Civil aviation fire-fighting emergency rescue device and use method thereof

Abstract

The invention discloses a civil aviation fire-fighting emergency rescue device and a use method thereof. The civil aviation fire-fighting emergency rescue device comprises a front-end user interface module, a data monitoring and preprocessing module, a risk assessment and decision module, a resource scheduling and execution module and a system management and detection module. The front-end user interface module is responsible for providing a graphical interface and supporting interaction between a user and the system, the data monitoring and preprocessing module is responsible for collecting data from various data sources and carrying out cleaning, denoising and preprocessing, and the risk assessment and decision making module is used for carrying out risk assessment and making a decision making strategy based on the preprocessed data and a real-time monitoring result. The resource scheduling and execution module is responsible for resource scheduling and execution, the system management and detection module is responsible for overall management and monitoring of the system, the possibility of fire occurrence is rapidly recognized through the time sequence analysis capability of LSTM and real-time anomaly detection of i Forest, comprehensive consideration can be carried out according to various factors by means of a Bayesian network and a decision tree, and the risk of fire disaster occurrence is reduced. And generating a risk assessment report.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency management and intelligent dispatching, and in particular to a civil aviation fire emergency rescue device and a use method thereof. Background Art

[0002] With the rapid development of the aviation transportation industry, airports are facing increasing fire safety challenges, especially in ensuring the safety of personnel and quickly resuming operations in emergency situations. In recent years, with the advancement of information technology, especially the application of big data analysis, machine learning and artificial intelligence technology, civil aviation fire emergency rescue equipment has been significantly improved. The application of these technologies has improved data processing capabilities, enhanced the accuracy of prediction and risk assessment, and improved the efficiency and flexibility of resource scheduling.

[0003] Some civil aviation fire emergency rescue devices in the existing technology have limited data processing capabilities: traditional systems have difficulty processing large-scale real-time data streams and cannot capture and analyze key information in a timely manner. Low accuracy of prediction and risk assessment: the lack of advanced prediction and risk assessment tools makes it difficult for decision makers to make quick and accurate decisions. Inefficient resource scheduling: information asymmetry often occurs in the resource scheduling process, resulting in waste of resources and scheduling delays. Weak emergency response capabilities: in the face of complex emergencies, traditional emergency response mechanisms are often slow to respond and lack flexibility and adaptability.

[0004] Therefore, a civil aviation fire emergency rescue device and a use method thereof are proposed. Summary of the invention

[0005] The purpose of the present invention is to solve the problems existing in the prior art, such as limited data processing capabilities: traditional systems have difficulty processing large-scale real-time data streams and are unable to capture and analyze key information in a timely manner. Low accuracy of prediction and risk assessment: the lack of advanced prediction and risk assessment tools makes it difficult for decision makers to make quick and accurate decisions. Inefficient resource scheduling: information asymmetry often occurs in the resource scheduling process, resulting in waste of resources and scheduling delays. Weak emergency response capabilities: in the face of complex emergencies, traditional emergency response mechanisms are often slow to respond and lack flexibility and adaptability. A civil aviation fire emergency rescue device and its use method are proposed.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A civil aviation fire emergency rescue device, comprising: a front-end user interface module, a data monitoring and preprocessing module, a risk assessment and decision-making module, a resource scheduling and execution module, and a system management and detection module;

[0008] The front-end user interface module is responsible for providing a graphical interface to support the interaction between the user and the system. The data monitoring and preprocessing module is responsible for collecting data from various data sources and performing cleaning, denoising and preprocessing. The risk assessment and decision-making module performs risk assessment and formulates decision-making strategies based on preprocessed data and real-time monitoring results. The resource scheduling and execution module is responsible for the scheduling and execution of resources. The system management and detection module is responsible for the overall management and monitoring of the system.

[0009] The front-end user interface module is the entrance for users to interact with the system. Users input operation instructions and query conditions through the interface and receive feedback from the system. The data monitoring and preprocessing module is responsible for collecting data from the data source and preprocessing it, providing high-quality data support for the risk assessment and decision-making module. The risk assessment and decision-making module performs risk assessment and decision-making based on preprocessed data and real-time data, and passes the results to the resource scheduling and execution module. The resource scheduling and execution module generates and executes scheduling instructions according to the decision-making strategy, and provides real-time feedback on the execution status. The system management and monitoring module is responsible for monitoring and management of the entire system.

[0010] The above technical solution further includes:

[0011] Preferably, the front-end user interface module includes a user interaction unit, an information display unit, and a query and report unit. The user interaction unit is used to provide a graphical interface for user login, authority verification, and operation instruction input. The information display unit is used to display real-time monitoring data, risk assessment results, and scheduling instruction execution status. The query and report unit is used to support historical data query and statistical report generation.

[0012] Preferably, the data monitoring and preprocessing module includes a data acquisition unit, a data cleaning unit, a data denoising unit, an LSTM prediction unit, and an iForest implementation anomaly detection unit. The data acquisition unit is responsible for collecting data from a data source in real time. The data cleaning unit is used to remove duplicate data, fill missing values, process abnormal values, etc. The data denoising unit is used to apply filtering technology to reduce data noise. The LSTM prediction unit analyzes historical time series data, learns the dependencies between data, and predicts indicators or trends at future time points based on this. The iForest real-time anomaly detection unit is used for real-time anomaly detection.

