Bridge cable fire prevention method and system based on localization large model

Abstract

The application discloses a bridge cable fire prevention method and system based on a localized large model, and relates to the technical field of bridge fire prevention. The fire prevention method comprises the following steps: S10: based on a large language model, key working condition parameters of a bridge cable vehicle fire scene are extracted and confirmed through multiple rounds of human-computer interaction, and a standardized working condition parameter dataset is obtained; S20: based on the working condition parameter dataset, a local knowledge base is called to verify parameters, and a deep learning model calling instruction for a bridge cable vehicle fire temperature prevention curve is generated; S30: based on the deep learning model calling instruction, a corresponding temperature field prediction model is loaded and trained, a temperature field space-time evolution simulation calculation is performed, and a temperature field prediction result is output; and S40: the working condition parameter dataset and the temperature field prediction result are fused to generate a report containing a parameter vector, a bridge cable vehicle fire temperature prevention curve, a prevention index and a prevention suggestion.

Description

Technical Field

[0001] This invention relates to the field of bridge fire protection technology, and in particular to a method and system for fire protection of bridge cables based on a localized large model. Background Technology

[0002] In recent years, with the widespread application of long-span cable-stayed bridges in important projects such as those spanning seas and rivers, fire safety issues have become increasingly prominent. Bridge-vehicle fires have occurred frequently in recent years, and the high-temperature flames and hot smoke generated by burning vehicles pose a serious threat to the bridge's critical load-bearing components (such as main cables, suspenders, stay cables, and bridge towers). The mechanical properties of steel are extremely sensitive to temperature. When the temperature of the cable exceeds its material critical value, its strength and stiffness will significantly degrade, potentially leading to overall structural failure and catastrophic consequences.

[0003] Therefore, in bridge fire protection design, scientifically determining the temperature effect of high temperatures during a fire on bridge cables, i.e., developing a temperature protection curve, is the prerequisite and key basis for carrying out fire-resistant design and taking targeted fire prevention measures.

[0004] In the field of bridge fire protection design, some technologies have been developed to address the shortcomings of traditional methods. However, these advanced technologies still face significant challenges in practical engineering applications: 1. Computational Fluid Dynamics (CFD) Numerical Simulation Technology: With the development of computer technology, it has become possible to use CFD software (such as FDS and Fluent) for detailed simulation of fire scenarios. This method can relatively accurately simulate the combustion of fire sources, smoke flow, and heat transfer processes, thereby obtaining the temperature field distribution under specific operating conditions. However, this method has an insurmountable bottleneck: High computational cost: A single complete CFD simulation can take hours or even days, and to obtain a defense curve that encompasses multiple possibilities, hundreds or thousands of simulations need to be performed, which is unacceptable in terms of time constraints and costs in engineering design. Highly specialized operation: CFD simulation demands extremely high levels of professional knowledge and experience from the operator; non-experienced engineers find it difficult to ensure the correctness of model setup and the reliability of results; Result integration is challenging. CFD simulation is essentially "condition-isolated," providing only point-to-point solutions. Extracting, integrating, and ultimately forming a representative defense curve from massive simulation results heavily relies on engineers' manual post-processing and experience-based judgment, a cumbersome process that is highly subjective and difficult to standardize.

[0005] 2. Preliminary Attempts at Intelligent Technologies: In recent years, research has attempted to utilize machine learning or deep learning models to rapidly predict fire temperature fields, aiming to replace time-consuming CFD simulations. These technologies have addressed the computational efficiency issue to some extent. However, they still face significant challenges in practical engineering applications: The "model silo" phenomenon: These advanced models often exist as isolated tools with limited functionality. They may be able to quickly predict the temperature distribution under a single operating condition, but they cannot automatically complete the full workflow necessary for generating defense curves: "multi-condition setting - batch calculation - envelope extraction - physical correction". Engineers still need to manually set up numerous operating conditions, drive model calculations, and manually compare the results, resulting in limited intelligence.

[0006] High barrier to entry: Preparing input parameters, calling the model, and parsing the results of professional models require certain programming or scripting knowledge, which keeps many front-line designers out.

[0007] Data security risks: If these models are provided as cloud APIs, sensitive design parameters such as bridge structural dimensions and geographical location must be uploaded to third-party platforms, posing serious privacy and data security risks. This is a fatal flaw for major infrastructure projects.

[0008] In summary, a significant gap exists in existing technologies: on the one hand, high-precision CFD simulation technology is difficult to directly apply to efficient defense analysis due to its low efficiency and operational complexity; on the other hand, emerging intelligent prediction technologies cannot be implemented due to insufficient system integration, poor usability, and security risks. Currently, there is a lack of an integrated solution that can deeply integrate the efficiency of artificial intelligence with the rigor and security requirements of engineering defense. Especially in localized deployment environments, how to organically integrate the intent understanding and task scheduling capabilities of large language models, the rapid prediction capabilities of professional models, and the normative constraints of knowledge bases using natural language as the interaction interface to form a secure, efficient, easy-to-use, and fully automated intelligent defense system, thereby completely liberating design productivity, remains a long-standing unsolved technical challenge in this field.

[0009] Therefore, developing a new method that can overcome the above-mentioned defects has become an urgent technical issue with significant practical value in the field of bridge fire safety. Summary of the Invention

[0010] This solution addresses the problems and needs raised above by proposing a bridge cable fire protection method and system based on a localized large model. The above technical objectives are achieved by adopting the following technical features, and other technical effects are also brought about.

[0011] One objective of this invention is to propose a fire protection method for bridge cables based on a localized large model, comprising the following steps: S10: Based on a large language model, key working condition parameters of bridge cable vehicle fire scenarios are extracted and confirmed through multiple rounds of human-computer interaction to obtain a standardized working condition parameter dataset. S20: Based on the aforementioned working condition parameter dataset, call the local knowledge base to verify the parameters and generate a deep learning model call instruction for the temperature protection curve of bridge cable vehicle fire. S30: Based on the deep learning model call instruction, load and obtain the corresponding temperature field prediction model and train it, perform temperature field spatiotemporal evolution simulation calculation, and output temperature field prediction results; S40: Integrate the working condition parameter dataset and the temperature field prediction results to generate a report containing parameter vectors, temperature protection curves for bridge cable vehicle fires, protection indicators, and protection recommendations.

