient temperature field of the cable surface in step S1 is input into the simplified equivalent model, and the time-temperature data inside the main cable is output. The simplified equivalent model is derived using a rectangular finite element model. The rectangular finite element model consists of the main cable steel wires and the air gaps between the steel wires. The internal steel wire dimensions and porosity are consistent with the experimental model. There is surface-to-surface thermal radiation between the steel wire gaps. The radiating surface is defined as a diffuse reflective surface and its emissivity is set. The radiation direction is set to be controlled by opacity. The opacities of the steel wires and air are set to opaque and transparent, respectively. The air is set to be non-flowing, and the thermal convection of the air is ignored. The upper and lower boundaries of the model are set to fixed temperatures T1 and T2, respectively, and the left and right boundaries are adiabatic. The average heat flux density q of the model can be obtained through finite element analysis, and the equivalent thermal conductivity can be obtained according to Fourier's law. ; The equivalent thermal conductivity is obtained through back-calculation using rectangular finite element analysis, and the calculation formula is as follows: in, The average heat flux density is denoted as H; H is the thickness of the main cable protective layer; T1 and T2 are the upper and lower boundary temperatures, respectively. Numerical fitting was performed on the equivalent thermal conductivity, and a fourth-order polynomial was used to obtain the formula for calculating the equivalent thermal conductivity as a function of temperature: in, , , , For temperature coefficient, It is a constant. The ambient temperature is the temperature field on the cable surface. In step S3, the time-temperature data (experimental data) inside the main cable obtained from the experiment can be acquired by establishing a hybrid test platform. The working parameters are controlled by adjusting the wind speed, angle, burner valve opening, and position of the combustion source during the experiment. Using integrated pressure sensors, infrared thermal imagers, distributed thermocouple arrays, and ultrasonic anemometers, the axial displacement, strain / stress data of the main cable model and the temperature / wind speed data of the fire scene are collected simultaneously. The stress, deformation, and temperature field distribution of the entire bridge are dynamically visualized through the Unity3D engine, and the load-displacement curve and the time-temperature curve (data) inside the main cable are generated simultaneously. The corresponding technology can be obtained from the Chinese invention patent "Real-time Hybrid Test Platform and Implementation Method of Vehicle-Fire-Wind Force for Substructure of Large Span Bridge" authorized by our research group with authorization announcement number CN120741035B. The time-temperature data inside the main cable output during the experiment is combined with the corresponding operating parameters, and the time-temperature data inside the main cable output in step S2 is combined with the corresponding operating parameters to form a dataset. The dataset is then preprocessed and divided into training set, validation set and test set in a ratio of 8:1:1. In step S4, such as Figure 2 , Figure 3 As shown, the main cable heat transfer inversion model has a generator and a discriminator, and adopts an alternating iterative update strategy of discriminator first and generator following, with 300-500 training rounds, and the minimum mean absolute error of the validation set is used as the early stopping criterion. The generator employs an LSTM autoregressive structure. The input layer receives a concatenated vector [z;c], which is formed by concatenating the noise vector z with a conditional scalar c. The conditional scalar c contains operating parameters including fire source intensity, fire source distance, and bridge surface wind speed. The initial hidden state and initial cell state of the LSTM are generated through a fully connected layer. At each time step, the predicted temperature from the previous time step is concatenated with the conditional scalar c as input. The hidden state and cell state are updated through an LSTM gating mechanism. The output layer uses a sigmoid activation function to generate normalized temperature values, thus obtaining the complete main cable temperature prediction sequence. ; The discriminator also uses a conditional LSTM structure. At each time step, the predicted temperature is concatenated with the conditional scalar c to form the input sequence. Temporal features are extracted by the LSTM encoder, and the final hidden state outputs the probability of truth through a fully connected layer and a sigmoid activation function. In step S5, when a fire occurs on a vehicle on the bridge, the working condition parameters of the fire are obtained. These working condition parameters can be confirmed by a large language model through multiple rounds of human-computer interaction. The large language model can be a dedicated model obtained by fine-tuning the existing open-source pre-trained model with text and code instructions in the field of bridge engineering, so as to be applicable to the field of bridge fire damage assessment. For key condition parameters that are not clearly defined, the questioning logic is initiated. For condition parameters that are ambiguous or exceed the reasonable range, the confirmation and correction logic is initiated, and finally, a complete and error-free set of standardized working condition parameters is output. For example, keyword extraction can be performed based on the large language model. This extraction method can convert language into text, and then the text can be calibrated to extract keywords such as "fire" and "location" to determine the main cable protection type and output standardized working condition parameters such as fire source intensity, bridge deck wind speed, and fire source distance. The fire condition parameters are input into the trained main cable heat transfer inversion model. The evaluation criteria for the main cable heat transfer inversion model on the test set are: mean square error (MSE) < 35, mean absolute error (MAE) < 5, and coefficient of determination (R²). 2 >0.9, and the calculation time for a single inversion should be controlled within 3 seconds; The curve of temperature change over time inside the main cable of the bridge is generated, which includes three characteristic sections: rapid heating stage, peak temperature plateau stage and slow cooling stage. It can accurately reflect the thermal inertia effect and heat transfer hysteresis characteristics of the main cable under the action of fire. In bridge vehicle fires, the location of the fire is random, and it is necessary to install temperature measurement equipment to measure the internal temperature of the main cable. First, it is difficult to install measurement equipment inside the main cable, and second, it is extremely costly to install measurement equipment throughout the entire cable. This main cable heat transfer inversion model can directly generate the internal temperature field of the main cable by only inputting the relevant operating parameters of the vehicle fire, thus solving the problem that it is not feasible to install temperature measurement equipment on the main cable. Example 1