[0013] Preferably, the risk assessment and decision-making module includes a risk assessment data preparation unit, a Bayesian network and decision tree combination unit, a decision strategy formulation unit, and a decision optimization unit. The risk assessment data preparation unit is used to integrate preprocessed data and historical data to construct a risk assessment data set. The Bayesian network and decision tree combination unit combines the prediction results of LSTM, the anomaly detection results of iForest and other data to perform risk assessment on emergency events and generate a risk assessment report. The decision strategy formulation unit is used to formulate a scheduling decision strategy based on the risk assessment results and preset emergency plans. The decision optimization unit is used to optimize and adjust the decision strategy by considering resource availability and response time factors.

[0014] Preferably, the resource scheduling and execution module includes a resource information management unit, a dynamic path planning unit, a scheduling instruction generation unit, an execution tracking and feedback unit, and a dynamic adjustment and optimization unit. The resource information management unit is used to maintain real-time information of rescue resources and personnel. The dynamic path planning unit is used to combine real-time traffic information, rescue needs and other factors and adopt a path planning algorithm to perform optimal path planning. The scheduling instruction generation unit is used to generate specific scheduling instructions based on decision-making strategies and path planning results. The execution tracking and feedback unit is used to track the execution status of the rescue team in real time and collect feedback information, such as arrival time and task completion status. The dynamic adjustment and optimization unit is used to dynamically adjust scheduling strategies and resource allocation based on execution status feedback and real-time data changes.

[0015] Preferably, the system management and monitoring module includes a user authority management unit, a system log management unit, a performance monitoring and optimization unit, and an external system integration unit. The user authority management unit is used to manage user accounts, roles and permissions; the system log management unit is used to record system logs for auditing and troubleshooting; the performance monitoring and optimization unit is used to monitor system performance indicators, perform performance tuning and fault warning; and the external system integration unit is responsible for the integration and interaction of external systems.

[0016] A method for using a civil aviation fire emergency rescue device comprises the following steps:

[0017] Further, S1: User interface initialization: Login and authority verification: Login to the system through the user interface of the dedicated fire emergency rescue device, and the system verifies according to the user role and authority to ensure that only authorized users can access the system; Interface navigation and layout: The user interface adopts a fire theme design, provides a clear navigation menu and layout, and specially marks the function modules related to fire emergency, so that firefighters can locate and operate the required functions;

[0018] Further, S2: Data collection and preprocessing: Data collection: Users set the fire data collection parameters for civil aviation airports and their surroundings through the user interface. The system monitors and collects these data in real time to provide a basis for subsequent emergency response; Preprocessing: During the preprocessing process, the system cleans, integrates and formats the collected fire data, removes noise data, and extracts key features. Users can view the progress of data cleaning and integration in real time on the interface, as well as an overview of the preprocessed fire data to ensure the accuracy and timeliness of the data;

[0019] Further, S3: Model construction and risk assessment: Model construction: The user interface displays the construction progress and results of the LSTM fire prediction model and the iForest fire anomaly detection model for civil aviation firefighting scenarios in real time. Risk assessment: Users view the results of the comprehensive fire risk assessment through the interface, and the system provides charts and reports to help firefighters understand the fire situation and formulate corresponding emergency response plans;

[0020] Further, S4: Emergency response decision making: Intelligent decision support: The user interface provides an intelligent decision support input interface for civil aviation fire emergency rescue, and users can adjust decision parameters based on real-time fire information and prediction results. The system uses algorithms and models to provide firefighters with a variety of emergency response plans and evaluate their feasibility and effectiveness; Decision making and display: During the decision-making process, users can view the simulated rescue effects under different parameter combinations on the interface, and the decision results are displayed on the user interface;

[0021] Further, S5: Execution and Feedback: Execution monitoring: Users monitor the execution of civil aviation fire emergency rescue in real time through the user interface. The system provides real-time data and video feedback to ensure that firefighters can fully grasp the dynamics of the rescue scene; Feedback adjustment: The user interface provides a feedback mechanism, allowing users to adjust the emergency response plan according to actual conditions. Firefighters can quickly make decisions and adjustments based on on-site feedback, and view the effects of the adjustments in real time;

[0022] Further, S6: Report output and experience summary: Report output: After the emergency response is completed, the user interface generates a detailed civil aviation fire emergency rescue report. Users can download or print the report as a basis for subsequent analysis and improvement; Experience summary: The user interface provides an input interface for experience summary, encouraging firefighters to record the experience and lessons of this emergency response. These experience summaries will be included in the system database for reference and sharing of similar events in the future, and continuously improve the capabilities and level of civil aviation fire emergency rescue.

[0023] Preferably, in S3, the specific steps of using the LSTM prediction model are:

[0024] Further, data collection and preprocessing: collect time series data X = {x 1 ,x 2 ,...,x T}, where x T represents the observation value at the Tth time point;

[0025] Furthermore, the collected data is cleaned and preprocessed, X'=preprocess(X), where preprocess represents the data cleaning and preprocessing function;

[0026] Further, data set division: the preprocessed data is divided into training set X train , validation set X val and the test set X test The training set is used for model training, the validation set is used for model parameter adjustment, and the test set is used to evaluate model performance;

[0027] Furthermore, LSTM model construction: designing the LSTM network structure, including the input layer, multiple LSTM layers, and a fully connected layer;

[0028] Furthermore, the number of neurons, learning rate, and optimizer of the LSTM layer are determined;

[0029] Furthermore, set the LSTM model parameters: number of network layers: L, number of LSTM units per layer: N l (l=1,2,...,L), learning rate η, optimizer ο, build LSTM model: LSTM=LSTM(L,N l ,η,ο);

[0030] Further, model training: Use the training set data to train the LSTM model: θ = train(LSTM, X train ,y train ), where θ represents the model parameters, y train is the label or target value corresponding to the training set. The network weights are adjusted through the back-propagation algorithm so that the model can learn the time dependency and nonlinear relationship in the data.

