Emergency shelter site selection evaluation system based on AI

Abstract

The invention provides an AI-based emergency shelter site selection evaluation system, and relates to the field of evaluation, and the system comprises the following modules: a multi-source data collection module which obtains building seismic grade data through a BIM system of a building department, and the data comprises a structure type, a design reference period, and reinforcement recording three-dimensional parameters; according to the method, through multi-source data fusion, various factors such as building seismic grade data, space safety, building spacing, crowd density, secondary disaster vulnerability index, ecological buffering efficiency, water source accessibility and emergency material reserve are integrated in allusion to natural earthquake disasters, and dynamic weight adjustment is combined, so that the seismic data of the building can be obtained. According to the method and the system, people can selectively list buildings with high dynamic safety scores into own journeys through the pushed dynamic safety scores in earthquake high-incidence areas and earthquake frequent-incidence areas, so that the buildings with high dynamic safety scores can be selectively listed into own journeys when people plan the journeys in the future; and the survival rate of people in a building with high safety and the convenience of people dodging can be improved, and the method has practical significance.

Description

Technical Field

[0001] The present invention relates to the field of new energy, and in particular to an AI-based emergency shelter site selection assessment system. Background Art

[0002] At present, when selecting shelters for natural disasters such as earthquakes, priority is given to open places, such as stadiums. These places are not easily hit by collapsed objects when an earthquake occurs, and the chance of survival is higher. However, judging whether to select a shelter based solely on whether it is open is too simple data, and open places in a city are too scarce. When major natural disasters such as earthquakes occur, there are too few shelters and too few options for people to choose from, which is not conducive to people's prevention of earthquakes and avoidance during earthquakes.

[0003] Therefore, we made improvements to this and proposed an AI-based emergency shelter site selection assessment system. Summary of the invention

[0004] The purpose of the present invention is to address the problems raised by the current background technology.

[0005] In order to achieve the above-mentioned purpose of the invention, the present invention provides an AI-based emergency shelter site selection assessment system to improve the above-mentioned problems.

[0006] The specific application is as follows: Includes the following modules: Multi-source data acquisition module: Obtain building seismic grade data through the BIM system of the housing and construction department, including structural type, design reference period, and three-dimensional parameters of reinforcement records; A 3D model with 0.1-meter accuracy was generated using drone oblique photography to calculate the building spacing index and open space ratio, where: , The minimum distance between adjacent buildings. is the area of ​​the region, is the evacuable area, is the total area; Deploy a distributed fiber optic sensor network with a node density of ≥50 / km2, collect regional population density (PDPM) in real time, and update data every 30 seconds; Dynamic weight allocation module: Establish a three-level weighting system: Core layer (weight 60%): seismic resistance level (weight 40%) + space safety (BSI×OSR, weight 20%); Risk layer (weight 25%): secondary disaster vulnerability index (weight 15%) + ecological buffer efficiency (weight 10%); Security layer (weight 15%): water accessibility (weight 8%) + emergency material reserves (weight 7%); Dynamic rating generation module: Constructing a spatiotemporal feature fusion neural network: Input layer: 65-dimensional feature vector, including earthquake resistance parameters (12 dimensions), spatial parameters (18 dimensions), environmental parameters (25 dimensions), and social parameters (10 dimensions); Hidden layer: Bidirectional LSTM units are used to capture the temporal characteristics of earthquake precursor data, and the graph convolutional network (GCN) is combined to extract spatial correlation; Output layer: Generates a dynamic safety score of 0-100 points, updated every 10 minutes, with an error of ≤±2.5 points.

[0007] As the preferred technical solution of this application, it also includes a hierarchical push module: The weekly dynamic safety score will be pushed to the mobile terminals of people in earthquake-prone areas and high-risk areas via text messages. If the dynamic safety score is less than 60 points, the area will be forced to push it.

[0008] As the preferred technical solution of this application, the secondary disaster vulnerability index calculation includes: Model of toxic gas diffusion in underground space: ,in, is the diffusion rate of poison gas, is the acceleration due to gravity, is the depth of underground pipeline, and Represent the density of air and poison gas respectively.

[0009] As the preferred technical solution of this application, the secondary disaster vulnerability index calculation also includes: Building collapse chain reaction: Prediction of shock wave propagation paths using a concrete crack growth simulator: ,in is the attenuation coefficient, is the seismic wave frequency, is the initial amplitude, Phase angle, Indicates time.