[0012] Furthermore, the bridge cable fire protection method and system based on a localized large model according to the present invention may also have the following technical features: In one example of the present invention, in step S10, key operating condition parameters of a bridge cable vehicle fire scenario are extracted and confirmed through multiple rounds of human-computer interaction based on a large language model, and a standardized operating condition parameter dataset is obtained. The specific process includes: Leveraging the semantic understanding capabilities of large language models, we can identify fire scene parameter entities that are implicit or explicit in user input. For key condition parameters that are not clearly defined, the inquiry logic is initiated; for condition parameters that are ambiguous or exceed a reasonable range, the confirmation and correction logic is initiated, and finally, a complete and error-free standard chemical condition parameter set is output.

[0013] In one example of the present invention, the standardized operating condition parameter set includes: fire source power, fire source distance, and bridge deck wind speed, wherein the fire source power is a fixed value; the fire source distance parameter set is obtained by sampling in a uniform distribution in [0,s], where s is the maximum fire source distance, and the maximum fire source distance is the distance between the adjacent lane of the emergency lane and the cable; the bridge deck wind speed is sampled according to the Weibull distribution. This involves arranging and combining three operating parameters to achieve comprehensive coverage of different fire scenarios through this hybrid sampling strategy.

[0014] In one example of the present invention, step S30 involves loading and training the corresponding temperature field prediction model, specifically including the following steps: S31: Construct a bridge fire simulation dataset: Generate bridge height-temperature data under different standardized working condition parameter combinations using simulation software; S32: Data preprocessing and standardization: The non-uniform height-temperature distribution in the simulation data is uniformly mapped to the standard height sequence through an interpolation algorithm, and the working parameters and temperature values ​​are normalized to form a working parameter set. The working parameter set is divided into a training set, a validation set and a test set. S33: Constructing a Conditional Generative Adversarial Network Model: A conditional generative adversarial network model is constructed by a generator and a discriminator. The generator is configured to generate predicted temperature curves based on noise vectors and normalized working condition parameters. The discriminator is configured to distinguish between the generated temperature curves and the actual temperature curves and output a discrimination probability value. S34: Define the weighted loss function: The generator adopts a weighted combined loss function that includes adversarial loss and physical constraints. The conditional generative adversarial network model is trained through the training set. The adversarial loss ensures that the generated distribution approximates the real distribution, and the physical constraints ensure that the generated curve is smooth and meets the corresponding physical laws. S35: Model Training and Optimization: Based on gradient descent optimization algorithms and their hyperparameter configuration strategies, a conditional generative adversarial network model is trained using a training set with an alternating iterative update strategy of discriminator and generator. During training, the training effect is monitored using a validation set to adjust the training parameters. The hyperparameter learning efficiency is 2×10⁻⁶. -4 The minimum number of training rounds is 500 to ensure prediction accuracy, and the maximum is 1000 to prevent overfitting. S36: Model Validation and Deployment: Evaluate the performance of the conditional generative adversarial network model on the test set, using the coefficient of determination R0. 2 The conditional generative adversarial network model is evaluated using the mean absolute error (MAE), where the coefficient of determination R0 is the mean absolute error (MAE). 2 >0.9, with a mean absolute error (MAE) <50℃. After successful verification, the weight file of the fully trained conditional generative adversarial network model will be deployed to a local professional model library for use by large language models.

[0015] In one example of the present invention, in step S33, the generator includes an input layer, three first fully connected hidden layers, and a first output layer; the input layer is a concatenated vector of a noise vector and normalized working condition parameters, used for data input and feature initialization; the three first fully connected hidden layers are connected sequentially, each followed by a batch normalization layer and a LeakyReLU activation function, used for feature extraction and nonlinear transformation at different levels, respectively; the first output layer includes a first fully connected layer and a Tanh activation function, used to output temperature prediction values ​​for multiple standard altitude points; The discriminator is a two-stream network structure, including a conditional stream, a data stream, a concatenation layer, two second fully connected hidden layers, and a second output layer. The conditional stream includes a second fully connected layer and a LeakyReLU activation function, used to process the operating condition vector and extract conditional features. The data stream includes a third fully connected layer and a LeakyReLU activation function, used to process the temperature curve vector and extract data features. The concatenation layer concatenates the feature vectors output from the conditional stream and the data stream into a fused feature vector. The two second fully connected hidden layers are connected sequentially, with each second fully connected hidden layer followed by a LeakyReLU activation function, used for feature fusion and discriminant analysis, respectively. The second output layer includes a fourth fully connected layer and a Sigmoid activation function, used to output the discriminant probability value.

[0016] In one example of the present invention, in step S34, the generator uses a weighted combination loss function L. total From the counter-loss L adv and physical constraint loss L smooth Weighted according to the following formula: Among them, the resistance loss L adv The calculation formula is: Among them, physical constraint loss L smooth The calculation formula is: In the formula, λ is the hyperparameter; E represents the mathematical expectation; P Z (Z) represents the probability distribution of random noise; P data (C) represents the data distribution under actual operating conditions; D(.) represents the output probability of the discriminator; G(.) represents the output of the generator; N represents the length of the temperature sequence; T i This represents the predicted temperature value at the i-th elevation point; The loss function used by the discriminator The expression is: In the formula, This represents a discriminator that outputs the probability of "true" under given operating conditions.

[0017] In one example of the present invention, generating the temperature protection curve for bridge cable vehicle fires in step S40 includes the following steps: S41: Abnormal data cleaning: Based on the correction method of statistical distribution and neighboring point interpolation, abnormal temperature points are identified and corrected for the distribution curves generated under all working conditions; for each standard height point, the distribution of temperature values ​​of all working conditions at that point is statistically analyzed, an anomaly discrimination threshold based on statistical quantiles is set, temperature points exceeding the threshold are identified as abnormal values, and the correction values ​​are calculated by linear interpolation using the temperature values ​​of the distribution curve at adjacent height points where the abnormal value is located and then replaced. S42: Envelope Extraction: Based on the cleaned temperature distribution curve dataset, extract the outer envelope of all curves along the height direction; S43: Envelope Physical Correction: Apply a monotonic physical constraint to the extracted preliminary outer envelope to ensure that it satisfies the law that temperature decays with increasing altitude.

[0018] In one example of the present invention, in step S43, a monotonic physical constraint is applied to the extracted preliminary outer envelope to ensure that it satisfies the law of temperature decay with increasing altitude, specifically including: Starting from the bridge deck and proceeding upwards, check each point one by one. If the temperature value of the current elevation point is lower than the temperature value of the subsequent elevation points, then correct the temperature value of the abnormal point to the maximum value in the temperature sequence of the subsequent elevation points.