[0031] Furthermore, during the training process, the validation set can be used to adjust the model parameters to find the optimal hyperparameter combination;

[0032] Further, model evaluation: Use the test set to evaluate the trained LSTM model: performance = evaluate(LSTM, θ, X test ,y test ), calculate the prediction accuracy and error rate performance indicators;

[0033] Furthermore, if the model performance does not meet the requirements, you can return to step 3 to readjust the model structure or hyperparameters, where performance represents the performance index;

[0034] Further, real-time prediction: the latest time series data X new Input into the trained LSTM model: Generate forecast results for a period of time in the future, Indicates the prediction result;

[0035] Furthermore, the prediction results are passed to the Bayesian network and decision tree combination unit for further processing.

[0036] Preferably, in S3, the specific steps of using the iForest real-time anomaly detection are:

[0037] Further, data integration: LSTM prediction results Real-time observation dataX real Integrate with other relevant data as the input dataset D for anomaly detection;

[0038] Further, iForest model construction: set iForest parameters, such as the number of trees n trees , subsample size ψ;

[0039] Further, the integrated data set is used to construct multiple isolation trees iForest=iForest(n trees ,ψ), each tree is built independently, and each tree randomly selects part of the data for training;

[0040] Further, anomaly detection: For each point d in the input dataset D i , calculate the path length h(d i ,tree j ), where j = 1, 2, ..., n trees , calculate the average path length Serves as a benchmark for the entire dataset;

[0041] Furthermore, the anomaly score is calculated based on the path length Where E[] represents the expectation, c(ψ) is the estimated path length based on the subsample size ψ, and the anomaly score is inversely proportional to the average path length, that is, the shorter the path, the higher the anomaly score;

[0042] Furthermore, a threshold τ is set and the anomaly score is compared with the preset threshold. If s(d i )>τ, the score is higher than the threshold, then the point d i It is an outlier.

[0043] Furthermore, result interpretation and response: analyze the causes of anomalies and trigger corresponding emergency responses based on the anomaly detection results.

[0044] Preferably, in S3, the specific steps of obtaining the risk assessment result using the decision tree and the Bayesian network are:

[0045] Further, data collection: collect all possible data relevant to the prediction or decision;

[0046] Further, feature selection: extract features useful for Bayesian network and decision tree modeling from the collected data. These features may include the average value of LSTM prediction results for a specific time period, the rate of change, and the abnormal indicators detected by iForest;

[0047] Furthermore, data preprocessing: cleaning the features, such as dealing with missing values ​​and outliers, and standardizing or normalizing to ensure the consistency and comparability of the data;

[0048] Further, Bayesian network construction: Define the network structure: Based on expert knowledge or data exploration results, define the nodes in the Bayesian network and the dependencies between them. Nodes can include LSTM prediction results, external factors, and system status;

[0049] Further, learning parameters: using historical data or expert knowledge to estimate the conditional probability distribution between nodes in the network;

[0050] Furthermore, model validation: a portion of the validation set is used to verify the accuracy and effectiveness of the Bayesian network model;

[0051] Further, decision tree construction: Feature selection: Select the features that have the most influence on the decision process as the input of the decision tree;

[0052] Further, build a decision tree: use the training data set to build a decision tree model to optimize the performance and interpretability of the tree;

[0053] Furthermore, model evaluation: Use the test set to evaluate the performance of the decision tree model, including accuracy and recall indicators;

[0054] Furthermore, Bayesian networks are combined with decision trees: Integrated model: The output probability distribution or specific prediction of the Bayesian network is used as one of the input features of the decision tree. This allows the decision tree to consider the uncertainty information and probability predictions provided by the Bayesian network when making decisions;

[0055] Further, optimization and adjustment: adjust the parameters of the Bayesian network and decision tree, as well as the integration method between them according to actual needs to achieve the best prediction or decision-making effect.

[0056] Furthermore, real-time decision making: when new data LSTM prediction results or real-time observation data arrive, they are first processed by the Bayesian network, and then the processing results are input into the decision tree for real-time decision making.

[0057] Further, result interpretation and action: interpret the results of the decision tree output and take corresponding actions or measures based on the results. At the same time, monitor and evaluate the effect of the decision in order to carry out subsequent model optimization and adjustment.

[0058] Preferably, the historical time series data is analyzed through LSTM (Long Short-Term Memory Network), the dependencies between the data are learned, and the indicators or trends at future time points are predicted based on this to obtain powerful data processing capabilities.

[0059] Preferably, an iForest (isolation forest) algorithm is used for real-time anomaly detection to discover abnormal behaviors.

[0060] Preferably, the prediction results of LSTM, the anomaly detection results of iForest and other data are combined to use the Bayesian network and decision tree model to conduct risk assessment on emergency events and generate a risk assessment report.

[0061] Preferably, a scheduling decision strategy is formulated based on the risk assessment results and preset emergency plans, and the decision strategy is optimized and adjusted by considering factors such as resource availability and response time.

