Emergency management intelligent detection alarm system and method based on multi-modal information

Abstract

The invention discloses an emergency management intelligent detection alarm system and method based on multi-modal information, and the system obtains and fuses the multi-source information of seismic waves, building structures, urban pipe networks and the like through a multi-modal seismic information collection and fusion module, and calculates the dynamic response through a dynamic analysis module based on an improved Newmark-beta algorithm. And a finite element simulation model building module is combined with city parameters to build a high-precision model. The LSTM-PINN hybrid analysis model module carries out deep analysis and prediction on data, the disaster risk assessment and classification module completes risk classification in combination with population and economic parameters, and finally the intelligent detection alarm module triggers corresponding alarms according to classification results and sends information in multiple channels. The method comprises the steps of information acquisition, dynamic analysis, model construction, disaster prediction, risk assessment and grading, alarm and the like, intelligent detection and accurate alarm of urban earthquake disasters are realized, and the emergency management efficiency and reliability are improved.

Description

Technical Field

[0001] The present invention relates to the field of urban earthquake disaster emergency management, and in particular to an emergency management intelligent detection and alarm system and method based on multimodal information. Background Art

[0002] As cities continue to expand and populations and infrastructure become increasingly concentrated, the potential threat posed by earthquakes is becoming increasingly severe. Urban environments, with their diverse building types and complex underground pipeline networks, create extremely complex dynamic responses during earthquakes. Traditional earthquake monitoring and emergency management methods are no longer sufficient to meet the needs of modern cities for efficient disaster prevention and mitigation. An intelligent system is urgently needed that can integrate multi-source information and accurately analyze disaster risks, enabling rapid and effective responses to earthquake disasters and safeguarding urban safety and the lives and property of residents.

[0003] Existing technologies have significant deficiencies in data processing and system coordination. On the one hand, data processing capabilities are weak. Most emergency management systems rely on a single or small number of sensor types to collect data, making it impossible to fully capture multimodal information such as earthquake wave characteristics, building structural conditions, and urban pipeline network operations. Furthermore, the collected data is difficult to effectively integrate, resulting in information fragmentation and the inability to form a complete and accurate assessment of the disaster situation. On the other hand, system coordination is poor, with each functional module relatively independent. Data transmission and interaction are subject to delays and obstacles. From data collection to analysis and prediction, to risk assessment and alarm issuance, each link is poorly connected, making it difficult to form an efficient emergency management chain. This significantly reduces the timeliness and accuracy of disaster warnings, making it difficult to provide timely and effective information support to relevant departments and the public during the critical period of a disaster. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of the prior art, the present invention provides an emergency management intelligent detection and alarm system and method based on multimodal information.

[0005] The technical solution adopted by the present invention is an emergency management intelligent detection and alarm system based on multimodal information, comprising:

[0006] Multimodal seismic information acquisition and fusion module, which uses different types of sensors deployed in various calibration areas of the city to synchronously collect the acceleration, velocity, and displacement information of seismic waves, the stress and strain information of building structures, and the pressure and flow information of the urban pipe network system, and preliminarily fuses the collected information; a dynamic analysis module based on the improved Newmark-β algorithm, which receives the information output by the multimodal seismic information acquisition and fusion module, uses the improved Newmark-β algorithm, and combines the mechanical parameters of urban buildings and infrastructure to perform calculations and analysis on the dynamic response under earthquake action; a finite element simulation model construction module, which constructs a high-precision urban finite element simulation model based on urban geographic information data, building design parameters, and geological structure parameters, and receives dynamic analysis based on the improved Newmark-β algorithm. The calculation results of the analysis module are used to set the boundary conditions and loads of the finite element simulation model; the LSTM-PINN hybrid analysis model module combines the simulation data generated by the finite element simulation model construction module with the original information of the multimodal earthquake information acquisition and fusion module as input data, and uses the LSTM-PINN hybrid analysis model to conduct in-depth analysis and prediction of urban earthquake disasters; the disaster risk assessment and classification module receives the output results of the LSTM-PINN hybrid analysis model module, combines the urban population distribution and economic density parameters, assesses the earthquake disaster risk, and classifies it according to preset standards; the intelligent detection and alarm module automatically triggers the corresponding level of alarm signal based on the classification results of the disaster risk assessment and classification module, and sends the alarm information to relevant departments and public terminals through different communication methods.

[0007] Furthermore, in the dynamic analysis module based on the improved Newmark-β algorithm, the improved Newmark-β algorithm introduces the damping correction coefficient α of different structural types of urban buildings. d , the dynamic equation is established as follows: Where M is the structural mass matrix, is the acceleration vector at time n+1, is the velocity vector at time n+1, u n+1 is the displacement vector at time n+1, C is the structural damping matrix, K is the structural stiffness matrix, F n+1 is the external force vector at time n+1.

[0008] Furthermore, in the finite element simulation model construction module, when constructing the finite element model of the urban underground pipe network system, the soil-pipeline interaction caused by seismic wave propagation is combined to establish the equation: K s-p {δ s-p}={F s-p}, where K s-pis the soil-pipe interaction stiffness matrix, {δ s-p} is the displacement vector of soil-pipe interaction, {F s-p} is the external force vector of soil-pipeline interaction.

[0009] Furthermore, in the LSTM-PINN hybrid analysis model module, the calculation equation from the input layer to the hidden layer of the LSTM network is: t =f(W xh x t +W hh h t-1 +b h ), where h t is the hidden layer state at time t, x t is the input layer data at time t, W xh is the weight matrix from the input layer to the hidden layer, W hh is the weight matrix from hidden layer to hidden layer, b h is the hidden layer bias vector, and f is the activation function. At the same time, the PINN part is based on the seismic design parameters of urban buildings and the characteristics of the seismic wave spectrum to construct the loss function equation: L = L physics +λL data , where L physics is the loss term based on physical laws, L data is the loss term based on data fitting, and λ is the weight coefficient.

[0010] Furthermore, in the disaster risk assessment and classification module, the urban population density distribution function ρ(x, y) and the economic density distribution function ω(x, y) are introduced to construct the risk assessment equation: R = ∫ S ∫ T P(t, x, y)ρ(x, y)ω(x, y)dxdy, where R is the risk assessment value, S is the urban area, T is the duration of the earthquake, and P(t, x, y) is the probability of an earthquake disaster occurring at time t and location (x, y).

[0011] Furthermore, in the intelligent detection alarm module, an alarm trigger threshold equation is established for different risk levels: T i =θ i R max , where T i is the alarm triggering threshold of the i-th level risk, θ i is the threshold coefficient of the i-th level risk, R maxis the maximum risk assessment value; when the risk assessment value output by the disaster risk assessment and classification module reaches or exceeds the alarm trigger threshold of the corresponding level, the intelligent detection alarm module automatically triggers the alarm signal of that level and sends the alarm information to relevant departments and public terminals through different communication methods such as mobile communication network, broadcast system and SMS platform.