[0019] In one example of the present invention, the generated report in step S40 includes: a project summary, a list of input operating parameters, a temperature protection curve, key safety indicator data, a risk level assessment, and specific graded protection recommendations.

[0020] Another objective of this invention is to propose a bridge cable fire protection system based on a localized large model, comprising: The human-computer interaction and parameter extraction module is configured to extract and confirm key working condition parameters of bridge cable vehicle fire scenarios through multiple rounds of human-computer interaction based on a large language model, and obtain a standardized working condition parameter dataset. The large language model scheduling and knowledge management module is configured to call the local knowledge base for parameter verification based on the working condition parameter dataset, and generate a deep learning model calling instruction for the temperature protection curve of bridge cable vehicle fire. The deep learning module is configured to load and train the corresponding temperature field prediction model based on the deep learning model call instruction, perform temperature field spatiotemporal evolution simulation calculation, and output temperature field prediction results. The structured report generation module is configured to integrate the working condition parameter dataset and the temperature field prediction results to generate a report containing parameter vectors, temperature protection curves for bridge cable vehicle fires, protection indicators, and protection recommendations.

[0021] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention represents a paradigm shift from "manually driven" to "intelligent automation." Traditional methods rely on engineers manually setting massive amounts of operational conditions and performing tedious post-processing. This invention uses a large language model as the intelligent scheduling hub, integrating multiple discrete steps such as parameter parsing, operational condition combination, model invocation, result integration, and report generation into a fully automated pipeline. Users only need to describe their defense requirements in natural language, and the system can automatically complete all the work that previously required weeks of manual labor. This reduces the efficiency of defense curve generation from days or even tens of days to minutes, completely solving the core pain point of low efficiency in traditional methods.

[0022] 2. A dual guarantee of "physical laws + data-driven" is constructed, significantly improving the scientific rigor and reliability of the defense curve. This invention is not a simple data fitting. On the one hand, the deep learning model incorporates physical law constraints during training, ensuring that its predictions conform to basic thermodynamic principles. On the other hand, during the defense curve generation stage, a multi-condition coverage strategy and conservative envelope extraction are used, along with strict physical corrections (such as monotonicity constraints), to ensure that the final defense curve covers various potential risk scenarios while meeting engineering conservatism requirements. Its scientific rigor and reliability far exceed those of traditional results relying on a single condition or empirical formula.

[0023] 3. It ensures the complete localization of core data and models, fundamentally eliminating the risk of sensitive information leakage. All processing, calculation, and storage of sensitive data involving bridge structural parameters, geographic information, etc., are completed within the organization's internal network server cluster, completely eliminating the risk of data leakage that may result from using cloud services. This feature makes this invention particularly suitable for key bridge engineering projects with extremely high information security requirements, providing a secure and reliable solution for the application of intelligent technology in critical infrastructure.

[0024] 4. Lowering the professional threshold and empowering frontline designers to produce expert-level results. Through a natural language interface and fully automated workflow, this invention encapsulates complex fire simulation and fortification curve generation technologies into an easy-to-use tool. Non-senior experts or frontline designers do not need in-depth knowledge of fire dynamics or numerical simulation to quickly generate professional fortification schemes that meet regulatory requirements, greatly lowering the technical application threshold and facilitating the promotion and popularization of advanced design methods.

[0025] 5. The system adopts a modular and standardized architecture, possessing excellent maintainability and scalability. Each core module is encapsulated using containerization technology and communicates through standard API interfaces. This highly cohesive and loosely coupled architecture design allows any component in the system (such as a new temperature prediction model or an updated design specification knowledge base) to be independently upgraded or replaced without requiring a complete system reconstruction. This ensures that the invention can continuously absorb technological advancements, adapt to future standard and requirement evolution, and possess a long lifespan and application value.

[0026] The preferred embodiments of the invention will be described in more detail below with reference to the accompanying drawings, so as to facilitate an understanding of the features and advantages of the invention. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. The drawings are merely illustrative of some embodiments of the present invention and are not intended to limit the scope of the present invention to all embodiments.

[0028] Figure 1 A flowchart illustrating a bridge cable fire protection method based on a localized large model according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the generator according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the discriminator according to an embodiment of the present invention. Detailed Implementation

[0029] 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.

[0030] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as 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.

[0031] According to a first aspect of the present invention, a method for fire protection of bridge cables based on a localized large model is provided, such as... Figure 1 As shown, it includes the following steps: S10: Based on a large language model, key working condition parameters of bridge cable vehicle fire scenarios are extracted and confirmed through multiple rounds of human-computer interaction to obtain a standardized working condition parameter dataset. S20: Based on the aforementioned working condition parameter dataset, call the local knowledge base to verify the parameters and generate a deep learning model call instruction for the temperature protection curve of bridge cable vehicle fire. S30: Based on the deep learning model call instruction, load and obtain the corresponding temperature field prediction model and train it, perform temperature field spatiotemporal evolution simulation calculation, and output temperature field prediction results; S40: Integrate the working condition parameter dataset and the temperature field prediction results to generate a report containing parameter vectors, temperature protection curves for bridge cable vehicle fires, protection indicators, and protection recommendations.

[0032] This fire protection design method represents a paradigm shift from "manually driven" to "intelligently automated." Traditional methods rely on engineers manually setting up massive amounts of operational conditions and performing tedious post-processing. This new method uses a large language model as the intelligent scheduling hub, integrating multiple discrete steps such as parameter parsing, operational condition combination, model invocation, result integration, and report generation into a fully automated pipeline. Users only need to describe their fire protection requirements in natural language, and the system can automatically complete all the work that previously required weeks of manual labor. This reduces the efficiency of fire protection curve generation from days or even tens of days to minutes, completely resolving the core pain point of inefficiency in traditional methods.

[0033] This fire protection design method establishes a dual guarantee of "physical laws + data-driven approach," significantly improving the scientific rigor and reliability of the design curve. This method is not simply data fitting. On one hand, the deep learning model incorporates physical constraints during training, ensuring its predictions conform to fundamental thermodynamic principles. On the other hand, during the design curve generation stage, a multi-condition coverage strategy and conservative envelope extraction are employed, along with stringent physical corrections (such as monotonicity constraints). This ensures that the final design curve covers various potential risk scenarios while meeting engineering conservatism requirements. Its scientific rigor and reliability far exceed traditional results relying on single conditions or empirical formulas.