[0062] The present invention has the following beneficial effects:

[0063] 1. In the present invention, potential safety threats, such as smoke, abnormal temperature rise or other fire precursor signals, are quickly identified through real-time monitoring and anomaly detection technology. Through the time series analysis capability of LSTM and the real-time anomaly detection function of iForest, the possibility of fire can be quickly identified and the early warning mechanism can be automatically triggered. This greatly reduces the time difference between discovering anomalies and initiating emergency response, and buys valuable initial disposal time for firefighters.

[0064] 2. In the present invention, by means of the analysis method of Bayesian network and decision tree, a comprehensive consideration can be made based on multiple factors, such as the specific situation of the fire scene, wind direction changes, building structural characteristics, etc., so as to generate a more accurate risk assessment report. These assessment reports can not only help decision makers understand the possibility of the development of the situation, but also guide them to formulate the most appropriate emergency plan. For example, under certain conditions, the best evacuation route is selected or the location of the fire source to be extinguished first is determined, which effectively avoids the losses that may be caused by blind actions. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1This is a structural schematic diagram of a civil aviation fire emergency rescue device proposed by the present invention;

[0066] Figure 2 The present invention is a schematic structural diagram of a method for using a civil aviation fire emergency rescue device. DETAILED DESCRIPTION

[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] Embodiment 1

[0069] like Figure 1-Figure 2 As shown, a civil aviation fire emergency rescue device proposed by the present invention includes: a front-end user interface module, a data monitoring and preprocessing module, a risk assessment and decision-making module, a resource scheduling and execution module, and a system management and detection module;

[0070] The front-end user interface module is responsible for providing a graphical interface to support the interaction between users and the system. The data monitoring and preprocessing module is responsible for collecting data from various data sources and performing cleaning, denoising and preprocessing. The risk assessment and decision-making module conducts risk assessment and formulates decision-making strategies based on preprocessed data and real-time monitoring results. The resource scheduling and execution module is responsible for the scheduling and execution of resources. The system management and detection module is responsible for the overall management and monitoring of the system.

[0071] The front-end user interface module is the entrance for users to interact with the system. Users input operating instructions and query conditions through the interface and receive feedback from the system. The data monitoring and preprocessing module is responsible for collecting and preprocessing data from the data source, and providing data support for the risk assessment and decision-making module. The risk assessment and decision-making module performs risk assessment and decision-making based on preprocessed data and real-time data, and passes the results to the resource scheduling and execution module. The resource scheduling and execution module generates and executes scheduling instructions according to the decision-making strategy, and provides real-time feedback on the execution status. The system management and monitoring module is responsible for monitoring and management of the entire system.

[0072] In one embodiment, the front-end user interface module includes a user interaction unit, an information display unit, and a query and report unit. The user interaction unit is used to provide a graphical interface for user login, authority verification, and operation instruction input. The information display unit is used to display real-time monitoring data, risk assessment results, and scheduling instruction execution status. The query and report unit is used to support historical data query and statistical report generation.

[0073] In one embodiment, the data monitoring and preprocessing module includes a data acquisition unit, a data cleaning unit, a data denoising unit, an LSTM prediction unit, and an iForest implementation anomaly detection unit. The data acquisition unit is responsible for collecting data from a data source in real time. The data cleaning unit is used to remove duplicate data, fill missing values, process abnormal values, etc. The data denoising unit is used to apply filtering technology to reduce data noise. The LSTM prediction unit analyzes historical time series data, learns the dependencies between data, and predicts indicators or trends at future time points based on this. The iForest real-time anomaly detection unit is used for real-time anomaly detection.

[0074] In one embodiment, the risk assessment and decision-making module includes a risk assessment data preparation unit, a Bayesian network and decision tree combination unit, a decision strategy formulation unit, and a decision optimization unit. The risk assessment data preparation unit is used to integrate preprocessed data and historical data to construct a risk assessment data set. The Bayesian network and decision tree combination unit combines the prediction results of LSTM, the anomaly detection results of iForest and other data to perform risk assessment on emergency events and generate a risk assessment report. The decision strategy formulation unit is used to formulate a scheduling decision strategy based on the risk assessment results and preset emergency plans. The decision optimization unit is used to consider resource availability and response time factors to optimize and adjust the decision strategy.

[0075] In one embodiment, the resource scheduling and execution module includes a resource information management unit, a dynamic path planning unit, a scheduling instruction generation unit, an execution tracking and feedback unit, and a dynamic adjustment and optimization unit. The resource information management unit is used to maintain real-time information of rescue resources and personnel. The dynamic path planning unit is used to combine real-time traffic information, rescue needs and other factors, and adopt a path planning algorithm to perform optimal path planning. The scheduling instruction generation unit is used to generate specific scheduling instructions based on decision-making strategies and path planning results. The execution tracking and feedback unit is used to track the execution status of the rescue team in real time and collect feedback information, such as arrival time and task completion status. The dynamic adjustment and optimization unit is used to dynamically adjust scheduling strategies and resource allocation based on execution status feedback and real-time data changes.

[0076] In one embodiment, the system management and monitoring module includes a user authority management unit, a system log management unit, a performance monitoring and optimization unit, and an external system integration unit. The user authority management unit is used to manage user accounts, roles and permissions. The system log management unit is used to record system logs for auditing and troubleshooting. The performance monitoring and optimization unit is used to monitor system performance indicators, perform performance tuning and fault warning, and the external system integration unit is responsible for the integration and interaction of external systems.