[0010] As the preferred technical solution of this application, the ecological buffering efficiency evaluation includes: Biological-geological coupling analysis: , Indicates that the tree has deep roots, Indicates root width, Indicates the inclination angle of the landslide surface.

[0011] As the preferred technical solution of this application, the ecological buffering efficiency evaluation also includes: Soil moisture monitoring: The soil moisture content is inverted by synthetic aperture radar (SAR). When the moisture content is >28% and the continuous rainfall is >3 hours, the ecological buffer efficiency weight is automatically increased to 12%.

[0012] As the preferred technical solution of this application, the dynamic score generation module also includes: Psychological Panic Modifier: Calculate regional panic index through social media sentiment analysis API: RPI=0.6×(negative keyword frequency)+0.4×(exclamation mark density).

[0013] As the preferred technical solution of this application, the specific implementation of the dynamic weight allocation and scoring generation module includes: S100, weight dynamic adjustment algorithm: Core layer weight update: When the regional population density is >200% and the duration is ≥5 minutes, the spatial safety weight is adjusted to 35%. If the regional population density is ≤150%, the spatial safety weight is adjusted to 20%.

[0014] As the preferred technical solution of this application, the poison gas diffusion model implementation standard is: Toxic gas sensors are deployed every 50 meters in underground pipelines, with an installation depth of ≥2 meters. When three consecutive sensors detect v>5m / s, an early warning is pushed.

[0015] As the preferred technical solution of this application, the fiber optic sensor deployment rules include: Commercial buildings: Node density is 70 / km², installation height is 4.5-5.0m; Residential buildings: node density is 50 / km², installation height is 3.0-3.5m; Industrial area buildings: The node density is 60 / km² and the installation height is 6.0-8.0m.

[0016] Compared with the prior art, the present invention has the following beneficial effects: In the scheme of this application: 1. Through multi-source data fusion, when dealing with earthquake natural disasters, we integrate multiple factors such as building seismic resistance data, spatial safety, building spacing, crowd density, secondary disaster vulnerability index, ecological buffer efficiency, water accessibility and emergency material reserves, and combine the adjustment of dynamic weights to enable people to push dynamic safety scores in earthquake-prone and frequent areas. In this way, when people plan their future itineraries, they can selectively include buildings with high dynamic safety scores in their itineraries. In this way, when an earthquake occurs, it can improve people's survival rate in safe buildings and the convenience of people's escape, which is of practical significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the AI-based emergency shelter site selection assessment system provided for this application. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0019] As described in the background technology, judging whether to select a shelter only by whether it is open is too simplistic, and open spaces in a city are too scarce. When major natural disasters such as earthquakes occur, there are too few shelters and too few options for people to choose from, which is not conducive to people's prevention of earthquakes and sheltering during earthquakes.

[0020] In order to solve this technical problem, the present invention provides an AI-based emergency shelter site selection assessment system, which is applied to the assessment field.

[0021] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0022] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features and technical solutions in the embodiments may be combined with each other.

[0023] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0024] Example 1, please refer to Figure 1, an AI-based emergency shelter site selection assessment system, including the following modules: Multi-source data acquisition module: Obtain building seismic grade data through the BIM system of the housing and construction department, including structural type, design reference period, and three-dimensional parameters of reinforcement records; A 3D model with 0.1m accuracy was generated using drone oblique photography to calculate the building spacing index (BSI) and open space ratio (OSR), where: , The minimum distance between adjacent buildings. is the area of ​​the region, is the evacuable area, is the total area; BIM Data Integration: Implementation method: Connect to the BIM system API of the housing and construction department (in compliance with GB / T 51269-2017 standard), extract the building structure type, design reference period, and three-dimensional parameters of reinforcement records, and update the data once an hour.