[0034] This fire protection method ensures the complete localization of core data and models, fundamentally eliminating the risk of sensitive information leakage. All processing, calculation, and storage of sensitive data involving bridge structural parameters, geographic information, etc., are completed within the organization's internal network server cluster, completely eliminating the risk of data leakage that may result from using cloud services. This characteristic makes this fire protection method particularly suitable for key bridge engineering projects with extremely high information security requirements, providing a secure and reliable solution for the application of intelligent technologies in critical infrastructure.

[0035] This fire protection design method lowers the professional threshold, empowering frontline designers to produce expert-level results. Through a natural language interface and a fully automated workflow, this method encapsulates complex fire simulation and design curve generation technologies into an easy-to-use tool. Non-senior experts or frontline designers do not need in-depth knowledge of fire dynamics or numerical simulation to quickly generate professional fire protection schemes that meet regulatory requirements, greatly reducing the technical application threshold and facilitating the promotion and popularization of advanced design methods.

[0036] This fire protection method employs a modular and standardized system architecture, boasting excellent maintainability and scalability. Each core module is encapsulated using containerization technology and communicates via standard API interfaces. This highly cohesive and loosely coupled architecture allows any component in the system (such as a new temperature prediction model or an updated design specification knowledge base) to be independently upgraded or replaced without requiring a complete system refactoring. This ensures that the fire protection method can continuously absorb technological advancements, adapt to future changes in specifications and requirements, and possesses a long lifespan and application value.

[0037] In one example of the present invention, in step S10, key operating condition parameters of a bridge cable vehicle fire scenario are extracted and confirmed through multiple rounds of human-computer interaction based on a large language model, and a standardized operating condition parameter dataset is obtained. The specific process includes: Leveraging the semantic understanding capabilities of large language models, we can identify fire scene parameter entities that are implicit or explicit in user input. For key condition parameters that are not clearly defined, the inquiry logic is initiated; for condition parameters that are ambiguous or exceed a reasonable range, the confirmation and correction logic is initiated, and finally, a complete and error-free standard chemical condition parameter set is output.

[0038] In one example of the present invention, the standardized operating condition parameter set includes: fire source power, fire source distance, and bridge deck wind speed, wherein the fire source power is a fixed value; the maximum fire source distance s is the distance between the adjacent lane of the emergency lane and the cable, and the fire source distance parameter set is obtained by sampling in a uniform distribution in [0,s]; the bridge deck wind speed is sampled according to a Weibull distribution; its probability density function is... The expression is as follows: In the formula, v represents the annual average wind speed of the design area, and k and c are two parameters of the Weibull distribution. k is called the shape parameter, and c is called the scale parameter. When c=1, it is called the standard Weibull distribution. The value of the shape parameter k is adjusted according to local conditions to make the wind speed distribution conform to reality. The smaller the k, the greater the variation in annual average wind speed; the larger the k, the smaller the variation in annual average wind speed.

[0039] This involves arranging and combining three operating parameters to achieve comprehensive coverage of different fire scenarios through this hybrid sampling strategy.

[0040] In one example of the present invention, step S30 involves loading and training the corresponding temperature field prediction model, specifically including the following steps: S31: Construct a bridge fire simulation dataset: Generate bridge height-temperature data under different standardized working condition parameter combinations using simulation software. The working condition parameters include fire source power, bridge deck wind speed, and fire source distance. S32: Data Preprocessing and Standardization: The non-uniform height-temperature distribution in the simulation data is uniformly mapped to the standard height sequence through an interpolation algorithm. The operating parameters and temperature values ​​are normalized to form an operating parameter set, which is then divided into a training set, a validation set, and a test set. For example, the same processing is performed on the experimental data to form an operating parameter set. The two sets of operating parameter sets are mixed and divided into a training set, a validation set, and a test set in a ratio of 8:1:1.

[0041] S33: Constructing a Conditional Generative Adversarial Network Model: This model is a generator-discriminator dual-path system, consisting of a generator and a discriminator forming a conditional generative adversarial network model. The generator is configured to generate predicted temperature curve values ​​based on noise vectors and normalized working condition parameters; the discriminator is configured to distinguish between the generated temperature curve and the real temperature curve and output a discrimination probability value. S34: Define the weighted loss function: The generator adopts a weighted combined loss function that includes adversarial loss and physical constraints. The conditional generative adversarial network model is trained through the training set. The adversarial loss ensures that the generated distribution approximates the real distribution, and the physical constraints ensure that the generated curve is smooth and meets the corresponding physical laws. S35: Model Training and Optimization: Based on gradient descent optimization algorithms and their hyperparameter configuration strategies, a conditional generative adversarial network model is trained using a training set with an alternating iterative update strategy of discriminator and generator. During training, the training effect is monitored using a validation set to adjust the training parameters. The hyperparameter learning efficiency is 2×10⁻⁶. -4 The minimum number of training rounds is 500 to ensure prediction accuracy, and the maximum is 1000 to prevent overfitting. S36: Model Validation and Deployment: Evaluate the performance of the conditional generative adversarial network model on the test set, using the coefficient of determination R0. 2 The conditional generative adversarial network model is evaluated using the mean absolute error (MAE), where the coefficient of determination R0 is the mean absolute error (MAE). 2 >0.9, with a mean absolute error (MAE) <50℃. After successful verification, the weight file of the fully trained conditional generative adversarial network model will be deployed to a local professional model library for use by large language models.

[0042] In one example of the present invention, step S31, constructing the bridge vehicle fire simulation model specifically includes: The two-dimensional cable plane is simplified into a one-dimensional vertical bar, and any position of the cable can be represented by the height position of the vertical bar, thus simplifying the two-dimensional spatial coordinates into a one-dimensional position. The vehicle is simplified into a cube with a fire source on its surface. Recommended maximum grid sizes are given for different vehicle types: 20cm for a maximum heat release rate of 0-5MW; 25cm for 5-10MW; 30cm for 10-15MW; 35cm for 15-20MW; and 50cm for 20-200MW. The maximum heat release efficiency of the flame is simulated for 8-12 minutes to equivalently replace the highest temperature of the actual flame simulated for 1.5 hours. Temperature measuring points are set at equal intervals along the height direction on the surface of the vertical rod according to the required accuracy.

[0043] In one example of the present invention, in step S31, the operating parameters include: fire source intensity, bridge deck wind speed, and fire source distance; wherein, the fire source intensity is densely sampled in a uniform distribution in the range of 0-50MW, and the fire source intensity is sparsely sampled in a uniform distribution in the range of 50-200MW; the bridge deck wind speed is sampled in a uniform distribution in the range of 0-15m / s; and the fire source distance is the distance between the emergency lane and the adjacent lane from the edge of the main cable.