[0012] Furthermore, the finite element simulation model construction module includes:

[0013] Geographic information modeling unit, which constructs a three-dimensional geometric model of urban topography based on high-precision urban geographic surveying data, provides a geographic spatial basis for finite element simulation, and determines the model's boundary range and the impact of terrain on seismic wave propagation;

[0014] Building parameter input unit, which receives the building design drawings and related parameters, including building structure type, material properties, and size specifications, and establishes a finite element model of the building. The finite element model truly reflects the mechanical properties of the building;

[0015] The geological structure analysis unit analyzes the urban underground geological structure. Based on geological exploration data, it establishes finite element models of different geological layers, determines the mechanical connection between the layers, and simulates the propagation characteristics of seismic waves in different geological layers.

[0016] The model integration and optimization unit integrates the models established by the geographic information modeling unit, the building parameter input unit and the geological structure analysis unit, checks the connection parts of the model, and optimizes the mesh division of the model.

[0017] Furthermore, the LSTM-PINN hybrid analysis model module includes:

[0018] A data preprocessing unit, which unifies the format of the simulation data generated by the finite element simulation model building module and the original information of the multimodal seismic information acquisition and fusion module, and performs data screening to remove invalid data;

[0019] LSTM network training unit, which uses preprocessed data to train the LSTM network according to the set network parameters to learn the time series characteristics and nonlinear relationships of earthquake disaster data;

[0020] PINN model construction unit, which builds the PINN model based on the physical laws of urban earthquake disasters, including dynamic equations and wave equations, and embeds physical constraints into the PINN model to ensure that the model prediction results are consistent with physical reality;

[0021] The model fusion and optimization unit fuses the trained LSTM network and the constructed PINN model, and optimizes the performance of the hybrid model by adjusting the fusion parameters.

[0022] Furthermore, the disaster risk assessment and classification module includes:

[0023] Basic data collection unit, which collects the city's population distribution data, economic data, building seismic resistance data and basic data related to disaster risk assessment to provide data support for risk assessment;

[0024] The risk factor analysis unit analyzes the collected basic data to identify the calibration factors that affect the risk of earthquake disasters, including densely populated areas, important economic facilities, and buildings with weak earthquake resistance;

[0025] Quantitative assessment calculation unit, which uses the established risk assessment equation and the risk factor analysis results to quantitatively calculate the urban earthquake disaster risk and obtain a specific risk assessment value;

[0026] The grading judgment execution unit compares the risk assessment value obtained by the quantitative assessment calculation unit with the preset grading standard, determines the earthquake disaster risk level, and outputs the corresponding grading result to provide a basis for intelligent detection and alarm.

[0027] The intelligent detection and alarm method for emergency management based on multimodal information includes the following steps:

[0028] In the first step, different types of sensors deployed in the multimodal earthquake information acquisition and fusion module are used to synchronously collect multimodal information related to urban earthquakes and perform preliminary fusion processing;

[0029] The second step is to transmit the information after preliminary fusion processing to the dynamic analysis module based on the improved Newmark-β algorithm. The improved Newmark-β algorithm is combined with the mechanical parameters of urban buildings and infrastructure to calculate the dynamic response under earthquake action.

[0030] The third step is to build a city finite element simulation model in the finite element simulation model construction module based on urban geographic information data, building design parameters and geological structure parameters, and set the model boundary conditions and loads according to the calculation results of the dynamic analysis module;

[0031] The fourth step is to combine the simulation data generated by the finite element simulation model with the original information from the multimodal earthquake information acquisition and fusion module and input it into the LSTM-PINN hybrid analysis model module. The LSTM-PINN hybrid analysis model is then used to conduct in-depth analysis and prediction of urban earthquake disasters.

[0032] In the fifth step, the disaster risk assessment and classification module receives the output of the hybrid analysis model module and combines it with the urban population distribution and economic density parameters to assess and classify the earthquake disaster risk.

[0033] In the sixth step, the intelligent detection and alarm module automatically triggers the corresponding level of alarm signal based on the classification results of the disaster risk assessment and classification module, and sends the alarm information to relevant departments and public terminals through different communication methods.

[0034] Beneficial Effects: The present invention proposes an intelligent detection and alarm system and method for emergency management based on multimodal information. The system utilizes a multimodal earthquake information acquisition and fusion module, deploying multiple sensors to simultaneously collect multiple types of information, including seismic waves, building structures, and urban pipeline networks, and then performs preliminary fusion. This overcomes the limitations of traditional single-source data acquisition and addresses the issues of insufficient and fragmented information sources. The improved Newmark-β algorithm, combined with the finite element simulation model construction module, can accurately calculate dynamic responses and construct high-precision simulation models based on the city's actual geographic, architectural, and geological parameters. This fully considers complex factors and significantly improves the accuracy of simulations of earthquake dynamic responses and disaster evolution processes compared to traditional simple analysis models. The LSTM-PINN hybrid analysis model module fuses and analyzes simulation data with raw information, deeply exploring disaster patterns and trends in the data and addressing the lack of depth and breadth in traditional analysis models. The disaster risk assessment and classification module combines demographic and economic parameters to scientifically assess and classify risks, providing support for accurate early warning. The modules are connected in an orderly manner, and data interacts smoothly, forming an efficient chain from information collection to alarm release. This completely solves the problems of poor coordination and delayed response in traditional systems, and realizes rapid and accurate detection and graded alarms of urban earthquake disasters, providing strong guarantees for urban emergency management and significantly improving the effectiveness of disaster prevention and mitigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a diagram of the modular unit composition of the present invention;

[0036] Figure 2 The figure is a flow chart of the method steps of the present invention. DETAILED DESCRIPTION

[0037] It should be noted that, unless there is a conflict, the embodiments in this application and the features described in the embodiments can be combined with each other. The application is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] like Figure 1 As shown, the emergency management intelligent detection and alarm system based on multimodal information includes:

[0039] A multimodal seismic information acquisition and fusion module uses different types of sensors deployed in designated areas of the city to simultaneously collect acceleration, velocity, and displacement information of seismic waves, stress and strain information of building structures, and pressure and flow information of urban pipe networks, and then performs preliminary fusion processing on the collected information;

[0040] Specifically, the multimodal seismic information acquisition and fusion module, as the primary link in the system's acquisition of raw data, undertakes critical data collection and preliminary integration. By deploying various types of sensors in key urban areas, such as high-rise building clusters, areas with dense underground pipeline networks, and transportation hubs like bridges, this module simultaneously collects acceleration, velocity, and displacement information of seismic waves in three mutually perpendicular directions. These information are core parameters reflecting seismic motion characteristics. Furthermore, strain gauges and stress sensors are used to collect real-time stress and strain states of building structures to assess their seismic loads. For urban pipeline systems, pressure and flow sensors are used to obtain pressure and flow data to monitor the network's operational status during earthquakes. This multi-source, heterogeneous information is initially fused to eliminate data conflicts and redundancies, forming structured data that provides an accurate and comprehensive foundation for subsequent analysis.