[0021] Step S1: Simulate and construct a bridge vehicle fire simulation model using FDS software. During modeling, the two-dimensional cable plane is simplified to a one-dimensional vertical bar, and any position of the cable can be represented by the height of the vertical bar, simplifying the two-dimensional spatial coordinates to a one-dimensional position. The vehicle is simplified to a cube, with a fire source placed on its surface. Recommended maximum mesh sizes are given for different vehicle types: 20cm for a maximum heat release rate of 0-5MW; 25cm for a maximum heat release rate of 5-10MW; 30cm for a maximum heat release rate of 10-15MW; 35cm for a maximum heat release rate of 15-20MW; and 50cm for a maximum heat release rate of 20-200MW. Input the operating parameters, including fire source intensity, fire source distance, and bridge deck wind speed, into the bridge vehicle fire simulation model, and output the ambient temperature field on the cable surface. In step S2, the heat transfer simulation model of the main cable adopts a simplified equivalent model, that is, the density of the main cable in the model is the same as the density of the steel. The equivalent specific heat capacity is corrected to 60%-80% of the original specific heat capacity according to the porosity of the model, preferably 75%. The main cable is simplified to an equivalent medium of "steel wire-void" using the equivalent heat transfer method. The equivalent thermal conductivity of the main cable model at different temperatures is derived using a rectangular finite element model. The radiating surface is defined as a diffuse reflective surface with an emissivity of 0.8. The radiation direction is set to be controlled by opacity. The opacity of the steel wire and the air is set to opaque and transparent, respectively. The air is set to be non-flowing, and the thermal convection of the air is ignored. Through numerical fitting, a fourth-order polynomial was obtained for the equivalent thermal conductivity as a function of temperature: By correcting the equivalent specific heat capacity and fitting the equivalent thermal conductivity, the steel wire-void structure replaces the original main cable structure, and the simplified equivalent model is established. The ambient temperature field of the cable surface in step S1 is output as the time-temperature data inside the main cable. As shown in Table 1, the simplified equivalent model is compared with the refined model. The error is within the allowable range, and the modeling efficiency and computation time are improved. Table 1. Comparison between the refined model and the simplified equivalent model. In step S3, the time-temperature data inside the main cable obtained from the experiment, the time-temperature data inside the main cable output from step S2, and the corresponding working parameters are used to construct a dataset. The experimental data can be obtained from the Chinese invention patent "Real-time Hybrid Test Platform and Implementation Method for Vehicle-Fire-Wind Force of Substructure of Long-span Bridge" authorized by our research group with authorization announcement number CN120741035B. That is, a full-scale or large-scale model fire resistance test of the cable under controllable parameter wind, fire, and force coupling conditions is carried out, and the temperature field data inside the main cable at the height point is directly collected. The repeatable, measurable, and high-fidelity fire environment provided by this platform can greatly ensure the diversity and reliability of the original data, and lay the sample set foundation for the training of the temperature field generation model. The dataset is then preprocessed, including: the time-temperature data inside the main cable is uniformly mapped to 91 standard time points with 1-minute intervals within 0-90 minutes using linear interpolation, and the temperature data is normalized; similarly, the corresponding operating parameters are also normalized, but the temperature data is retained after the mapping process. In step S4, such as Figure 2 As shown, the generator takes the operating parameters as input as a conditional scalar c, along with a one-dimensional Gaussian noise vector. The concatenated vector [z;c] is used to generate the initial hidden state and initial cell state of the LSTM. This vector [z;c] is then mapped through a fully connected layer and interpreted as the starting point of the autoregressive generation process. The LSTM units inside the generator are then expanded sequentially over time, and at each time step, the predicted temperature from the previous time step is concatenated with the conditional scalar c. As the current input, the hidden state and cell state are updated through a gating mechanism, and then the normalized temperature value for this moment is generated through the fully connected layer and Sigmoid activation. After several time steps of iteration, the final output is obtained with a length of... Main cable temperature prediction sequence ; The discriminator compares the temperature value with the same conditional scalar c at each time step. The sequences are concatenated and used as the input sequence for the LSTM