[0044] In one example of this invention, in step S32, experimental data is collected using the Chinese invention patent application number 2025111956407, "Real-time Hybrid Test Platform and Implementation Method for Vehicle-Fire-Wind Force Mixed Test of Substructure of Long-Span Bridge," to conduct a full-scale or large-scale model fire resistance test of the cable under controllable parameter wind, fire, and force coupling conditions, directly collecting temperature field data at height points. The repeatable, measurable, and high-fidelity fire environment provided by this platform greatly ensures the diversity and reliability of the original data, laying the foundation for a sample set for training the temperature field generation model.

[0045] In one example of the present invention, in step S32, the preprocessing of the dataset includes: mapping the non-uniform height-temperature data output by the simulation model to a standard height sampled at equal intervals within 0-35m using linear interpolation; and normalizing the working condition parameter matrix and the temperature value matrix respectively.

[0046] In one example of the present invention, in step S33, as Figure 2 and Figure 3 As shown, the generator includes an input layer, three first fully connected hidden layers, and a first output layer. The input layer is a concatenated vector of a received noise vector and normalized working condition parameters, used for data input and feature initialization. The three first fully connected hidden layers are connected sequentially, with each first fully connected hidden layer followed by a batch normalization layer and a LeakyReLU activation function, used for feature extraction and nonlinear transformation at different levels, respectively. The first output layer includes a first fully connected layer and a Tanh activation function, used to output temperature prediction values ​​for multiple standard altitude points. The discriminator is a two-stream network structure, including a conditional stream, a data stream, a concatenation layer, two second fully connected hidden layers, and a second output layer. The conditional stream includes a second fully connected layer and a LeakyReLU activation function, used to process the operating condition vector and extract conditional features. The data stream includes a third fully connected layer and a LeakyReLU activation function, used to process the temperature curve vector and extract data features. The concatenation layer concatenates the feature vectors output from the conditional stream and the data stream into a fused feature vector for feature fusion. The two second fully connected hidden layers are connected sequentially, with each second fully connected hidden layer followed by a LeakyReLU activation function, used for feature fusion and discriminant analysis, respectively. The second output layer includes a fourth fully connected layer and a Sigmoid activation function, used to output the discriminant probability value.

[0047] Specifically, the generator structure is as follows: the input layer receives a concatenated vector of noise vector and normalized operating condition parameters; it passes sequentially through first fully connected hidden layers of 256 nodes, 512 nodes, and 256 nodes, with each first fully connected hidden layer followed by a batch normalization layer and a LeakyReLU activation function; the output layer is a first fully connected layer of 36 nodes, using the corresponding Tanh activation function; the discriminator is a two-stream network structure, where the conditional stream is a second fully connected layer of 64 nodes used to process the operating condition vector; the data stream is a third fully connected layer of 64 nodes used to process the temperature curve vector; the 64-dimensional feature vector output by the two-stream network structure is concatenated into 128 dimensions, and then sequentially passes through fourth fully connected layers of 256 nodes and 128 nodes, with each fourth fully connected layer followed by a batch normalization layer and a LeakyReLU activation function, finally outputting a discriminant probability value through a Sigmoid activation function.

[0048] In one example of the present invention, in step S34, the generator uses a weighted combination loss function L. total From the counter-loss L adv and physical constraint loss L smooth Weighted according to the following formula: Among them, the resistance loss L adv The calculation formula is: Among them, physical constraint loss L smooth The calculation formula is: In the formula, λ is the hyperparameter; E represents the mathematical expectation; P Z (Z) represents the probability distribution of random noise; P data(C) represents the data distribution under actual operating conditions; D(.) represents the output probability of the discriminator; G(.) represents the output of the generator; N represents the length of the temperature sequence; T i This represents the predicted temperature value at the i-th elevation point; The loss function used by the discriminator in the temperature prediction model The expression is: In the formula, This represents a discriminator that outputs the probability of "true" under given operating conditions.

[0049] In one example of the present invention, generating the temperature protection curve for bridge cable vehicle fires in step S40 includes the following steps: S41: Abnormal Data Cleaning: Correction methods based on statistical distribution and neighboring point interpolation. For example, for the maximum temperature at each altitude point, if it exceeds the second highest temperature by more than 25%, it is corrected by interpolation. The temperature at this point is corrected by averaging the temperatures of the two nearest altitude points of the curve. Abnormal temperature points are identified and corrected for the distribution curves generated under all working conditions. For each standard altitude point, the distribution of temperature values ​​at this point under all working conditions is statistically analyzed. An anomaly discrimination threshold based on statistical quantiles is set. Temperature points exceeding the threshold are identified as abnormal values, and the correction values ​​are calculated by linear interpolation using the temperature values ​​of the distribution curve at adjacent altitude points to replace them. S42: Envelope Extraction: Based on the cleaned temperature distribution curve dataset, extract the outer envelope of all curves along the height direction; S43: Envelope Physical Correction: Apply a monotonic physical constraint to the extracted preliminary outer envelope to ensure that it satisfies the law that temperature decays with increasing altitude.

[0050] In one example of the present invention, in step S43, a monotonic physical constraint is applied to the extracted preliminary outer envelope to ensure that it satisfies the law of temperature decay with increasing altitude, specifically including: Starting from the bridge deck and proceeding upwards, check each point one by one. If the temperature value of the current elevation point is lower than the temperature value of the subsequent elevation points, then correct the temperature value of the abnormal point to the maximum value in the temperature sequence of the subsequent elevation points, thereby ensuring the physical rationality and conservatism of the final fortification curve.

[0051] In one example of the present invention, the generated report (e.g., a Word document) in step S40 includes: a project summary, a list of input operating parameters, a temperature protection curve, key safety indicator data, a risk level assessment, and specific graded protection recommendations.

[0052] In one example of the present invention, the deployment environment of the fire protection method is a local server cluster under the internal network of the organization. All model parameters, calculation data and knowledge base content are stored in local storage devices to ensure offline processing and privacy security throughout the data processing chain. The large language model is an open-source generative pre-trained model that has been fine-tuned with text and code instructions in the field of bridge engineering, and has professional bridge fire protection terminology understanding and logical reasoning capabilities.

[0053] In one example of the present invention, in step S40, the safety assessment logic for generating the report is as follows: the temperature protection curve is compared with the material critical temperature threshold, and the fire risk level is automatically determined based on the protection specification clauses stored in the local knowledge base, and a graded protection recommendation is given.