[0041] In terms of implementation, a scientific sensor deployment plan must first be developed based on the city's geographical layout, building distribution, and pipeline network planning to ensure that key areas are effectively covered. When selecting sensors, performance indicators such as accuracy, range, and anti-interference ability must be considered. For example, seismic wave sensors must have high sensitivity to capture weak seismic signals, and their range must meet the requirements of possible strong earthquakes; building structure sensors must have good durability and stability to adapt to complex building environments. The collected data is transmitted to the data processing center via wired or wireless communication networks. At the data processing center, data alignment, feature extraction, and other technologies are used to perform preliminary data fusion. For example, different types of sensor data are synchronized based on timestamps, and feature extraction algorithms are used to extract key data features to form a standardized data format for subsequent module calls.

[0042] A dynamic analysis module based on the improved Newmark-β algorithm. This module receives the information output by the multimodal seismic information acquisition and fusion module and uses the improved Newmark-β algorithm in combination with the mechanical parameters of urban buildings and infrastructure to perform calculations and analysis on the dynamic response under earthquake action.

[0043] Specifically, the dynamic analysis module based on the improved Newmark-β algorithm is the core unit for dynamic calculations of multi-modal data. This module receives multi-source data after preliminary fusion, and uses the improved Newmark-β algorithm to accurately calculate the dynamic response of urban buildings and infrastructure under earthquakes, combining their specific mechanical parameters such as the elastic modulus and Poisson's ratio of the material, the mass distribution of the structure, the stiffness matrix, etc. The improved algorithm introduces a damping correction mechanism for different building structure types, which can more accurately simulate the energy consumption characteristics of the structure during earthquake motion. By calculating the acceleration, velocity and displacement response of the structure at each moment of the earthquake, it provides key dynamic parameters for the subsequent finite element simulation model construction, thereby gaining a deeper understanding of the dynamic behavior and failure mechanism of the structure under earthquake action.

[0044] During implementation, the first step is to establish a mechanical parameter database for urban buildings and infrastructure, obtaining accurate mechanical parameters through architectural design drawings, construction materials, and on-site inspections. The data output by the multimodal seismic information acquisition and fusion module is imported into the dynamic analysis module. Based on the characteristics of the seismic waves and the structural mechanical parameters, the calculation parameters of the improved Newmark-β algorithm are set. For example, the time step must be reasonably selected based on the frequency components of the seismic waves and the natural vibration period of the structure to ensure calculation accuracy and efficiency. The algorithm uses iterative calculations to gradually solve the dynamic response of the structure under earthquake action. Each calculation step considers the nonlinear factors and damping corrections of the structure, and ultimately outputs the structural dynamic response data, which is stored in the form of a data file or database table for reading and use by the finite element simulation model construction module.

[0045] Finite element simulation model construction module, which builds a high-precision urban finite element simulation model based on urban geographic information data, building design parameters and geological structure parameters. It also receives the calculation results of the dynamic analysis module based on the improved Newmark-β algorithm and sets the boundary conditions and loads of the finite element simulation model;

[0046] Specifically, the finite element simulation model construction module is dedicated to establishing a high-precision three-dimensional finite element simulation model of the city to intuitively and accurately simulate the earthquake disaster process. Based on detailed urban geographic information data, including topography and elevation data; building design parameters such as building floor plan layout, structural form, and component dimensions; and geological structure parameters such as soil layer distribution and geomechanical properties, this module uses professional finite element modeling software to construct a detailed finite element model covering various urban elements. At the same time, it receives the calculation results output by the dynamic analysis module based on the improved Newmark-β algorithm and uses them as the boundary conditions and loads of the model to simulate the propagation of seismic waves in the urban area and the response of the structure under the action of the earthquake, thereby achieving a realistic reproduction of the urban earthquake disaster scene and providing a visual and quantitative model foundation for subsequent disaster analysis and prediction.

[0047] In specific implementation, a 3D geometric model of the urban topography is constructed using Geographic Information System (GIS) data and a Digital Elevation Model (DEM), providing a spatial framework for the finite element model. For buildings, 3D solid modeling is performed in modeling software based on design drawings, with precise meshing, material properties, and mechanical parameters assigned. For geological structures, models of different geological layers are created based on geological exploration data, and interlayer contact relationships are determined. Ground motion parameters, such as acceleration time histories, calculated using an improved Newmark-β algorithm, are applied to the model boundaries as seismic excitation. The model undergoes mesh optimization and parameter verification, and its accuracy and reliability are ensured by comparing it with actual engineering cases or theoretical calculation results. Ultimately, a complete urban finite element simulation model is generated and stored in a specific format for use in the LSTM-PINN hybrid analysis model module.

[0048] The LSTM-PINN hybrid analysis model module combines the simulation data generated by the finite element simulation model construction module with the original information from the multimodal earthquake information acquisition and fusion module as input data, and uses the LSTM-PINN hybrid analysis model to conduct in-depth analysis and prediction of urban earthquake disasters;

[0049] Specifically, the LSTM-PINN hybrid analysis model module is a key component for in-depth analysis and prediction of urban earthquake hazards. This module combines the simulation data generated by the finite element simulation model construction module with the raw information from the multimodal earthquake information acquisition and fusion module as input. The simulation data contains information about earthquake disaster processes simulated based on physical laws, while the raw information is earthquake-related data acquired through actual monitoring, complementing each other. The LSTM (Long Short-Term Memory) network automatically learns the time series characteristics and nonlinear relationships in the data, exploring patterns in data change over time. The PINN (Physical Information Neural Network) embeds physical constraints into the neural network based on the physical laws of urban earthquake hazards, such as dynamic equations and wave equations. This ensures that the model's predictions are consistent with both data characteristics and physical laws. This organic integration of the two enables accurate predictions of key information, such as the development trend and impact range of urban earthquake hazards, providing an important basis for disaster risk assessment and classification.

[0050] In terms of implementation, the input data is first normalized, mapping data of different types and magnitudes to the same numerical range to improve model training efficiency and stability. An LSTM-PINN hybrid network architecture is then constructed, and parameters such as the number of layers and neurons are determined. The hybrid model is trained using the normalized data. During training, the LSTM component uses a backpropagation algorithm to update network weights and learn the time series characteristics of the data. The PINN component adjusts network parameters by minimizing a loss function based on physical laws, ensuring that the model's predictions meet physical constraints. Model parameters are continuously optimized through multiple rounds of iterative training until the model's prediction accuracy meets the set standard. The resulting LSTM-PINN hybrid analysis model is capable of accurately analyzing and predicting urban earthquake disasters, processing new input data in real time and outputting predictions.