[0025] Technical significance: Through digital mapping of seismic resistance levels, reliable basic data is provided for weight allocation Drone Modeling: Equipment parameters: DJI Matrice 350 RTK drone, flight altitude 100±2 meters, heading overlap rate 80%±3%; Algorithm flow: Generate a 3D model using ContextCapture and calculate the building spacing index (BSI) and open space ratio (OSR); Verification rules: Daily comparison of OpenStreetMap data, triggering manual review when deviation > 5% Deploy a distributed fiber optic sensor network with a node density of ≥50 / km2, collect regional population density (PDPM) in real time, and update data every 30 seconds; Dynamic weight allocation module: Establish a three-level weighting system: Core layer (weight 60%): seismic resistance level (weight 40%) + space safety (BSI×OSR, weight 20%); Risk layer (weight 25%): secondary disaster vulnerability index (15%) + ecological buffer efficiency (10%); Security layer (weight 15%): water accessibility (8%) + emergency supplies reserves (7%); Dynamic weight allocation module: Three-level weighting system: Core layer (weight 60%): Seismic level (40%): Automatic classification based on BIM data (I-IV); Spatial safety (20%): Dynamic product of BSI × OSR. Weight increase is triggered when the distance between buildings is less than 15 meters.

[0026] Risk layer (weight 25%): Secondary disaster vulnerability index (15%): weighted calculation of the gas diffusion rate and the probability of building collapse; Ecological buffering efficiency (10%): TSCI index and SAR moisture content are jointly evaluated.

[0027] Guarantee layer (weight 15%): Water source accessibility (8%): ​​Fire hydrant density within 1 km ≥ 2 / km²; Emergency supplies reserves (7%): The average per capita reserve in the shelter should be ≥ 3 days’ supply.

[0028] Dynamic rating generation module: Construct spatiotemporal feature fusion neural network (ST-FNN): Input layer: 65-dimensional feature vector, including earthquake resistance parameters (12 dimensions), spatial parameters (18 dimensions), environmental parameters (25 dimensions), and social parameters (10 dimensions); Hidden layer: Bidirectional LSTM units are used to capture the temporal characteristics of earthquake precursor data, and the graph convolutional network (GCN) is combined to extract spatial correlation; Output layer: Generates a dynamic safety score (DSS) of 0-100 points, updated every 10 minutes, with an error of ≤±2.5 points; Abnormal scenario processing: When the instantaneous increase in crowd density is monitored to exceed 200%, the Bayesian network is activated to dynamically adjust the weight distribution, giving priority to increasing the spatial safety weight to 35%.

[0029] Dynamic rating generation module: ST-FNN Neural Network: Model structure: Input layer: 65-dimensional vector (12-dimensional earthquake resistance parameters + 18-dimensional spatial parameters + 25-dimensional environmental parameters + 10-dimensional social parameters); Hidden layer: Bidirectional LSTM (128 units) extracts temporal features, and GCN (graph convolutional network) captures spatial correlation; Output layer: Sigmoid function generates a 0-100 DSS score.

[0030] Training specifications: Dataset: 28,500 positive samples (DSS ≥ 70) + 3,500 negative samples (DSS ≤ 40); Hyperparameters: learning rate 0.001 (Adam optimizer), batch size 64, dropout rate 0.3.

[0031] Psychological Panic Modifiers: RPI index calculation: Negative keyword library ("collapse", "suffocation", "aftershock") matching frequency; Exclamation mark density = number of exclamation marks per thousand words of text × 100%; VR calming training: Meta Quest Pro device provides 5 minutes of breathing guidance, temporary safety bonus +8 points.

[0032] It also includes a hierarchical push module: The weekly dynamic safety score is pushed to the mobile terminals of people in earthquake-prone areas and high-risk areas via text messages. If the dynamic safety score is <60 points, the area will be forced to push it, and the national encryption SM4 algorithm will be used to encrypt the user location data (accuracy ±3 meters) (this technology is already known, so it will not be repeated).

[0033] Through multi-source data fusion, when dealing with earthquake natural disasters, we have integrated multiple factors such as building seismic resistance data, spatial safety, building spacing, crowd density, secondary disaster vulnerability index, ecological buffer efficiency, water accessibility and emergency material reserves. Combined with the adjustment of dynamic weights, people can get dynamic safety scores through push notifications in earthquake-prone and earthquake-frequent areas. In this way, when people plan their future itineraries, they can selectively include buildings with high dynamic safety scores in their itineraries. This will improve people's survival rate in safe buildings and the convenience of people's escape when an earthquake occurs, which is of practical significance.

[0034] For example, in areas where earthquakes occur frequently, when many small earthquakes occur, people can plan their future itineraries based on the dynamic safety score, and can give priority to buildings with high dynamic safety scores as activity areas. This can increase people's survival rate when an earthquake occurs and reduce people's panic. In particular, the dynamic safety score is dynamically updated, which can also prevent high-scoring buildings from having their scores lowered when the population density increases. In this way, people can choose other high-scoring buildings to reduce congestion, and have autonomous selectivity, which is more reasonable.