[0054] In one example of the present invention, step S50 is also included: a continuous learning mechanism: capable of recording user feedback and modifications to the generated report, and using this data to incrementally fine-tune the large language model scheduling and deep learning model to optimize the subsequent prediction accuracy and interactive intelligence of the system.

[0055] According to a second aspect of the present invention, a bridge cable fire protection system based on a localized large model includes: The human-computer interaction and parameter extraction module is configured to extract and confirm key working condition parameters of bridge cable vehicle fire scenarios through multiple rounds of human-computer interaction based on a large language model, and obtain a standardized working condition parameter dataset. The large language model scheduling and knowledge management module is configured to call the local knowledge base for parameter verification based on the working condition parameter dataset, and generate a deep learning model calling instruction for the temperature protection curve of bridge cable vehicle fire. The deep learning module is configured to load and train the corresponding temperature field prediction model based on the deep learning model call instruction, perform temperature field spatiotemporal evolution simulation calculation, and output temperature field prediction results. The structured report generation module is configured to integrate the working condition parameter dataset and the temperature field prediction results to generate a report containing parameter vectors, temperature protection curves for bridge cable vehicle fires, protection indicators, and protection recommendations.

[0056] This fire protection system represents a paradigm shift from "manually driven" to "intelligently automated." Traditional methods rely on engineers manually setting up massive amounts of operational conditions and performing tedious post-processing. The fire protection system, using a large language model as its intelligent scheduling hub, integrates multiple discrete steps—parameter analysis, operational condition combination, model invocation, result integration, and report generation—into a fully automated pipeline. Users only need to describe their fire protection requirements in natural language, and the system can automatically complete all the work that previously required weeks of manual labor. This reduces the efficiency of fire protection curve generation from days or even tens of days to minutes, completely resolving the core pain point of inefficiency in traditional methods.

[0057] This fire protection system employs a dual approach of "physical laws + data-driven" to significantly enhance the scientific rigor and reliability of the design curve. This system is not simply a data fit. On one hand, the deep learning model incorporates physical constraints during training, ensuring its predictions conform to fundamental thermodynamic principles. On the other hand, during the design curve generation phase, a multi-condition coverage strategy and conservative envelope extraction, along with stringent physical corrections (such as monotonicity constraints), ensure that the final design curve covers various potential risk scenarios while meeting engineering conservatism requirements. Its scientific rigor and reliability far surpass traditional results relying on single conditions or empirical formulas.

[0058] This fire protection system ensures the complete localization of core data and models, fundamentally eliminating the risk of sensitive information leakage. All processing, calculation, and storage of sensitive data involving bridge structural parameters, geographic information, etc., are completed within the organization's internal network of server clusters, completely eliminating the risk of data leakage that might occur due to the use of cloud services. This feature makes the fire protection system particularly suitable for key bridge engineering projects with extremely high information security requirements, providing a secure and reliable solution for the application of intelligent technologies in critical infrastructure.

[0059] This fire protection system lowers the professional threshold, empowering frontline designers to produce expert-level results. Through a natural language interface and fully automated workflow, the system encapsulates complex fire simulation and design curve generation technologies into an easy-to-use tool. Non-senior experts or frontline designers do not need in-depth knowledge of fire dynamics or numerical simulation to quickly generate professional fire protection schemes that meet regulatory requirements, significantly reducing the technical application threshold and facilitating the promotion and popularization of advanced design methods.

[0060] This fire protection system adopts a modular and standardized system architecture, possessing excellent maintainability and scalability. Each core module of the system is encapsulated using containerization technology and communicates through standard API interfaces. This highly cohesive and loosely coupled architecture design allows any component in the system (such as a new temperature prediction model or an updated design specification knowledge base) to be independently upgraded or replaced without requiring a complete system reconstruction. This ensures that the fire protection system can continuously absorb technological advancements, adapt to future changes in specifications and requirements, and possesses a long lifespan and application value.

[0061] In one example of the present invention, it further includes: a continuous learning mechanism module, configured to record user feedback and modifications to the generated report, and to use this data to incrementally fine-tune the large language model scheduling and deep learning module to optimize the subsequent prediction accuracy and interactive intelligence of the system.

[0062] In one example of the present invention, each module of the fire protection system is encapsulated using containerization technology, and data exchange and process scheduling are performed through application programming interfaces to achieve high cohesion and low coupling between modules, thereby ensuring the maintainability and scalability of the system.

[0063] In one example of the present invention, the safety assessment logic of the structured report generation module is as follows: the temperature protection curve is compared with the material critical temperature threshold, and the fire risk level is automatically determined based on the protection code clauses stored in the local knowledge base, and a graded protection recommendation is given.

[0064] In one example of the present invention, the deep learning module includes: a bridge terminology library, a technical parameter terminology and database, bridge fire protection standards, and physical properties of bridge cable materials.

[0065] It should be noted that the bridge cable fire protection system based on localized large model of the present invention can also perform any of the processing described in the previously described bridge cable fire protection method based on localized large model, and the specific details are not repeated here.

[0066] A specific case study on the automatic generation and report output of temperature protection curves for bridge cable vehicle fires based on a localized large model: This embodiment takes the main cable of a typical long-span cable-stayed bridge as the object. A localized large language model, knowledge base, and professional deep learning model library are deployed on a local server cluster within the organization's internal network to automatically generate and output structured reports of temperature protection curves for bridge cable-stayed vehicle fires. The deep learning model uses the conditional generative adversarial network temperature field prediction model from the aforementioned "Method and System for Generating Temperature-Height Curves for Bridge Cable-Stayed Vehicle Fires" as the professional model component in this embodiment. The specific steps are as follows: S10: Human-Computer Interaction and Operating Parameter Extraction: Users describe fire scenarios in natural language through the localized large language model interface, for example: "A vehicle fire has occurred on the bridge. The estimated fire source power is 30MW, the wind speed on the bridge is about 5m / s, the fire source is located in the emergency lane, about 2.5m away from the main cable. Please generate the main cable temperature protection curve and provide protection suggestions." The localized large language model performs semantic recognition on the input statement, extracting key operating condition parameter entities such as fire source power, bridge wind speed, and fire source distance. For missing parameters (such as the maximum distance s due to lane width, wind speed statistics, etc.), it initiates follow-up logic; for parameters that are ambiguous or exceed a reasonable range, it initiates confirmation and correction logic, and finally outputs a complete standardized operating condition parameter dataset. A standardized output example is: fire source power P=30MW (fixed value), fire source distance d=2.5m (constrained within [0,s]), bridge wind speed u=5m / s (unit unified in m / s), and provides the parameter source and confirmation record.