[0051] Disaster risk assessment and classification module, which receives the output of the LSTM-PINN hybrid analysis model module, combines urban population distribution and economic density parameters to assess earthquake disaster risks and classify them according to preset standards;

[0052] Specifically, the Disaster Risk Assessment and Classification module uses the output of the LSTM-PINN hybrid analysis model module and the city's actual conditions to scientifically assess and classify earthquake disaster risks. This module comprehensively considers multiple factors, including urban population distribution, economic density, and building seismic resistance. By establishing a risk assessment model, it quantifies the potential losses incurred by different regions and types of facilities in the event of an earthquake disaster. This module uses urban population distribution data to determine casualty risk, combines economic density data to assess economic loss risk, and uses building seismic resistance data to determine building damage risk. This comprehensive analysis of these factors categorizes earthquake disaster risk into different levels (low, medium, high, and extremely high) according to pre-set risk grading criteria. This provides a clear and specific basis for emergency management departments to formulate targeted response measures and early warning plans.

[0053] In specific implementation, a basic urban information database is first established, integrating relevant data such as population distribution, economic data, and building seismic resistance ratings. Upon receiving the prediction results from the LSTM-PINN hybrid analysis model module, the predicted earthquake disaster impact range and intensity information is correlated with the data in the basic information database. Using risk assessment algorithms such as the Analytic Hierarchy Process (AHP) and the Fuzzy Comprehensive Evaluation Method (Fuzzy Comprehensive Evaluation Method), each risk factor is weighted and comprehensively calculated to derive a risk assessment value for each area or facility. This calculated risk assessment value is compared with the preset risk classification threshold to determine the corresponding risk level. A risk assessment report is then generated, detailing the risk level of each area or facility, the risk factor analysis results, and other information, which is then transmitted to the intelligent detection and alarm module.

[0054] Intelligent detection and alarm module, which automatically triggers the corresponding level of alarm signal based on the classification results of the disaster risk assessment and classification module, and sends the alarm information to relevant departments and public terminals through different communication methods.

[0055] Specifically, the intelligent detection and alarm module is the last link in the system to realize the emergency warning function. It sends out alarm signals in a timely and accurate manner according to the classification results of the disaster risk assessment and classification module. This module sets corresponding alarm trigger thresholds and alarm methods for different risk levels. When the received risk assessment level reaches or exceeds a certain threshold, the alarm signal of the corresponding level is automatically triggered. For low-risk levels, reminder information may be sent to the public through SMS or APP push; for medium-risk levels, in addition to SMS and APP push, warnings will also be issued through community broadcasts; for high-risk and extremely high-risk levels, a full range of alarm mechanisms will be activated, including television, radio, SMS, APP push, sirens and other methods to ensure that relevant departments and the public can obtain disaster warning information in a timely manner so that they can quickly take response measures to reduce the losses caused by earthquake disasters.

[0056] During the implementation process, the first step is to establish a comprehensive communication network and alarm equipment management system. The communication network must cover multiple communication methods such as mobile communication networks, the Internet, and broadcast networks to ensure that alarm information can be transmitted stably and quickly. Alarm equipment, such as sirens and broadcast systems, must be regularly maintained and inspected to ensure normal operation. After the disaster risk assessment and classification module outputs risk level information, the intelligent detection and alarm module calls the corresponding communication interface and alarm equipment based on the preset correspondence between risk level and alarm method, and generates and sends alarm information. At the same time, the system will record information such as the alarm sending time, recipient, and alarm content to facilitate subsequent inquiries and traceability of emergency management work.

[0057] The output end of the multimodal seismic information acquisition and fusion module is connected to the input end of the dynamic analysis module based on the improved Newmark-β algorithm, the output end of the dynamic analysis module based on the improved Newmark-β algorithm is connected to the input end of the finite element simulation model construction module, the output end of the finite element simulation model construction module is connected to the input end of the LSTM-PINN hybrid analysis model module, the output end of the LSTM-PINN hybrid analysis model module is connected to the input end of the disaster risk assessment and classification module, and the output end of the disaster risk assessment and classification module is connected to the input end of the intelligent detection and alarm module.

[0058] Preferably, in the dynamic analysis module based on the improved Newmark-β algorithm, the improved Newmark-β algorithm introduces the damping correction coefficient α of different structural types of urban buildings. d , the dynamic equation is established as follows: Where M is the structural mass matrix, is the acceleration vector at time n+1, is the velocity vector at time n+1, u n+1 is the displacement vector at time n+1, C is the structural damping matrix, K is the structural stiffness matrix, F n+1 is the external force vector at time n+1. This equation is used to calculate the dynamic response of urban buildings and infrastructure under earthquakes, providing basic data for the subsequent finite element simulation model construction.

[0059] Specifically, the improved Newmark-β algorithm incorporates damping correction coefficients for different structural types of urban buildings. These coefficients are specifically designed to account for the varying energy dissipation characteristics of diverse urban structures, such as brick-concrete, frame, and steel structures, during earthquakes. The structural mass matrix, damping matrix, and stiffness matrix are determined based on the building's actual design parameters and material properties. These parameters are obtained from building design drawings, construction materials, and material test reports, accurately reflecting the building's mechanical properties. The acceleration, velocity, and displacement vectors represent the structure's motion at different moments during an earthquake, while the external force vector is derived from the loads exerted by seismic waves on the structure. During implementation, preliminary fused earthquake-related information, including seismic wave characteristics and partial building response data, is first obtained from the multimodal seismic information acquisition and fusion module. These data are then combined with corresponding mechanical parameters retrieved from a database of building and infrastructure mechanical parameters and substituted into the improved dynamic equations. During the calculation process, an appropriate time step is selected based on the duration and characteristics of the seismic waves, and iterative calculations are performed to gradually solve for the acceleration, velocity, and displacement responses of the structure at each moment during the earthquake. These calculation results provide key dynamic basic data for the subsequent construction of finite element simulation models, enabling the finite element model to more realistically simulate the dynamic behavior of buildings and infrastructure during earthquakes, laying a solid foundation for the entire system's analysis and prediction of earthquake disasters.

[0060] Preferably, in the finite element simulation model building module, when building a finite element model for the urban underground pipe network system, the soil-pipeline interaction caused by seismic wave propagation is combined to establish the equation: K s-p {δ s-p}={F s-p}, where K s-p is the soil-pipe interaction stiffness matrix, {δ s-p} is the displacement vector of soil-pipe interaction, {F s-p} is the external force vector for the soil-pipeline interaction. This equation accurately simulates the mechanical behavior of underground pipeline systems under earthquakes. Combined with the pipe diameter, material, and burial depth parameters of urban pipelines, it improves the finite element simulation model, enhances simulation accuracy, and provides more reliable input data for the LSTM-PINN hybrid analysis model.