[0035] Example 2 further optimizes the AI-based emergency shelter site selection assessment system provided in Example 1. Specifically, the secondary disaster vulnerability index calculation includes: Model of toxic gas diffusion in underground space: ,in, is the diffusion rate of poison gas, is the acceleration due to gravity, is the depth of underground pipeline, and Represent the density of air and poison gas respectively.

[0036] The secondary disaster vulnerability index calculation also includes: Building collapse chain reaction: Prediction of shock wave propagation paths by Concrete Crack Extension Simulator (CCES): ,in is the attenuation coefficient, is the seismic wave frequency, is the initial amplitude, Phase angle, Indicates time.

[0037] Ecological buffer effectiveness assessment includes: Biological-geological coupling analysis: , Indicates that the tree has deep roots, Indicates root width, Indicates the inclination angle of the landslide surface. When TSCI<0.4, an ecological reinforcement warning is triggered. When TSCI<0.4, a tree planting plan is automatically generated (such as pine trees with a root depth ≥1.5 meters). SAR moisture content monitoring: when the moisture content is >28% and the rainfall is >3 hours, the ecological weight is increased to 12%.

[0038] Ecological buffer effectiveness assessment also includes: Soil moisture monitoring: The soil moisture content is inverted by synthetic aperture radar (SAR). When the moisture content is >28% and the continuous rainfall is >3 hours, the ecological buffer efficiency weight is automatically increased to 12%.

[0039] Embodiment 3 further optimizes the AI-based emergency shelter site selection evaluation system provided in Embodiment 1 or 2. Specifically, the dynamic score generation module further includes: Psychological Panic Modifier: Calculate the Regional Panic Index (RPI) via the Social Media Sentiment Analysis API: RPI=0.6×(negative keyword frequency)+0.4×(exclamation mark density). When RPI>65, the virtual reality calmness training module is activated. After completing the training, +8 temporary safety points can be obtained. Negative keywords include "collapse", "suffocation" and "aftershock". The exclamation mark density is the number of exclamation marks per thousand words of text.

[0040] The specific implementation of the dynamic weight allocation and score generation module includes: S100, weight dynamic adjustment algorithm: Core layer weight update: When the regional population density is >200% and the duration is ≥5 minutes, the spatial safety weight is adjusted to 35%. If the regional population density is ≤150%, the spatial safety weight is adjusted to 20%. Bayesian network parameters: The prior probability is calculated based on the data of the past 30 days, and the dimension of the conditional probability table (CPT) is 12×8×6; S200, ST-FNN neural network training specifications: Training dataset: Positive samples: Global earthquake safe area data from 2010 to 2024 (n=28,500), with DSS score ≥70; Negative samples: dangerous area data (n=3,500), annotated with DSS score ≤40; Hyperparameter settings: Learning rate: 0.001 (Adam optimizer), batch size: 64, Dropout rate: 0.3; Real-time inference acceleration: Using NVIDIA TensorRT 8.6 quantized model, the inference time is ≤ 0.8 seconds at FP16 precision.

[0041] Gas diffusion model implementation standards: Honeywell SGP30 toxic gas sensors are deployed every 50 meters in underground pipelines, with an installation depth of ≥2 meters. When three consecutive sensors detect v>5m / s, an early warning is pushed.

[0042] Fiber optic sensor deployment rules include: Commercial buildings: Node density is 70 / km², installation height is 4.5-5.0m; Residential buildings: node density is 50 / km², installation height is 3.0-3.5m; Industrial area buildings: The node density is 60 / km² and the installation height is 6.0-8.0m.

[0043] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral one; it can be a mechanical connection, an electrical connection, or communication with each other; it can be a direct connection, or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0044] Obviously, the embodiments described above are only some embodiments of the present invention, rather than all embodiments. The preferred embodiments of the present invention are given in the accompanying drawings, but they do not limit the patent scope of the present invention. The present invention can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive. Although the present invention has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific embodiments, or to perform equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of the present invention, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present invention.