[0077] In one embodiment, a method for using a civil aviation fire emergency rescue device includes the following steps:

[0078] S1: User interface initialization: Login and authority verification: Log in to the system through the user interface of the dedicated fire emergency rescue device. The system verifies the user role and authority to ensure that only authorized users can access the system; Interface navigation and layout: The user interface adopts a fire theme design, provides a clear navigation menu and layout, and specially marks the function modules related to fire emergency, so that firefighters can locate and operate the required functions;

[0079] S2: Data collection and preprocessing: Data collection: Users set fire data collection parameters for civil aviation airports and their surrounding environments through the user interface. The system monitors and collects these data in real time to provide a basis for subsequent emergency response; Preprocessing: During the preprocessing process, the system cleans, integrates and formats the collected fire data, removes noise data, and extracts key features. Users can view the progress of data cleaning and integration in real time on the interface, as well as an overview of the preprocessed fire data to ensure the accuracy and timeliness of the data;

[0080] S3: Model construction and risk assessment: Model construction: The user interface displays the construction progress and results of the LSTM fire prediction model and the iForest fire anomaly detection model for civil aviation firefighting scenarios in real time. Risk assessment: Users view the results of the comprehensive fire risk assessment through the interface. The system provides charts and reports to help firefighters understand the fire situation and formulate corresponding emergency response plans;

[0081] S4: Emergency response decision making: Intelligent decision support: The user interface provides an intelligent decision support input interface for civil aviation fire emergency rescue. Users can adjust decision parameters based on real-time fire information and prediction results. The system uses algorithms and models to provide firefighters with a variety of emergency response plans and evaluate their feasibility and effectiveness; Decision making and display: During the decision-making process, users can view the simulated rescue effects under different parameter combinations on the interface, and the decision results are displayed on the user interface;

[0082] S5: Execution and Feedback: Execution monitoring: Users monitor the execution of civil aviation fire emergency rescue in real time through the user interface. The system provides real-time data and video feedback to ensure that firefighters can fully grasp the dynamics of the rescue scene; Feedback adjustment: The user interface provides a feedback mechanism, allowing users to adjust the emergency response plan according to actual conditions. Firefighters can quickly make decisions and adjustments based on on-site feedback, and view the effects of the adjustments in real time;

[0083] S6: Report output and experience summary: Report output: After the emergency response is completed, the user interface generates a detailed civil aviation fire emergency rescue report. Users can download or print the report as a basis for subsequent analysis and improvement; Experience summary: The user interface provides an input interface for experience summary, encouraging firefighters to record the experience and lessons of this emergency response. These experience summaries will be included in the system database for reference and sharing of similar incidents in the future, and continuously improve the capabilities and level of civil aviation fire emergency rescue.

[0084] In one embodiment, in S3, the specific steps of using the LSTM prediction model are:

[0085] Data collection and preprocessing: Time series data X = {x 1 ,x 2 ,...,x T}, where x T represents the observation value at the Tth time point;

[0086] Clean and preprocess the collected data, X'=preprocess(X), where preprocess represents the data cleaning and preprocessing function;

[0087] Dataset division: Divide the preprocessed data into training set X train , validation set X val and the test set X test The training set is used for model training, the validation set is used for model parameter adjustment, and the test set is used to evaluate model performance;

[0088] LSTM model construction: Design the LSTM network structure, including the input layer, multiple LSTM layers, and the fully connected layer;

[0089] Determine the number of neurons, learning rate, and optimizer of the LSTM layer;

[0090] Set LSTM model parameters: number of network layers: L, number of LSTM units per layer: N l (l=1,2,...,L), learning rate η, optimizer ο, build LSTM model: LSTM=LSTM(L,N l ,η,ο);

[0091] Model training: Use the training set data to train the LSTM model: θ = train(LSTM, X train ,y train ), where θ represents the model parameters, y train is the label or target value corresponding to the training set. The network weights are adjusted through the back-propagation algorithm so that the model can learn the time dependency and nonlinear relationship in the data.

[0092] During the training process, the validation set can be used to adjust the model parameters to find the optimal hyperparameter combination;

[0093] Model evaluation: Use the test set to evaluate the trained LSTM model: performance = evaluate(LSTM, θ, X test ,y test ), calculate the prediction accuracy and error rate performance indicators;

[0094] If the model performance does not meet the requirements, you can return to step 3 to readjust the model structure or hyperparameters, where performance represents the performance index;

[0095] Real-time prediction: The latest time series data X new Input into the trained LSTM model: Generate forecast results for a period of time in the future, Indicates the prediction result;

[0096] The prediction results are passed to the Bayesian network and decision tree combination unit for further processing.

[0097] In one embodiment, in S3, the specific steps of using iForest real-time anomaly detection are:

[0098] Data integration: LSTM prediction results Real-time observation dataX real Integrate with other relevant data as the input dataset D for anomaly detection;

[0099] iForest model construction: set iForest parameters, such as the number of trees n trees , subsample size ψ;

[0100] Use the integrated data set to construct multiple isolation trees iForest = iForest(n trees ,ψ), each tree is built independently, and each tree randomly selects part of the data for training;

[0101] Anomaly detection: For each point d in the input dataset D i , calculate the path length h(d i ,tree j ), where j = 1, 2, ..., n trees , calculate the average path length Serves as a benchmark for the entire dataset;

[0102] Calculate anomaly score based on path length Where E[] represents the expectation, c(ψ) is the estimated path length based on the subsample size ψ, and the anomaly score is inversely proportional to the average path length, that is, the shorter the path, the higher the anomaly score;

[0103] Set a threshold τ and compare the anomaly score with the preset threshold. If s(d i )>τ, the score is higher than the threshold, then the point d i It is an outlier.