[0061] Specifically, the finite element simulation model construction module focuses on the urban underground pipeline system. When constructing the finite element model, it fully considers the soil-pipeline interaction caused by seismic wave propagation. This is because the mechanical behavior of underground pipelines during earthquakes is closely related to the interaction with the surrounding soil. The soil-pipeline interaction stiffness matrix reflects the ability of the soil and pipeline to resist relative deformation. Its value is closely related to parameters such as soil type and density, as well as pipeline diameter, material, and burial depth. The interaction displacement vector reflects the relative displacement between the soil and pipeline under earthquake action, while the interaction external force vector is a comprehensive reflection of the force exerted by the soil on the pipeline and the inertial force generated by the pipeline itself due to the earthquake. In specific implementation, the pipeline parameters such as diameter, material, and burial depth are first determined based on the design drawings and construction data of the urban underground pipeline network. Geological exploration data is also used to determine the type and mechanical properties of the soil surrounding the pipeline. These parameters are then substituted into the equations describing the soil-pipeline interaction to calculate the mechanical response data of the pipeline network under earthquake action. Next, using specialized finite element modeling software, the underground pipe network model was integrated with models of other parts of the city, combining urban geographic information data, design parameters of other buildings, and geological structural parameters to construct a complete finite element simulation model of the city. This approach accurately simulated the mechanical behavior of the underground pipe network system under earthquakes, improving the overall accuracy of the finite element simulation model and providing more reliable input data for the subsequent LSTM-PINN hybrid analysis model, facilitating more accurate analysis and prediction of earthquake impacts on the city's underground pipe network.

[0062] Preferably, in the LSTM-PINN hybrid analysis model module, the calculation equation from the input layer to the hidden layer of the LSTM network is: t =f(W xh x t +W hh h t-1 +b h ), where h t is the hidden layer state at time t, x t is the input layer data at time t, W xh is the weight matrix from the input layer to the hidden layer, W hh is the weight matrix from hidden layer to hidden layer, b h is the hidden layer bias vector, and f is the activation function. At the same time, the PINN part is based on the seismic design parameters of urban buildings and the characteristics of the seismic wave spectrum to construct the loss function equation: L = L physics +λL data , where L physics is the loss term based on physical laws, L datais the loss term based on data fitting, and λ is the weight coefficient. The above equation effectively integrates LSTM and PINN to conduct in-depth analysis and prediction of urban earthquake disasters, providing a basis for disaster risk assessment and classification.

[0063] Specifically, the LSTM-PINN hybrid analysis model module combines an LSTM network with a PINN to achieve in-depth analysis and prediction of urban earthquake hazards. In the LSTM network, the hidden layer state at time t is calculated from the current input layer data, the hidden layer state at the previous time, and the corresponding weight matrix and bias vector. The activation function introduces nonlinearity into the network, enabling the LSTM network to learn the complex time series characteristics and nonlinear relationships of earthquake disaster data and exploit the inherent patterns of data change over time. The loss function equation constructed by the PINN component uses a physics-based loss term to ensure that the model's prediction results conform to the physical laws related to earthquake hazards, such as dynamic equations and wave equations. A data-fitting loss term ensures that the model fits the actual collected and simulated data well. Weight coefficients are used to balance the importance of the two. During implementation, the simulated data generated by the finite element simulation model construction module is first integrated with the raw data from the multimodal earthquake information acquisition and fusion module as input data for the LSTM-PINN hybrid analysis model. This data is then preprocessed, including normalization. The LSTM network and PINN model are then trained separately. The LSTM network continuously adjusts its weight matrix and bias vector through a backpropagation algorithm to minimize the error between predicted and actual values, thereby learning the time series characteristics of earthquake disaster data. The PINN model embeds physical constraints into the model by adjusting its own parameters to minimize the loss term based on physical laws and the loss term based on data fitting. Finally, the trained LSTM network and the constructed PINN model are integrated by adjusting the weight coefficients in the loss function to optimize the performance of the hybrid model, enabling it to accurately analyze and predict key information such as the development trend and impact range of urban earthquake disasters, providing an important basis for disaster risk assessment and classification.

[0064] Preferably, the disaster risk assessment and classification module introduces the urban population density distribution function ρ(x, y) and the economic density distribution function ω(x, y) to construct the risk assessment equation: R = ∫ S ∫ TP(t, x, y)ρ(x, y)ω(x, y)dxdy, where R is the risk assessment value, S is the urban area, T is the earthquake duration, and P(t, x, y) is the probability of an earthquake disaster occurring at time t and location (x, y). This equation, combined with parameters such as urban population distribution and economic density, quantitatively assesses earthquake disaster risk and categorizes it according to preset grading standards, including low risk, medium risk, high risk, and extremely high risk, providing accurate classification signals for intelligent detection and alarms.

[0065] Specifically, the disaster risk assessment and classification module constructs a risk assessment equation by incorporating urban population density and economic density distribution functions, comprehensively considering the important role of urban population and economic factors in earthquake disaster risk. The risk assessment value is calculated by integrating the probability of an earthquake disaster occurring at different times and locations within the urban area and during the earthquake duration with the population density and economic density functions. The urban area defines the geographical boundaries for the assessment, while the earthquake duration reflects the duration of the seismic effect. The probability of an earthquake disaster is determined based on historical earthquake data, geological structure, and predictions from the LSTM-PINN hybrid analysis model module. The population density and economic density distribution functions are determined using urban census data and economic statistics. Implementation begins with establishing a comprehensive urban basic information database, collecting various information relevant to disaster risk assessment, including population distribution data, economic data, historical earthquake data, and geological structure data. Upon receiving the earthquake disaster prediction results output by the LSTM-PINN hybrid analysis model module, such as the earthquake's impact range and intensity, they are correlated with data in the basic information database. Using the risk assessment equation, each parameter is substituted into the calculation, and a risk assessment value for each area or facility is obtained through integration. Finally, the calculated risk assessment value is compared with the pre-set risk grading standard, and the risk is classified into low risk, medium risk, high risk, and extremely high risk levels, generating a detailed risk assessment report. This report provides accurate classification signals for the intelligent detection and alarm module and provides a scientific basis for emergency management departments to formulate targeted disaster response strategies and resource allocation plans.

[0066] Preferably, in the intelligent detection alarm module, an alarm trigger threshold equation is established for different risk levels: T i =θ i R max , where T i is the alarm triggering threshold of the i-th level risk, θ i is the threshold coefficient of the i-th level risk, R maxis the maximum risk assessment value. When the risk assessment value output by the Disaster Risk Assessment and Grading Module reaches or exceeds the corresponding alarm trigger threshold, the Intelligent Detection and Alarm Module automatically triggers an alarm signal for that level and sends the alarm information to relevant departments and public terminals via mobile communication networks, broadcast systems, and SMS platforms, enabling rapid and accurate disaster warnings.