Claims

1. An AI-based emergency shelter site selection assessment system, characterized in that: Includes the following modules: Multi-source data acquisition module: Obtain building seismic grade data through the BIM system of the housing and construction department, including structural type, design reference period, and three-dimensional parameters of reinforcement records; A 3D model with 0.1m accuracy was generated using drone oblique photography to calculate the building spacing index and open space ratio, where: , The minimum distance between adjacent buildings. is the area of ​​the region, is the evacuable area, is the total area; Deploy a distributed fiber optic sensor network with a node density of ≥50 / km2, collect regional population density (PDPM) in real time, and update data every 30 seconds; Dynamic weight allocation module: Establish a three-level weighting system: Core layer (weight 60%): seismic resistance level (weight 40%) + space safety (BSI×OSR, weight 20%); Risk layer (weight 25%): secondary disaster vulnerability index (weight 15%) + ecological buffer efficiency (weight 10%); Security layer (weight 15%): water accessibility (weight 8%) + emergency material reserves (weight 7%); Dynamic rating generation module: Constructing a spatiotemporal feature fusion neural network: Input layer: 65-dimensional feature vector, including earthquake resistance parameters (12 dimensions), spatial parameters (18 dimensions), environmental parameters (25 dimensions), and social parameters (10 dimensions); Hidden layer: Bidirectional LSTM units are used to capture the temporal characteristics of earthquake precursor data, and graph convolutional networks (GCNs) are combined to extract spatial correlations; Output layer: Generates a dynamic safety score of 0-100 points, updated every 10 minutes, with an error of ≤±2.5 points.

2. The AI-based emergency shelter site selection assessment system according to claim 1, characterized in that: It also includes a hierarchical push module: The weekly dynamic safety score will be pushed to the mobile terminals of people in earthquake-prone areas and high-risk areas via text messages. If the dynamic safety score is less than 60 points, the area will be forced to push it.

3. The AI-based emergency shelter site selection assessment system according to claim 1, characterized in that: The secondary disaster vulnerability index calculation includes: Model of toxic gas diffusion in underground space: ,in, is the diffusion rate of poison gas, is the acceleration due to gravity, is the depth of underground pipeline, and Represent the density of air and poison gas respectively.

4. The AI-based emergency shelter site selection assessment system according to claim 3, characterized in that: The secondary disaster vulnerability index calculation also includes: Building collapse chain reaction: Prediction of shock wave propagation paths using a concrete crack growth simulator: ,in is the attenuation coefficient, is the seismic wave frequency, is the initial amplitude, Phase angle, Indicates time.

5. The AI-based emergency shelter site selection assessment system according to claim 1, characterized in that: Ecological buffer effectiveness assessment includes: Biological-geological coupling analysis: , Indicates that the tree has deep roots, Indicates root width, Indicates the inclination angle of the landslide surface.

6. The AI-based emergency shelter site selection assessment system according to claim 5, characterized in that: Ecological buffer effectiveness assessment also includes: Soil moisture monitoring: The soil moisture content is inverted by synthetic aperture radar (SAR). When the moisture content is >28% and the continuous rainfall is >3 hours, the ecological buffer efficiency weight is automatically increased to 12%.

7. The AI-based emergency shelter site selection assessment system according to claim 1, characterized in that: The dynamic rating generation module also includes: Psychological Panic Modifier: Calculate regional panic index through social media sentiment analysis API: RPI=0.6×(negative keyword frequency)+0.4×(exclamation mark density).

8. The AI-based emergency shelter site selection assessment system according to claim 1, characterized in that: The specific implementation of the dynamic weight allocation and score generation module includes: S100, weight dynamic adjustment algorithm: Core layer weight update: When the regional population density is >200% and the duration is ≥5 minutes, the spatial safety weight is adjusted to 35%. If the regional population density is ≤150%, the spatial safety weight is adjusted to 20%.

9. The AI-based emergency shelter site selection assessment system according to claim 3, characterized in that: Gas diffusion model implementation standards: Toxic gas sensors are deployed every 50 meters in underground pipelines, with an installation depth of ≥2 meters. When three consecutive sensors detect v>5m / s, an early warning is pushed.

10. The AI-based emergency shelter site selection assessment system according to claim 1, characterized in that: Fiber optic sensor deployment rules include: Commercial buildings: node density is 70 / km², installation height is 4.5-5.0m; Residential buildings: node density is 50 / km², installation height is 3.0-3.5m; Industrial area buildings: The node density is 60 / km² and the installation height is 6.0-8.0m.

Images