[0104] Result interpretation and response: Analyze the causes of abnormal points and trigger corresponding emergency responses based on the abnormal detection results.

[0105] In one embodiment, in S3, the specific steps of obtaining the risk assessment result using the decision tree and the Bayesian network are:

[0106] Data collection: Gather all possible data relevant to the prediction or decision;

[0107] Feature selection: Extract features useful for Bayesian network and decision tree modeling from the collected data. These features may include the average value of LSTM prediction results for a specific time period, the rate of change, and the abnormal indicators detected by iForest;

[0108] Data preprocessing: Clean the features, such as dealing with missing values ​​and outliers, and standardizing or normalizing them to ensure the consistency and comparability of the data;

[0109] Bayesian network construction: Define the network structure: Based on expert knowledge or data exploration results, define the nodes in the Bayesian network and the dependencies between them. Nodes can include LSTM prediction results, external factors, and system status;

[0110] Learning parameters: using historical data or expert knowledge to estimate the conditional probability distribution between nodes in the network;

[0111] Model validation: Use a portion of the validation set to verify the accuracy and effectiveness of the Bayesian network model;

[0112] Decision tree construction: Feature selection: Select the features that have the greatest impact on the decision-making process as the input of the decision tree;

[0113] Build a decision tree: Use the training dataset to build a decision tree model to optimize the performance and interpretability of the tree;

[0114] Model evaluation: Use the test set to evaluate the performance of the decision tree model, including accuracy and recall indicators;

[0115] Combining Bayesian networks with decision trees: Integrated model: The output probability distribution or specific prediction of the Bayesian network is used as one of the input features of the decision tree. This allows the decision tree to consider the uncertainty information and probability predictions provided by the Bayesian network when making decisions;

[0116] Optimization and adjustment: Adjust the parameters of the Bayesian network and decision tree, as well as the integration method between them, according to actual needs to achieve the best prediction or decision-making effect.

[0117] Real-time decision-making: When new data LSTM prediction results or real-time observation data arrive, they are first processed by the Bayesian network, and then the processing results are input into the decision tree for real-time decision-making.

[0118] Result interpretation and action: Interpret the results of the decision tree output and take corresponding firefighting actions or measures based on the results. At the same time, monitor and evaluate the effect of the decision in order to carry out subsequent model optimization and adjustment.

[0119] In one embodiment, the user interaction unit receives the user's login information and operation instructions, the information display unit receives data from other modules and displays it to the user in real time, the query and report unit receives data from other modules for generating reports, the data acquisition unit collects data from various data sources such as sensors and databases in real time, the data cleaning unit receives data from the data acquisition unit and performs cleaning, the data denoising unit receives data from the data cleaning unit and performs denoising, the LSTM prediction unit receives data from the data denoising unit and performs prediction, the iForest real-time anomaly detection unit receives data from the LSTM prediction unit and performs anomaly detection, the risk assessment data preparation unit receives data from the data monitoring and preprocessing module and constructs a risk assessment data set, the Bayesian network and decision tree combination unit receives data from the risk assessment data preparation unit, combines LSTM and iForest results to perform risk assessment, and generates a risk assessment report, and the decision strategy formulation unit receives data from the data monitoring and preprocessing module and constructs a risk assessment data set. The data from the Bayesian network and decision tree combination unit is used to formulate a scheduling decision strategy. The decision optimization unit receives data from the decision strategy formulation unit and optimizes and adjusts the strategy. The resource information management unit maintains real-time information on rescue resources and personnel. The dynamic path planning unit receives data from the decision optimization unit and performs path planning. The scheduling instruction generation unit receives data from the dynamic path planning unit and generates scheduling instructions. The execution tracking and feedback unit receives data from the scheduling instruction generation unit and tracks the execution status. The dynamic adjustment and optimization unit receives data from the execution tracking and feedback unit and dynamically adjusts the strategy. The user authority management unit manages user accounts, roles and permissions. The system log management unit records logs such as system operation status, user operations, and abnormal information. The performance monitoring and optimization unit monitors system performance indicators, performs performance tuning and fault warning. The external system integration unit is responsible for the integration and interaction with external systems such as geographic information systems (GIS), communication systems, and other emergency management systems.

[0120] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A civil aviation fire emergency rescue device, characterized in that: include: Front-end user interface module, data monitoring and preprocessing module, risk assessment and decision-making module, resource scheduling and execution module, system management and detection module; The front-end user interface module is responsible for providing a graphical interface to support the interaction between the user and the system. The data monitoring and preprocessing module is responsible for collecting data from various data sources and performing cleaning, denoising and preprocessing. The risk assessment and decision-making module performs risk assessment and formulates decision-making strategies based on preprocessed data and real-time monitoring results. The resource scheduling and execution module is responsible for the scheduling and execution of resources. The system management and detection module is responsible for the overall management and monitoring of the system. The front-end user interface module is the entrance for users to interact with the system. Users input operating instructions and query conditions through the interface and receive feedback from the system. The data monitoring and preprocessing module is responsible for collecting and preprocessing data from the data source, and providing data support for the risk assessment and decision-making module. The risk assessment and decision-making module performs risk assessment and decision-making based on preprocessed data and real-time data, and passes the results to the resource scheduling and execution module. The resource scheduling and execution module generates and executes scheduling instructions according to the decision-making strategy, and provides real-time feedback on the execution status. The system management and monitoring module is responsible for monitoring and management of the entire system.