[0067] Specifically, the intelligent detection and alarm module establishes alarm trigger threshold equations for different risk levels, aiming to achieve precise, graded alarms for earthquake disaster risks. The alarm trigger threshold for level i is calculated by multiplying the threshold coefficient for level i by the maximum risk assessment value. The threshold coefficient is determined through extensive data analysis and empirical research, based on factors such as the impact of historical earthquake disasters, the response capabilities of emergency management departments, and social tolerance. Different threshold coefficients correspond to different risk levels, reflecting the varying severity of each risk. The maximum risk assessment value is determined based on a comprehensive consideration of the city's size, building density, and the distribution of critical infrastructure, representing the potential risk level a city could face in the event of a severe earthquake disaster. During implementation, when the disaster risk assessment and grading module outputs a risk assessment value for a specific area or facility, the intelligent detection and alarm module compares this value with each level of threshold value calculated using the alarm trigger threshold equation. Once the risk assessment value reaches or exceeds a certain level of alarm trigger threshold, the intelligent detection and alarm module automatically triggers an alarm signal for the corresponding level. For low-risk levels, the system may use relatively mild methods such as text messages and mobile phone application push to send disaster warning information to the public; for medium-risk levels, in addition to text messages and application push, the warning range will be expanded through community broadcasting and other methods; and for high-risk and extremely high-risk levels, a full-scale, high-intensity alarm mechanism will be activated, including television broadcasts, group text messages, sirens, mobile phone emergency notifications and other communication methods to ensure that relevant departments, rescue agencies and the public can obtain accurate disaster warning information at the first time, so as to quickly take effective response measures and minimize casualties and property losses caused by earthquake disasters.

[0068] Preferably, the finite element simulation model building module includes:

[0069] Geographic information modeling unit, which constructs a three-dimensional geometric model of urban topography based on high-precision urban geographic surveying data, provides an accurate geographic spatial basis for finite element simulation, and determines the model's boundary range and the impact of terrain on seismic wave propagation;

[0070] Building parameter input unit, which receives the building's design drawings and related parameters, including building structure type, material properties, and size specifications, and accurately establishes the building's finite element model to ensure that the model can truly reflect the building's mechanical properties;

[0071] The geological structure analysis unit analyzes the urban underground geological structure. Based on geological exploration data, it establishes finite element models of different geological layers, determines the mechanical connection between the layers, and simulates the propagation characteristics of seismic waves in different geological layers.

[0072] Model integration and optimization unit, which integrates the models established by the geographic information modeling unit, the building parameter input unit and the geological structure analysis unit, checks the connection parts of the model, optimizes the mesh division of the model, and improves the overall accuracy and calculation efficiency of the finite element simulation model.

[0073] Specifically, the finite element simulation model construction module comprises multiple functional units working together. The geographic information modeling unit constructs a three-dimensional geometric model of the city's topography based on high-precision urban geographic mapping data, including satellite imagery and topographic survey data. This model accurately reproduces the city's topographical features, such as topographic undulations and elevation changes, providing an accurate geospatial foundation for finite element simulations and defining the model's boundaries. It also considers the impact of topography on the propagation path and velocity of seismic waves. The building parameter input unit receives detailed building design drawings and relevant parameters, including structural type, material elastic modulus, Poisson's ratio, and component dimensions. Using specialized finite element modeling software, it accurately creates a finite element model of each building, ensuring that the model accurately reflects the building's mechanical properties, including its stress and deformation under earthquake action. The geological structure analysis unit analyzes the city's underground geological structure based on geological exploration data, such as borehole data and geological radar detection results. It creates finite element models of different geological layers and accurately determines the mechanical connections between these layers. This model simulates the propagation characteristics of seismic waves in these layers, including wave velocity variations and energy attenuation. The Model Integration and Optimization Unit integrates the models created by the Geographic Information Modeling Unit, the Building Parameter Entry Unit, and the Geological Structure Analysis Unit. It checks the accuracy and rationality of the connections between the models and optimizes the model's meshes, improving computational efficiency while ensuring accuracy. Through the close collaboration of these units, a highly accurate and comprehensive urban finite element simulation model is constructed, providing a reliable foundation for subsequent analysis and prediction of urban earthquake hazards.

[0074] Preferably, the LSTM-PINN hybrid analysis model module includes:

[0075] Data preprocessing unit, which unifies the format of the simulation data generated by the finite element simulation model building module and the original information of the multimodal seismic information acquisition and fusion module, removes invalid data, and provides high-quality data input for subsequent model training;

[0076] LSTM network training unit, which uses preprocessed data to train the LSTM network according to the set network parameters, learns the time series characteristics and nonlinear relationships of earthquake disaster data, and improves the model's ability to predict the evolution of earthquake disasters;

[0077] PINN model construction unit, which builds the PINN model based on the physical laws of urban earthquake disasters, including dynamic equations and wave equations, and embeds physical constraints into the model to ensure that the model prediction results are consistent with physical reality;

[0078] The model fusion and optimization unit fuses the trained LSTM network and the constructed PINN model, optimizes the performance of the hybrid model by adjusting the fusion parameters, and achieves more accurate analysis and prediction of urban earthquake disasters.

[0079] Specifically, the various units in the LSTM-PINN hybrid analysis model module have clear divisions of labor and collaborate with each other. The data preprocessing unit processes the simulation data generated by the finite element simulation model construction module and the raw information from the multimodal seismic information acquisition and fusion module. Because these data come from different sources, have different formats and data types, and may contain invalid or erroneous data, the data preprocessing unit removes invalid data through operations such as format unification, data cleaning, and normalization. It converts different types of data into the same format and maps the data to appropriate numerical ranges, providing high-quality, standardized data input for subsequent model training. The LSTM network training unit uses the preprocessed data to train the LSTM network according to pre-set parameters such as the number of network layers, number of neurons, and learning rate. During training, the LSTM network continuously adjusts the weight matrix and bias vector to learn the time series characteristics and nonlinear relationships of earthquake disaster data, exploring the patterns of data evolution over time, thereby improving the model's ability to predict the evolution of earthquake disasters. The PINN model construction unit constructs the PINN model based on the physical laws related to urban earthquake hazards, such as dynamic equations and wave equations. These physical constraints are mathematically embedded into the neural network to ensure that the model's predictions are consistent with physical reality and avoid irrational predictions that violate physical laws. The model fusion and optimization unit fuses the trained LSTM network with the constructed PINN model. By adjusting fusion parameters, such as adjusting the weights of LSTM and PINN predictions in the loss function, the performance of the hybrid model is optimized. This allows the hybrid model to fully leverage the advantages of LSTM learning data features and PINN following physical laws, achieving more accurate analysis and prediction of urban earthquake hazards and providing reliable results for disaster risk assessment.

[0080] Preferably, the disaster risk assessment and classification module includes:

[0081] Basic data collection unit, which collects the city's population distribution data, economic data, building seismic rating data and basic data related to disaster risk assessment, providing comprehensive data support for risk assessment;

[0082] The risk factor analysis unit analyzes the collected basic data to identify the calibration factors that affect the risk of earthquake disasters, including densely populated areas, important economic facilities, and buildings with weak earthquake resistance;

[0083] Quantitative assessment calculation unit, which uses the established risk assessment equation and the risk factor analysis results to quantitatively calculate the urban earthquake disaster risk and obtain a specific risk assessment value;

[0084] The grading judgment execution unit compares the risk assessment value obtained by the quantitative assessment calculation unit with the preset grading standard, determines the earthquake disaster risk level, and outputs the corresponding grading result to provide a basis for intelligent detection and alarm.