encoder. After traversing the time dimension, the hidden state at the last time step is taken as the high-dimensional feature representation of the entire sequence under the current working condition. Then, it is mapped to the truth probability between 0 and 1 through two fully connected layers and a Sigmoid activation. It is used to measure the degree of consistency between the input temperature sequence and the given operating parameters; Specifically, a training dataset was constructed based on 109 sets of main cable heat transfer simulation results, with a time range of 0 to 90 minutes and a sampling time step of 1 minute. After resampling and interpolation, a main cable temperature sequence y with a length of T=91 was obtained. Each set of data corresponds to a set of operating parameters. The operating parameters are used as the conditional scalar c of the sequence-level condition input. After normalization of each sample sequence, a "condition-sequence" sample pair (c,y) is formed. The generator uses an LSTM autoregression structure with noise and conditions; for each sample, the conditional scalar c and the noise vector z with dimension d_z=32 are concatenated along the feature dimension to form a joint vector [z;c]; First, the joint vector [z;c] is mapped through a fully connected layer to generate the initial hidden state and the initial cell state: in, , For the LSTM hidden unit dimension, the split() function splits the fully connected layer output into hidden state h0 and cell state c0, W. init This is the initialization of the weight matrix, b init It is a bias term; The generator progressively generates the main cable temperature prediction sequence in an autoregressive manner. That is, at each time step The temperature value generated at the previous moment The current input of the gating mechanism, together with the conditional scalar c. If the temperature value of the previous time step is set as the normalized value corresponding to the initial ambient temperature of the main cable, then the current input... : The current input x t Feeding into LSTM gating mechanism update ,Right now: Among them, the input gate Controls the extent to which the current input is written into the cell state; forget gate Determines which dimensions the memory from the previous moment is retained; Output gate Determine which information in the cell state is visible to the current hidden state; Here, is the Sigmoid activation function; tanh(·) is the hyperbolic tangent function; ⊙ represents the Hadamard element-wise multiplication; W ix W fx W ox W cx W represents the input weight matrix. ih W fh W oh W ch Let b represent the hidden state weight matrix of the previous time step. i b f b o b c Indicates bias; h t-1 c t-1 The previous hidden state and cell state; h t c t The current state and cell state are hidden. Candidate cell state; Introducing cell state c at each time step t and input gate Forgotten Gate and output gate Three types of gating mechanisms can automatically "remember" key stages (such as the rapid heating period) related to the temperature rise of the main cable and "forget" disturbances that are less related to the final result during the long-term temperature evolution from 0 to 90 minutes, thereby realizing time-series modeling of the entire process of "rapid heating, slow climbing, and gradual cooling" of the main cable temperature. Then, the normalized temperature prediction value for this moment is given by outputting a fully connected layer and sigmoid activation: in, , This is the transpose of the output weight matrix. As a bias, during training, the value is adjusted based on the global minimum-maximum. Inversely normalized to the physical temperature range, through an autoregressive cycle T=91 Next, the main cable temperature prediction sequence was obtained: The generator adopts an autoregressive generation method that uses the output of the previous time step as the input of the next time step, so that the generator naturally satisfies the time smoothness and continuity constraints in terms of numerical values, and can better reproduce the characteristics of gradual temperature rise and slow temperature drop of the main cable in the solid heat transfer finite element model. like Figure 3 As shown, the discriminator uses an LSTM structure to encode the entire time series, concatenating the predicted temperature at each moment with the conditional scalar c to form the input vector. Repeat the concatenation operation along the time dimension to obtain the input sequence. ; input vector The data is fed into the LSTM encoder sequentially. in, The cell state at the previous moment as determined by the discriminator. The hidden state of the discriminator at the previous time step is then used to output the "realism" probability through two fully connected layers and a sigmoid activation: in The LeakyReLU activation function is used. It is considered as a high-dimensional feature representation of the entire sequence under a given working condition, and W1 is the weight matrix of the first fully connected layer. b1 and b2 are the transpose of the weight matrix of the second fully connected layer, and b1 and b2 are the biases of each fully connected layer.