2. A civil aviation fire emergency rescue device according to claim 1, characterized in that: The front-end user interface module includes a user interaction unit, an information display unit, and a query and report unit. The user interaction unit is used to provide a graphical interface for user login, authority verification, and operation instruction input. The information display unit is used to display real-time monitoring data, risk assessment results, and scheduling instruction execution status. The query and report unit is used to support historical data query and statistical report generation.

3. A civil aviation fire emergency rescue device according to claim 1, characterized in that: The data monitoring and preprocessing module includes a data acquisition unit, a data cleaning unit, a data denoising unit, an LSTM prediction unit, and an iForest implementation anomaly detection unit. The data acquisition unit is responsible for collecting data from the data source in real time. The data cleaning unit is used to remove duplicate data, fill missing values, process abnormal values, etc. The data denoising unit is used to apply filtering technology to reduce data noise. The LSTM prediction unit analyzes historical time series data, learns the dependencies between data, and predicts indicators or trends at future time points based on this. The iForest real-time anomaly detection unit is used for real-time anomaly detection.

4. The civil aviation fire emergency rescue device according to claim 1, characterized in that: The risk assessment and decision-making module includes a risk assessment data preparation unit, a Bayesian network and decision tree combination unit, a decision strategy formulation unit, and a decision optimization unit. The risk assessment data preparation unit is used to integrate preprocessed data and historical data to construct a risk assessment data set. The Bayesian network and decision tree combination unit combines the prediction results of LSTM, the anomaly detection results of iForest and other data to perform risk assessment on emergency events and generate a risk assessment report. The decision strategy formulation unit is used to formulate a scheduling decision strategy based on the risk assessment results and the preset emergency plan. The decision optimization unit is used to consider resource availability and response time factors to optimize and adjust the decision strategy.

5. The civil aviation fire emergency rescue device according to claim 1, characterized in that: The resource scheduling and execution module includes a resource information management unit, a dynamic path planning unit, a scheduling instruction generation unit, an execution tracking and feedback unit, and a dynamic adjustment and optimization unit. The resource information management unit is used to maintain real-time information of rescue resources and personnel. The dynamic path planning unit is used to combine real-time traffic information, rescue needs and other factors, and adopt a path planning algorithm to perform optimal path planning. The scheduling instruction generation unit is used to generate specific scheduling instructions based on decision-making strategies and path planning results. The execution tracking and feedback unit is used to track the execution status of the rescue team in real time and collect feedback information, such as arrival time and task completion status. The dynamic adjustment and optimization unit is used to dynamically adjust scheduling strategies and resource allocation based on execution status feedback and real-time data changes.

6. The civil aviation fire emergency rescue device according to claim 1, characterized in that: The system management and monitoring module includes a user authority management unit, a system log management unit, a performance monitoring and optimization unit, and an external system integration unit. The user authority management unit is used to manage user accounts, roles and permissions; the system log management unit is used to record system logs for auditing and troubleshooting; the performance monitoring and optimization unit is used to monitor system performance indicators, perform performance tuning and fault warning; and the external system integration unit is responsible for the integration and interaction of external systems.

7. A method for using a civil aviation fire emergency rescue device according to claims 1-6, characterized in that: The following steps are involved: S1: User interface initialization: Login and authority verification: Log in to the system through the user interface of the dedicated fire emergency rescue device. The system verifies the user role and authority to ensure that only authorized users can access the system; Interface navigation and layout: The user interface adopts a fire theme design, provides a clear navigation menu and layout, and specially marks the function modules related to fire emergency, so that firefighters can locate and operate the required functions; S2: Data collection and preprocessing: Data collection: Users set fire data collection parameters for civil aviation airports and their surrounding environments through the user interface. The system monitors and collects these data in real time to provide a basis for subsequent emergency response; Preprocessing: During the preprocessing process, the system cleans, integrates and formats the collected fire data, removes noise data, and extracts key features. Users can view the progress of data cleaning and integration in real time on the interface, as well as an overview of the preprocessed fire data, to ensure the accuracy and timeliness of the data; S3: Model construction and risk assessment: Model construction: The user interface displays the construction progress and results of the LSTM fire prediction model and the iForest fire anomaly detection model for civil aviation firefighting scenarios in real time. Risk assessment: Users view the results of the comprehensive fire risk assessment through the interface. The system provides charts and reports to help firefighters understand the fire situation and formulate corresponding emergency response plans; S4: Emergency response decision making: Intelligent decision support: The user interface provides an intelligent decision support input interface for civil aviation fire emergency rescue. Users can adjust decision parameters based on real-time fire information and prediction results. The system uses algorithms and models to provide firefighters with a variety of emergency response plans and evaluate their feasibility and effectiveness; Decision making and display: During the decision-making process, users can view the simulated rescue effects under different parameter combinations on the interface, and the decision results are displayed on the user interface; S5: Execution and Feedback: Execution monitoring: Users monitor the execution of civil aviation fire emergency rescue in real time through the user interface. The system provides real-time data and video feedback to ensure that firefighters can fully grasp the dynamics of the rescue scene; Feedback adjustment: The user interface provides a feedback mechanism, allowing users to adjust the emergency response plan according to actual conditions. Firefighters can quickly make decisions and adjustments based on on-site feedback, and view the effects of the adjustments in real time; S6: Report output and experience summary: Report output: After the emergency response is completed, the user interface generates a detailed civil aviation fire emergency rescue report. Users can download or print the report as a basis for subsequent analysis and improvement; Experience summary: The user interface provides an input interface for experience summary, encouraging firefighters to record the experience and lessons of this emergency response. These experience summaries will be included in the system database for reference and sharing of similar incidents in the future, and continuously improve the capabilities and level of civil aviation fire emergency rescue.