[0085] Specifically, the various units in the Disaster Risk Assessment and Classification Module work together to assess and classify earthquake disaster risks. The Basic Data Collection Unit is responsible for extensively collecting basic urban data relevant to disaster risk assessment, including but not limited to urban population distribution data. Detailed population size, density, and distribution are obtained through censuses, household registration management, and other channels. Economic data covers the distribution of various economic facilities, economic output, and other information, obtained from economic statistics departments and corporate surveys. Building seismic rating data determines the seismic resistance of each building through building design information and seismic appraisal reports. The Risk Factor Analysis Unit conducts in-depth analysis of the large amount of basic data collected, applying data analysis algorithms and specialized knowledge to identify key factors influencing earthquake disaster risk, such as densely populated areas, important economic facilities, and buildings with weak seismic resistance. These factors are important bases for assessing earthquake disaster risk. The Quantitative Assessment Calculation Unit uses the established risk assessment equation, combined with the analysis results of the Risk Factor Analysis Unit, to substitute the relevant parameters into the equation for calculation. During the calculation process, the weights and interrelationships of various risk factors are fully considered, and the urban earthquake disaster risk is quantified to produce a specific risk assessment value. This value can intuitively reflect the potential risk level of different areas or facilities in earthquake disasters. The grading judgment execution unit strictly compares the risk assessment value obtained by the quantitative assessment calculation unit with the preset grading standards, and determines the risk level according to low risk, medium risk, high risk, and extremely high risk, and outputs the corresponding grading results. These grading results provide accurate alarm basis for the intelligent detection and alarm module, and also provide important reference for emergency management departments to formulate scientific and reasonable emergency response strategies and deploy rescue resources.

[0086] like Figure 2 As shown, the intelligent detection and alarm method for emergency management based on multimodal information includes the following steps:

[0087] In the first step, different types of sensors deployed in the multimodal earthquake information acquisition and fusion module are used to synchronously collect multimodal information related to urban earthquakes and perform preliminary fusion processing;

[0088] The second step is to transmit the information after preliminary fusion processing to the dynamic analysis module based on the improved Newmark-β algorithm. The improved Newmark-β algorithm is combined with the mechanical parameters of urban buildings and infrastructure to calculate the dynamic response under earthquake action.

[0089] The third step is to build a city finite element simulation model in the finite element simulation model construction module based on urban geographic information data, building design parameters and geological structure parameters, and set the model boundary conditions and loads according to the calculation results of the dynamic analysis module;

[0090] The fourth step is to combine the simulation data generated by the finite element simulation model with the original information from the multimodal earthquake information acquisition and fusion module and input it into the LSTM-PINN hybrid analysis model module. The LSTM-PINN hybrid analysis model is then used to conduct in-depth analysis and prediction of urban earthquake disasters.

[0091] In the fifth step, the disaster risk assessment and classification module receives the output of the hybrid analysis model module and combines the urban population distribution and economic density parameters to assess and classify the earthquake disaster risk;

[0092] In the sixth step, the intelligent detection and alarm module automatically triggers the corresponding level of alarm signal based on the classification results of the disaster risk assessment and classification module, and sends the alarm information to relevant departments and public terminals through different communication methods.

[0093] The intelligent detection and alarm system and method for emergency management based on multimodal information, at the data acquisition level, the system deploys multiple types of sensors to build a multimodal earthquake information acquisition and fusion module, and simultaneously obtains multi-source data such as seismic waves, building structures and urban pipelines, effectively solving the problem of single information source and severe fragmentation in traditional systems. Through preliminary fusion processing, it provides comprehensive and relevant basic data for subsequent analysis.

[0094] In the analysis and prediction phase, the system deeply integrates an improved Newmark-β algorithm, finite element simulation, and an LSTM-PINN hybrid analysis model. The dynamic analysis module, based on the improved Newmark-β algorithm, combines the mechanical parameters of urban buildings and infrastructure to accurately calculate the dynamic response of earthquakes. The finite element simulation model construction module constructs a high-precision model based on geographical, architectural, and geological parameters, taking into account complex factors such as soil-pipeline interactions. Compared to traditional simple models, this significantly improves the accuracy and reliability of simulations. The LSTM-PINN hybrid analysis model module combines simulation data with original information, using LSTM to learn the time series characteristics of the data and PINN to incorporate physical constraints. This enables in-depth analysis and accurate prediction of earthquake disasters, overcoming the lack of depth and breadth of traditional analysis models.

[0095] In terms of risk assessment and emergency response, the Disaster Risk Assessment and Classification module combines parameters such as urban population distribution and economic density to scientifically quantify and classify risks. The Intelligent Detection and Alarm module automatically triggers alarms of corresponding levels based on the classification results, rapidly transmitting information through multiple communication channels. The close connection between modules and efficient data exchange establish a complete and seamless emergency management chain, fundamentally changing the poor coordination and delayed response of traditional systems. This system enables intelligent and precise management of the entire process, from disaster monitoring, analysis and prediction, to graded alarms. This significantly improves the efficiency and reliability of urban earthquake disaster emergency management, providing a solid foundation for the safe operation of cities and the safety of residents' lives and property.

[0096] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0097] While embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Emergency management intelligent detection and alarm system based on multimodal information, characterized by: include: A multimodal seismic information acquisition and fusion module uses different types of sensors deployed in designated areas of the city to simultaneously collect acceleration, velocity, and displacement information of seismic waves, stress and strain information of building structures, and pressure and flow information of urban pipe networks, and then performs preliminary fusion processing on the collected information; A dynamic analysis module based on the improved Newmark-β algorithm. This module receives the information output by the multimodal seismic information acquisition and fusion module and uses the improved Newmark-β algorithm in combination with the mechanical parameters of urban buildings and infrastructure to perform calculations and analysis on the dynamic response under earthquake action. Finite element simulation model construction module, which builds a high-precision urban finite element simulation model based on urban geographic information data, building design parameters and geological structure parameters. It also receives the calculation results of the dynamic analysis module based on the improved Newmark-β algorithm and sets the boundary conditions and loads of the finite element simulation model; LSTM-PINN hybrid analysis model module, which combines the simulation data generated by the finite element simulation model construction module with the original information of the multimodal earthquake information acquisition and fusion module as input data, and uses the LSTM-PINN hybrid analysis model to conduct in-depth analysis and prediction of urban earthquake disasters; disaster risk assessment and classification module, which receives the output results of the LSTM-PINN hybrid analysis model module, combines urban population distribution and economic density parameters to assess earthquake disaster risks, and classifies them according to preset standards; intelligent detection and alarm module, which automatically triggers the corresponding level of alarm signal based on the classification results of the disaster risk assessment and classification module, and sends the alarm information to relevant departments and public terminals through different communication methods.