[0067] S20: Local Knowledge Base Verification and Model Call Instruction Generation: The system calls the local knowledge base for verification and completion based on the standardized working condition parameter dataset obtained in step S10. This includes verification of parameter unit and dimension consistency, value range verification, and calculation and constraint of the upper limit s of distance associated with bridge design geometry (e.g., determined by the distance from the emergency lane and adjacent lane to the cable edge). When inconsistencies are found, the system returns a verification prompt to the user and provides correction suggestions. After verification, the localized large language model generates a deep learning model call instruction. The model call instruction can be encapsulated in a structured format (e.g., JSON or function call format): including model identifier (temperature-height curve prediction model), input working condition parameter vector, output height sequence (0-35m, step size 1m, 36 points in total), batch calculation strategy (single working condition prediction or multi-working condition sampling prediction), and result feedback path, etc.

[0068] S30: Loading of Temperature Field Prediction Model and Batch Generation of Temperature Curves: After receiving the model call instruction, the system loads the corresponding temperature field prediction model weight file from the local professional model library; when needed, the model can be incrementally trained or retrained according to the instruction (e.g., updating the weights after adding simulation / experiment samples), and then the temperature field fast inference calculation is performed to output the temperature prediction values ​​of 36 standard height points under the described working condition, and generate temperature-height curves. In this embodiment, in order to obtain a defense curve with engineering conservatism, the system samples the fire source distance d in a uniform distribution within the range of [0, s] while keeping the fire source power P fixed, and samples the bridge surface wind speed u in a Weibull distribution to form multiple sets of working condition parameter combinations; for each set of working conditions, the generator is called to output a temperature-height curve, thereby forming a temperature distribution curve dataset.

[0069] S40: Fortification Curve Generation, Index Calculation, and Report Output: The system integrates the operating condition parameter dataset and temperature field prediction results to generate the fortification curve. First, abnormal data is cleaned (temperature distribution is statistically analyzed at each standard height point, significant anomalies are identified and corrected, and linear interpolation of nearby height points can be used for correction). Then, the outer envelope of the cleaned curve set is extracted along the height direction as the preliminary fortification curve. Finally, a monotonic physical constraint is applied to the preliminary fortification curve to ensure that the temperature decreases with increasing height, thus obtaining the final temperature fortification curve. Based on this, the system automatically calculates key fortification indicators (e.g., peak temperature near the bridge deck, height range where temperature exceeds the threshold, risk level, etc.) based on the material critical temperature thresholds and fortification specification clauses in the local knowledge base, and provides graded fortification recommendations (e.g., fireproof coating / covering height range, water spraying or heat insulation measures recommendations, operational monitoring recommendations, etc.).

[0070] Finally, the structured report generation module generates and writes the project summary, input condition parameter list, temperature protection curve, key safety index data, risk level assessment and protection recommendations into a Word document. This Word report is generated and stored in a local offline environment, avoiding the uploading of bridge structural parameters and design information to external networks, thereby meeting the requirements for engineering data security.

[0071] S50: Continuous learning mechanism: If a user revises the protection recommendations in the report (e.g., adjusts the recommended fireproof coating height or supplements on-site wind environment information), the system records the revision content and corresponding working conditions as feedback data; under the premise of satisfying permissions and security policies, it can be used to incrementally fine-tune the parameter extraction strategy of the localized large language model and the temperature prediction model in the professional model library, thereby continuously improving the interactive accuracy and prediction reliability of subsequent projects.

[0072] As can be seen from the above embodiments, the present invention can automatically complete the extraction and verification of working condition parameters, the invocation of temperature field prediction models, the generation of multi-working condition temperature curves, the envelope and physical correction of defense curves, and the output of structured reports in a localized deployment environment, using natural language interaction as the entry point. This significantly reduces the professional threshold of defense analysis and improves engineering efficiency.

[0073] The foregoing description, with reference to preferred embodiments, details an exemplary implementation of the bridge cable fire protection method and system based on a localized large model proposed in this 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 this invention, and various combinations can be made to the various technical features and structures proposed in this invention without exceeding the protection scope of this invention, which is determined by the appended claims.

Claims

1. A method for fire protection of bridge cables based on a localized large model, characterized in that, Includes the following steps: S10: Based on a large language model, key working condition parameters of bridge cable vehicle fire scenarios are extracted and confirmed through multiple rounds of human-computer interaction to obtain a standardized working condition parameter dataset. S20: Based on the aforementioned working condition parameter dataset, call the local knowledge base to verify the parameters and generate a deep learning model call instruction for the temperature protection curve of bridge cable vehicle fire. S30: Based on the deep learning model call instruction, load and obtain the corresponding temperature field prediction model and train it, perform temperature field spatiotemporal evolution simulation calculation, and output temperature field prediction results; S40: Integrate the working condition parameter dataset and the temperature field prediction results to generate a report containing parameter vectors, temperature protection curves for bridge cable vehicle fires, protection indicators, and protection recommendations.

2. The bridge cable fire protection method based on a localized large model according to claim 1, characterized in that, In step S10, key operating condition parameters for a bridge cable vehicle fire scenario are extracted and confirmed through multiple rounds of human-computer interaction based on a large language model, and a standardized operating condition parameter dataset is obtained. The specific process includes: Leveraging the semantic understanding capabilities of large language models, we can identify fire scene parameter entities that are implicit or explicit in user input. For key condition parameters that are not clearly defined, the inquiry logic is initiated; for condition parameters that are ambiguous or exceed a reasonable range, the confirmation and correction logic is initiated, and finally, a complete and error-free standard chemical condition parameter set is output.

3. The bridge cable fire protection method based on a localized large model according to claim 2, characterized in that, The standardized operating condition parameter set includes: fire source power, fire source distance, and bridge deck wind speed, wherein the fire source power is a fixed value; the fire source distance parameter set is obtained by sampling in a uniform distribution in [0,s], where s is the maximum fire source distance, and the maximum fire source distance is the distance between the adjacent lane of the emergency lane and the cable; the bridge deck wind speed is sampled according to the Weibull distribution. This involves arranging and combining three operating parameters to achieve comprehensive coverage of different fire scenarios through this hybrid sampling strategy.