[0022] The discriminator introduces a conditional scalar c explicitly at the input layer. It can learn the physical correspondence between "operating parameters - temperature rise curve inside the main cable". When the temperature rise amplitude or cooling rate of a certain generated sequence does not match the current operating parameters, the discriminator will give a lower authenticity score, thereby pushing the generator to move closer to the real COMSOL distribution in adversarial training.

[0023] The generator introduces a reconstruction term for the main cable temperature sequence on top of the adversarial loss. The weighted combined loss function of adversarial loss and physical constraints can be written as: The discriminator loss uses the standard binary cross-entropy form: Where E represents the mathematical expectation operator; z is the noise vector; c is the conditional scalar; D(·) represents the authenticity probability function of the discriminator output; G(·) represents the internal temperature sequence function of the main cable generated by the generator; y represents the physical constraint weight hyperparameter; y represents the actual main cable temperature sequence sampled from the real data distribution; Pdata represents the data distribution of the actual internal temperature sequence of the main cable, representing the overall characteristics of the temperature data obtained from COMSOL simulation or experiment. During the training phase of the heat transfer inversion model for the main cable, the Adam optimization algorithm can be used, with a learning rate set to 2×10. -4 During the evaluation phase of the main cable heat transfer inversion model, the mean square error (MSE), mean absolute error (MAE), and coefficient of determination were calculated using the inversely normalized temperature sequences of the validation and test sets. The minimum MAE of the validation set is used as the criterion for selecting the optimal model. At the same time, comparison charts of "real curves and generated curves" under several typical working conditions are output to intuitively verify the model's fitting effect on the characteristics of the main cable's temperature rise and cooldown stages. Example 2

[0024] Four representative operating conditions were selected as test samples, with the maximum ambient temperature of the main cable being 700℃ (45MW, 0 m / s, 0.5m), 900℃ (45MW, 3m / s, 0.5m), 1000℃ (100MW, 6m / s, 1.5m), and 1100℃ (100MW, 3m / s, 1.5m) under different combinations of operating parameters (fire source intensity, bridge deck wind speed, and fire source distance). The heat transfer inversion model of the main cable in this method was compared and analyzed with the COMSOL transient heat conduction simulation results, that is, the comparison chart of the predicted temperature and the simulated temperature under the four sets of temperatures was obtained. The selected operating conditions cover a range from moderate operating conditions with relatively small temperature rise and short fire duration to extreme unfavorable operating conditions with high ambient temperature and long duration of main cable heating, which can reflect the model's generalization ability in typical and extreme scenarios to a certain extent. Based on the overall curve comparison, a quantitative statistical evaluation was performed on the temperature time series of all main cables within the test set. The results show that the mean squared error (MSE) of the model on the test set is 32.29, the mean absolute error (MAE) is 4.77, and the coefficient of determination (R²) is [missing value]. 2=0.9926; Converted to physical quantities, the average deviation of the model at each time step in the entire temperature history is about 5℃, which can statistically reproduce the temperature response of the main cable obtained by finite element calculation well, providing sufficient numerical accuracy reserve for post-disaster assessment; On the same computing platform, the internal temperature history of the main cable from 0 to 90 minutes can be generated within 2 seconds by inputting the operating conditions parameters once, so that the temperature change curve of the main cable after the fire occurs and decays can be quickly obtained in the post-disaster assessment scenario; like Figure 4 As shown in the temperature curves under typical operating conditions, the trend of the main cable temperature predicted by the model in this method over time is highly consistent with the results of the COMSOL transient thermal conductivity simulation. The network prediction curve can accurately reproduce the rapid temperature rise stage at the beginning of the fire, the slow rise stage near the peak, and the slow cooling stage after the fire decays, without any obvious phase lag or advance. Near the peak temperature and its occurrence time, which are of most concern in engineering design, the envelope shapes of the two are basically overlapping, indicating that the model has a strong fitting ability for key temperature response characteristics such as peak size, occurrence time, and cooling slope. The two maintain good overlap for most of the entire process, and the error range is within a reasonable range acceptable for engineering.