2. The intelligent detection and alarm system for emergency management based on multimodal information according to claim 1 is characterized in that: In the dynamic analysis module based on the improved Newmark-β algorithm, the improved Newmark-β algorithm introduces the damping correction coefficient α of different structural types of urban buildings. d , the dynamic equation is established as follows: Where M is the structural mass matrix, is the acceleration vector at time n+1, is the velocity vector at time n+1, u n+1 is the displacement vector at time n+1, C is the structural damping matrix, K is the structural stiffness matrix, F n+1 is the external force vector at time n+1.

3. The intelligent detection and alarm system for emergency management based on multimodal information according to claim 1 is characterized in that: In the finite element simulation model construction module, when constructing the finite element model of the urban underground pipe network system, the soil-pipeline interaction caused by seismic wave propagation is combined to establish the equation: K s-p {δ s-p }={F s-p }, where K s-p is the soil-pipe interaction stiffness matrix, {δ s-p } is the displacement vector of soil-pipe interaction, {F s-p } is the external force vector of soil-pipeline interaction.

4. The intelligent detection and alarm system for emergency management based on multimodal information according to claim 1 is characterized in that: In the LSTM-PINN hybrid analysis model module, the calculation equation from the input layer to the hidden layer of the LSTM network is: t =f(W xh x t +W hh h t-1 +b h ), where h t is the hidden layer state at time t, x t is the input layer data at time t, W xh is the weight matrix from the input layer to the hidden layer, W hh is the weight matrix from hidden layer to hidden layer, b h is the hidden layer bias vector, and f is the activation function. At the same time, the PINN part is based on the seismic design parameters of urban buildings and the characteristics of the seismic wave spectrum to construct the loss function equation: L = L physics +λL data , where L physics is the loss term based on physical laws, L data is the loss term based on data fitting, and λ is the weight coefficient.

5. The intelligent detection and alarm system for emergency management based on multimodal information according to claim 1 is characterized in that: In the disaster risk assessment and classification module, the urban population density distribution function ρ(x, y) and the economic density distribution function ω(x, y) are introduced to construct the risk assessment equation: R = ∫ S ∫ T P(t, x, y)ρ(x, y)ω(x, y)dxdy, where R is the risk assessment value, S is the urban area, T is the duration of the earthquake, and P(t, x, y) is the probability of an earthquake disaster occurring at time t and location (x, y).

6. The intelligent detection and alarm system for emergency management based on multimodal information according to claim 1 is characterized in that: In the intelligent detection alarm module, an alarm trigger threshold equation is established for different risk levels: T i =θ i R max , where T i is the alarm triggering threshold for level i risk, θ i is the threshold coefficient of the i-th level risk, R max is the maximum risk assessment value; when the risk assessment value output by the disaster risk assessment and classification module reaches or exceeds the alarm trigger threshold of the corresponding level, the intelligent detection alarm module automatically triggers the alarm signal of that level and sends the alarm information to relevant departments and public terminals through different communication methods such as mobile communication network, broadcast system and SMS platform.

7. The intelligent detection and alarm system for emergency management based on multimodal information according to claim 1 is characterized in that: The finite element simulation model building module includes: Geographic information modeling unit, which constructs a three-dimensional geometric model of urban topography based on high-precision urban geographic surveying data, provides a geographic spatial basis for finite element simulation, and determines the model's boundary range and the impact of terrain on seismic wave propagation; Building parameter input unit, which receives the building design drawings and related parameters, including building structure type, material properties, and size specifications, and establishes a finite element model of the building. The finite element model truly reflects the mechanical properties of the building; The geological structure analysis unit analyzes the urban underground geological structure. Based on geological exploration data, it establishes finite element models of different geological layers, determines the mechanical connection between the layers, and simulates the propagation characteristics of seismic waves in different geological layers. The model integration and optimization unit integrates the models established by the geographic information modeling unit, the building parameter input unit and the geological structure analysis unit, checks the connection parts of the model, and optimizes the mesh division of the model.

8. The intelligent detection and alarm system for emergency management based on multimodal information according to claim 1 is characterized in that: The LSTM-PINN hybrid analysis model module includes: A data preprocessing unit, which unifies the format of the simulation data generated by the finite element simulation model building module and the original information of the multimodal seismic information acquisition and fusion module, and performs data screening to remove invalid data; LSTM network training unit, which uses preprocessed data to train the LSTM network according to the set network parameters to learn the time series characteristics and nonlinear relationships of earthquake disaster data; PINN model construction unit, which builds the PINN model based on the physical laws of urban earthquake disasters, including dynamic equations and wave equations, and embeds physical constraints into the PINN model to ensure that the model prediction results are consistent with physical reality; The model fusion and optimization unit fuses the trained LSTM network and the constructed PINN model, and optimizes the performance of the hybrid model by adjusting the fusion parameters.

9. The intelligent detection and alarm system for emergency management based on multimodal information according to claim 1 is characterized in that: The disaster risk assessment and classification module includes: Basic data collection unit, which collects the city's population distribution data, economic data, building seismic resistance data and basic data related to disaster risk assessment to provide data support for risk assessment; The risk factor analysis unit analyzes the collected basic data to identify the calibration factors that affect the risk of earthquake disasters, including densely populated areas, important economic facilities, and buildings with weak earthquake resistance; Quantitative assessment calculation unit, which uses the established risk assessment equation and the risk factor analysis results to quantitatively calculate the urban earthquake disaster risk and obtain a specific risk assessment value; The grading judgment execution unit compares the risk assessment value obtained by the quantitative assessment calculation unit with the preset grading standard, determines the earthquake disaster risk level, and outputs the corresponding grading result to provide a basis for intelligent detection and alarm.

10. An intelligent detection and alarm method for emergency management based on multimodal information, characterized in that: The following steps are involved: In the first step, different types of sensors deployed in the multimodal earthquake information acquisition and fusion module are used to synchronously collect multimodal information related to urban earthquakes and perform preliminary fusion processing; The second step is to transmit the information after preliminary fusion processing to the dynamic analysis module based on the improved Newmark-β algorithm. The improved Newmark-β algorithm is combined with the mechanical parameters of urban buildings and infrastructure to calculate the dynamic response under earthquake action. The third step is to build a city finite element simulation model in the finite element simulation model construction module based on urban geographic information data, building design parameters and geological structure parameters, and set the model boundary conditions and loads according to the calculation results of the dynamic analysis module; The fourth step is to combine the simulation data generated by the finite element simulation model with the original information from the multimodal earthquake information acquisition and fusion module and input it into the LSTM-PINN hybrid analysis model module. The LSTM-PINN hybrid analysis model is then used to conduct in-depth analysis and prediction of urban earthquake disasters. In the fifth step, the disaster risk assessment and classification module receives the output of the hybrid analysis model module and combines the urban population distribution and economic density parameters to assess and classify the earthquake disaster risk; In the sixth step, the intelligent detection and alarm module automatically triggers the corresponding level of alarm signal based on the classification results of the disaster risk assessment and classification module, and sends the alarm information to relevant departments and public terminals through different communication methods.

Images