4. The bridge cable fire protection method based on a localized large model according to claim 1, characterized in that, In step S30, the corresponding temperature field prediction model is loaded and trained, specifically including the following steps: S31: Construct a bridge fire simulation dataset: Generate bridge height-temperature data under different standardized working condition parameter combinations using simulation software; S32: Data preprocessing and standardization: The non-uniform height-temperature distribution in the simulation data and experimental data is uniformly mapped to the standard height sequence through an interpolation algorithm. The operating parameters and temperature values ​​are normalized to form an operating parameter set, which is then divided into a training set, a validation set, and a test set. S33: Constructing a Conditional Generative Adversarial Network Model: A conditional generative adversarial network model is constructed by a generator and a discriminator. The generator is configured to generate predicted temperature curves based on noise vectors and normalized working condition parameters. The discriminator is configured to distinguish between the generated temperature curves and the actual temperature curves and output a discrimination probability value. S34: Define the weighted loss function: The generator adopts a weighted combined loss function that includes adversarial loss and physical constraints. The conditional generative adversarial network model is trained through the training set. The adversarial loss ensures that the generated distribution approximates the real distribution, and the physical constraints ensure that the generated curve is smooth and meets the corresponding physical laws. S35: Model Training and Optimization: Based on gradient descent optimization algorithms and their hyperparameter configuration strategies, a conditional generative adversarial network model is trained using a training set with an alternating iterative update strategy of discriminator and generator. During training, the training effect is monitored using a validation set to adjust the training parameters. The hyperparameter learning efficiency is 2×10⁻⁶. -4 The minimum number of training rounds is 500 to ensure prediction accuracy, and the maximum is 1000 to prevent overfitting. S36: Model Validation and Deployment: Evaluate the performance of the conditional generative adversarial network model on the test set, using the coefficient of determination R0. 2 The conditional generative adversarial network model is evaluated using the mean absolute error (MAE), where the coefficient of determination R0 is the mean absolute error (MAE). 2 >0.9, with a mean absolute error (MAE) <50℃. After successful verification, the weight file of the fully trained conditional generative adversarial network model will be deployed to a local professional model library for use by large language models.

5. The bridge cable fire protection method based on a localized large model according to claim 4, characterized in that, In step S33, the generator includes an input layer, three first fully connected hidden layers, and a first output layer. The input layer is a concatenated vector of a received noise vector and normalized working condition parameters, used for data input and feature initialization. The three first fully connected hidden layers are connected sequentially, with each first fully connected hidden layer followed by a batch normalization layer and a LeakyReLU activation function, used for feature extraction and nonlinear transformation at different levels, respectively. The first output layer includes a first fully connected layer and a Tanh activation function, used to output temperature prediction values ​​for multiple standard altitude points. The discriminator is a two-stream network structure, including a conditional stream, a data stream, a splicing layer, two second fully connected hidden layers, and a second output layer. The conditional stream includes a second fully connected layer and a LeakyReLU activation function, which are used to process the operating condition vector and extract conditional features. The data stream includes a third fully connected layer and a LeakyReLU activation function, which are used to process the temperature curve vector and extract data features. The concatenation layer concatenates the feature vectors output from the conditional stream and the data stream into a fused feature vector; the two second fully connected hidden layers are connected sequentially, with each second fully connected hidden layer followed by a LeakyReLU activation function, which are used for feature fusion and discriminant analysis, respectively; the second output layer includes a fourth fully connected layer and a Sigmoid activation function, which are used to output the discriminant probability value.

6. The bridge cable fire protection method based on a localized large model according to claim 1, characterized in that, In step S34, the generator uses a weighted combination loss function L. total From the counter-loss L adv and physical constraint loss L smooth Weighted according to the following formula: Among them, the resistance loss L adv The calculation formula is: Among them, physical constraint loss L smooth The calculation formula is: In the formula, λ is the hyperparameter; E represents the mathematical expectation; P Z (Z) represents the probability distribution of random noise; P data (C) represents the data distribution under actual operating conditions; D(.) represents the output probability of the discriminator; G(.) represents the output of the generator; N represents the length of the temperature sequence; T i This represents the predicted temperature value at the i-th elevation point; The loss function used by the discriminator The expression is: In the formula, This represents a discriminator that outputs the probability of "true" under given operating conditions.

7. The bridge cable fire protection method based on a localized large model according to claim 1, characterized in that, In step S40, generating the temperature protection curve for bridge cable vehicle fires includes the following steps: S41: Abnormal data cleaning: Based on the correction method of statistical distribution and neighboring point interpolation, abnormal temperature points are identified and corrected for the distribution curves generated under all working conditions; for each standard height point, the distribution of temperature values ​​of all working conditions at that point is statistically analyzed, an anomaly discrimination threshold based on statistical quantiles is set, temperature points exceeding the threshold are identified as abnormal values, and the correction values ​​are calculated by linear interpolation using the temperature values ​​of the distribution curve at adjacent height points where the abnormal value is located and then replaced. S42: Envelope Extraction: Based on the cleaned temperature distribution curve dataset, extract the outer envelope of all curves along the height direction; S43: Envelope Physical Correction: Apply a monotonic physical constraint to the extracted preliminary outer envelope to ensure that it satisfies the law that temperature decays with increasing altitude.

8. The bridge cable fire protection method based on a localized large model according to claim 1, characterized in that, In step S43, a monotonic physical constraint is applied to the extracted preliminary outer envelope to ensure that it satisfies the law of temperature decay with increasing altitude, specifically including: Starting from the bridge deck and proceeding upwards, check each point one by one. If the temperature value of the current elevation point is lower than the temperature value of the subsequent elevation points, then correct the temperature value of the abnormal point to the maximum value in the temperature sequence of the subsequent elevation points.

9. The bridge cable fire protection method based on a localized large model according to claim 1, characterized in that, In step S40, the generated report includes: a project summary, a list of input operating parameters, a temperature protection curve, key safety indicator data, a risk level assessment, and specific graded protection recommendations.

10. A bridge cable fire protection system based on a localized large model, characterized in that, include: The human-computer interaction and parameter extraction module is configured to extract and confirm key working condition parameters of bridge cable vehicle fire scenarios through multiple rounds of human-computer interaction based on a large language model, and obtain a standardized working condition parameter dataset. The large language model scheduling and knowledge management module is configured to call the local knowledge base for parameter verification based on the working condition parameter dataset, and generate a deep learning model calling instruction for the temperature protection curve of bridge cable vehicle fire. The deep learning module is configured to load and train the corresponding temperature field prediction model based on the deep learning model call instruction, perform temperature field spatiotemporal evolution simulation calculation, and output temperature field prediction results. The structured report generation module is configured to integrate the working condition parameter dataset and the temperature field prediction results to generate a report containing parameter vectors, temperature protection curves for bridge cable vehicle fires, protection indicators, and protection recommendations.

Images