[0025] The foregoing description, with reference to preferred embodiments, details an exemplary implementation of the method for inverting the internal temperature of the main cable under a bridge vehicle fire proposed by the present invention. However, those skilled in the art will understand that various modifications and alterations can be made to the above specific embodiments without departing from the concept of the present invention, and various combinations can be made to the various technical features and structures proposed by the present invention without exceeding the protection scope of the present invention, which is determined by the appended claims.

Claims

1. A method for inverting the internal temperature of the main cable under a bridge vehicle fire, characterized in that, Specifically, the following steps are included: S1, obtain the working condition parameters, which include fire source intensity, fire source distance, and bridge deck wind speed. Input the working condition parameters into the bridge vehicle fire simulation model and output the ambient temperature field of the cable surface. S2, input the cable surface temperature field into the main cable heat transfer simulation model, and output the time-temperature data inside the main cable; S3, construct a dataset by combining the time-temperature data inside the main cable obtained from the experiment, the time-temperature data inside the main cable output from step S2, and the corresponding operating parameters, preprocess the dataset, and divide it into a training set, a validation set, and a test set; S4. Construct a main cable heat transfer inversion model based on LSTM conditional generative adversarial network. Use a weighted combination loss function that includes adversarial loss and physical constraints to train the generator of the LSTM conditional generative adversarial network model to obtain the trained main cable heat transfer inversion model. S5: Obtain the operating parameters of the bridge vehicle fire scenario, input them into the trained main cable heat transfer inversion model, output the predicted temperature value inside the main cable within a preset time range, and generate the time-temperature change curve inside the bridge main cable based on the predicted temperature value.

2. The method for inverting the internal temperature of the main cable under a bridge vehicle fire according to claim 1, characterized in that, In step S4, the main cable heat transfer inversion model has a generator and a discriminator; The generator uses an LSTM autoregressive structure and includes the following steps: The input layer receives a conditional scalar c defined by operating parameters, and generates the initial hidden state and initial cell state of the LSTM through a fully connected layer. At each time step, the predicted temperature from the previous time step is concatenated with the conditional scalar c as input. The hidden state and cell state are updated through the LSTM gating mechanism. The output layer uses the Sigmoid activation function to generate normalized temperature values, thus obtaining the complete main cable temperature prediction sequence. The discriminator uses a conditional LSTM structure, specifically including the following steps: At each time step, the temperature value is concatenated with the conditional scalar c to form the input sequence. Temporal features are extracted by an LSTM encoder, and the final hidden state outputs the probability of authenticity through a fully connected layer and a Sigmoid activation function.

3. The method for inverting the internal temperature of the main cable under a bridge vehicle fire according to claim 2, characterized in that, In the generator, the conditional scalar c is concatenated with the noise vector z to form a joint vector [z;c], which is then input into the input layer.

4. The method for inverting the internal temperature of the main cable under a bridge vehicle fire according to claim 3, characterized in that, The generator introduces a reconstruction term of the main cable temperature sequence on top of the adversarial loss, and the comprehensive loss is: The discriminator loss uses the standard binary cross-entropy form: Where E represents the mathematical expectation operator; D(·) represents the probability function of the discriminator's output; G(·) represents the temperature sequence function of the main cable's internal temperature generated by the generator; y represents the physical constraint weight hyperparameter; y represents the actual internal temperature sequence of the main cable sampled from the real data distribution; Pdata represents the data distribution of the actual main cable temperature sequence.