8. The method for using the civil aviation fire emergency rescue device according to claim 7, characterized in that: In S3, the specific steps of using the LSTM prediction model are: Data collection and preprocessing: Time series data X = {x1, x2, ..., x T }, where x T represents the observation value at the Tth time point; Clean and preprocess the collected data, X'=preprocess(X), where preprocess represents the data cleaning and preprocessing function; Dataset Partition: Divide the preprocessed data into training set X train , validation set X val and the test set X test The training set is used for model training, the validation set is used for model parameter adjustment, and the test set is used to evaluate model performance; LSTM model construction: Design the LSTM network structure, including the input layer, multiple LSTM layers, and the fully connected layer; Determine the number of neurons, learning rate, and optimizer of the LSTM layer; Set LSTM model parameters: number of network layers: L, number of LSTM units per layer: N l (l=1,2,...,L), learning rate η, optimizer ο, build LSTM model: LSTM=LSTM(L,N l ,η,ο); Model training: Use the training set data to train the LSTM model: θ = train(LSTM, X train ,y train ), where θ represents the model parameters, y train is the label or target value corresponding to the training set. The network weights are adjusted through the back-propagation algorithm so that the model can learn the time dependency and nonlinear relationship in the data. During the training process, the validation set can be used to adjust the model parameters to find the optimal hyperparameter combination; Model evaluation: Use the test set to evaluate the trained LSTM model: performance = evaluate(LSTM, θ, X test ,y test ), calculate the prediction accuracy and error rate performance indicators; If the model performance does not meet the requirements, you can return to step 3 to readjust the model structure or hyperparameters, where performance represents the performance index; Real-time prediction: The latest time series data X new Input into the trained LSTM model: Generate forecast results for a period of time in the future, Indicates the prediction result; The prediction results are passed to the Bayesian network and decision tree combination unit for further processing.

9. The method for using the civil aviation fire emergency rescue device according to claim 7, characterized in that: In S3, the specific steps of using the iForest real-time anomaly detection are: Data integration: LSTM prediction results Real-time observation dataX real Integrate with other relevant data as the input dataset D for anomaly detection; iForest model construction: set iForest parameters, such as the number of trees n trees , subsample size ψ; Use the integrated data set to construct multiple isolation trees iForest = iForest(n trees ,ψ), each tree is built independently, and each tree randomly selects part of the data for training; Anomaly detection: For each point d in the input dataset D i , calculate the path length h(d i ,tree j ), where j = 1, 2, ..., n trees , calculate the average path length Serves as a benchmark for the entire dataset; Calculate anomaly score based on path length Where E[] represents the expectation, c(ψ) is the estimated path length based on the subsample size ψ, and the anomaly score is inversely proportional to the average path length, that is, the shorter the path, the higher the anomaly score; Set a threshold τ and compare the anomaly score with the preset threshold. If s(d i )>τ, the score is higher than the threshold, then the point d i It is an outlier. Result interpretation and response: Analyze the causes of abnormal points and trigger corresponding emergency responses based on the abnormal detection results.

10. The method for using the civil aviation fire emergency rescue device according to claim 7, characterized in that: In S3, the specific steps of obtaining the risk assessment result using the decision tree and Bayesian network are: Data collection: Gather all possible data relevant to the prediction or decision; Feature selection: Extract features useful for Bayesian network and decision tree modeling from the collected data. These features may include the average value of LSTM prediction results for a specific time period, the rate of change, and the abnormal indicators detected by iForest; Data preprocessing: Clean the features, such as dealing with missing values ​​and outliers, and standardizing or normalizing them to ensure the consistency and comparability of the data; Bayesian network construction: Define the network structure: Based on expert knowledge or data exploration results, define the nodes in the Bayesian network and the dependencies between them. Nodes can include LSTM prediction results, external factors, and system status; Learning parameters: using historical data or expert knowledge to estimate the conditional probability distribution between nodes in the network; Model validation: Use a portion of the validation set to verify the accuracy and effectiveness of the Bayesian network model; Decision tree construction: Feature selection: Select the features that have the greatest impact on the decision-making process as the input of the decision tree; Build a decision tree: Use the training dataset to build a decision tree model to optimize the performance and interpretability of the tree; Model evaluation: Use the test set to evaluate the performance of the decision tree model, including accuracy and recall indicators; Combining Bayesian networks with decision trees: Integrated model: The output probability distribution or specific prediction of the Bayesian network is used as one of the input features of the decision tree. This allows the decision tree to consider the uncertainty information and probability predictions provided by the Bayesian network when making decisions; Optimization and adjustment: Adjust the parameters of the Bayesian network and decision tree, as well as the integration method between them, according to actual needs to achieve the best prediction or decision-making effect. Real-time decision-making: When new data LSTM prediction results or real-time observation data arrive, they are first processed by the Bayesian network, and then the processing results are input into the decision tree for real-time decision-making. Result interpretation and action: Interpret the results of the decision tree output and take corresponding actions or measures based on the results. At the same time, monitor and evaluate the effect of the decision in order to carry out subsequent model optimization and adjustment.

Images