5. The method for inverting the internal temperature of the main cable under a bridge vehicle fire according to claim 4, characterized in that, The generator progressively generates a main cable temperature prediction sequence in an autoregressive manner. Specifically, it includes the following steps: The joint vector [z;c] generates the initial hidden state and the initial cell state through a fully connected layer: in, , For the LSTM hidden unit dimension, the split() function splits the fully connected layer output into hidden state h0 and cell state c0, W. init This is the initialization of the weight matrix, b init It is a bias term; At each time step The temperature value generated at the previous moment With conditional scalar c The current inputs that make up the gating mechanism: The current input x t Feeding into LSTM gating mechanism update ; Then, the normalized temperature prediction value for this moment is given by outputting a fully connected layer and sigmoid activation: in, , It is the Sigmoid activation function. h is the transpose of the output weight matrix. t Hide the current state. For bias; During training, the global minimum and maximum values ​​are then used to... Inversely normalized to the physical temperature range, through the autoregressive cycle T The main cable temperature prediction sequence was obtained next: 。 6. The method for inverting the internal temperature of the main cable under a bridge vehicle fire according to claim 4, characterized in that, In the discriminator, the formula for calculating the "authenticity" probability of the output through two fully connected layers and Sigmoid activation is as follows: in, It is the Sigmoid activation function. Use the LeakyReLU activation function; It is considered as a high-dimensional feature representation of the entire sequence under a given working condition, and W1 is the weight matrix of the first fully connected layer. b1 and b2 are the transpose of the weight matrix of the second fully connected layer, and b1 and b2 are the biases of each fully connected layer.

7. A method for inverting the internal temperature of the main cable under a bridge vehicle fire according to any one of claims 1 to 6, characterized in that, In step S2, the main cable structure is regarded as a "steel wire-void" porous medium in the heat transfer simulation model of the main cable. By fitting the equivalent thermal conductivity and correcting the equivalent specific heat capacity, a simplified equivalent model of the steel wire-void main cable structure is established. Finally, the ambient temperature field of the cable surface in step S1 is input into the simplified equivalent model, and the time-temperature data inside the main cable is output.

8. The method for inverting the internal temperature of the main cable under a bridge vehicle fire according to claim 7, characterized in that, The simplified equivalent model is derived using a rectangular finite element model. The rectangular finite element model consists of the main cable steel wires and the air gaps between the steel wires. The internal steel wire dimensions and porosity are consistent with the experimental model. There is surface-to-surface thermal radiation between the steel wire gaps. The radiating surface is defined as a diffuse reflective surface and its emissivity is set. The radiation direction is set to be controlled by opacity. The opacities of the steel wires and air are set to opaque and transparent, respectively. The air is set to be non-flowing, and the thermal convection of the air is ignored. The equivalent thermal conductivity of the simplified equivalent model is derived through the rectangular finite element model, and its calculation formula is as follows: in, The average heat flux density is denoted as H; H is the thickness of the main cable protective layer; T1 and T2 are the upper and lower boundary temperatures, respectively. Numerical fitting was performed on the equivalent thermal conductivity, and a fourth-order polynomial was used to obtain the formula for calculating the equivalent thermal conductivity as a function of temperature: in, , , , For temperature coefficient, It is a constant. Ambient temperature; The equivalent specific heat capacity is corrected to 60%-80% of the original specific heat capacity based on the porosity of the model.

9. A method for inverting the internal temperature of the main cable under a bridge vehicle fire according to any one of claims 1 to 6, characterized in that, In step S3, the dataset is preprocessed. Specifically, the time-temperature data calculated and output by the main cable heat transfer simulation model is uniformly mapped to 91 standard time points with 1-minute intervals within 0-90 minutes using linear interpolation, and the temperature data is normalized.

10. The method for inverting the internal temperature of the main cable under a bridge vehicle fire according to claim 9, characterized in that, In step S4, the training rounds are 300-500, and the minimum mean absolute error of the validation set is used as the early stopping criterion. In step S5, the evaluation index of the main cable heat transfer inversion model on the test set is: mean square error (MSE) < 35, mean absolute error (MAE) < 5, and coefficient of determination (R²). 2 >0.9, and the calculation time for a single inversion is controlled within 3 seconds.

Images