Earthquake disaster data driven multi-state emergency assessment state generation method and system

Abstract

The application discloses a method and system for generating an emergency evaluation state of a seismic disaster data driving polymorphism, and relates to the technical field of emergency management and early warning. The method comprises the following steps: obtaining initial disaster evaluation result data; generating credibility data and performing threshold comparison, and marking the data meeting the conditions as to-be-confirmed evaluation data; gathering continuous to-be-confirmed data in a spatial region as a dimension to form regional evaluation sequence data; performing stability analysis on the sequence to generate evaluation stability data, and converting the data meeting the conditions into confirmed evaluation data through threshold comparison; performing trend fitting on the sequence corresponding to the confirmed data to generate evolution direction data and an emergency level dynamic adjustment instruction; and generating final emergency state data by comprehensively combining the confirmed data and the adjustment instruction. The application effectively suppresses false alarms and response instruction swings by constructing a seismic emergency evaluation digital twin, and improves the robustness and reliability of emergency decision-making.

Description

Technical Field

[0001] This invention relates to the field of emergency management and early warning technology, specifically to a method and system for generating multi-state emergency assessment status driven by earthquake disaster data. Background Technology

[0002] In the early stages following an earthquake disaster, rapid and accurate disaster assessment is crucial for initiating emergency response and allocating rescue resources. With the development of sensing technology, driven by multi-source sensing data, crowdsourced information, and seismic motion monitoring data, earthquake assessment has gradually evolved from traditional on-site surveys towards automation and real-time processing, greatly improving the efficiency of emergency response.

[0003] Existing technologies primarily acquire static or instantaneous dynamic data such as epicenter location, magnitude, and surrounding population density, using attenuation models or rapid damage assessment algorithms to generate preliminary disaster assessment results, which then serve as the basis for triggering different levels of emergency response. These methods can achieve relatively quick preliminary assessments in routine earthquake disasters, possessing a certain degree of timeliness. However, in the initial stages of a strong earthquake, disaster data often exhibits extreme volatility and high uncertainty. Due to complex physical interference at the earthquake site, as well as objective factors such as damaged sensing terminals and communication congestion, the collected disaster data stream is often accompanied by numerous false anomalies or logically contradictory noise, causing the initial assessment results to fluctuate violently within a short period. This "non-steady-state" fluctuation in assessment results leads to frequent false alarms or repeated level jumps in the emergency response decision-making system, preventing rescue resources from accurately pinpointing the true disaster location during the critical "golden disaster relief period" due to wavering decision-making.

[0004] Existing technologies fail to adequately verify the state of earthquake data streams across time, making it difficult to effectively filter out false signals and identify the true evolution trend of the disaster situation amidst highly dynamic data changes. This can easily lead to blind or delayed emergency resource allocation, increasing the risk of secondary disaster prevention failures and low rescue efficiency. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for generating multi-state emergency assessment states driven by earthquake disaster data.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] In a first aspect, this invention discloses a method for generating multi-state emergency assessment states driven by earthquake disaster data, comprising the following steps:

[0008] Obtain spatial data of the disaster area and at least one initial disaster assessment result;

[0009] Based on each initial disaster assessment result data, a corresponding disaster assessment credibility data is generated, and the credibility data is compared with a first preset threshold to generate a first execution rule; the first execution rule is used to mark initial disaster assessment result data that are greater than the first preset threshold as assessment data to be confirmed.

[0010] Using the disaster-affected spatial region as a dimension, the continuous assessment data to be confirmed are collected in chronological order to form a regional assessment sequence data.

[0011] The volatility of the disaster level is measured in the regional assessment sequence data within a preset time window to generate assessment stability data that characterizes the smoothness of the change in disaster level. The assessment stability data is then compared with a second preset threshold to generate a second execution rule. The second execution rule is used to convert the assessment data to be confirmed that is greater than the second preset threshold into confirmed assessment data.

[0012] Based on the regional assessment sequence data corresponding to the confirmed assessment data, trend fitting is performed to generate disaster evolution direction data, and forward-looking adjustments are made according to the disaster evolution direction data to generate emergency level dynamic adjustment instructions.

[0013] The final emergency status data for the current moment is generated by combining the confirmed assessment data and the emergency level dynamic adjustment instructions.

[0014] Secondly, this invention discloses a method and system for generating multi-state emergency assessment states driven by earthquake disaster data. The method, using the aforementioned earthquake disaster data-driven multi-state emergency assessment state generation method, includes:

[0015] The data processing module is used to acquire data on the spatial area of ​​the disaster and the initial disaster assessment results;

[0016] The first state transition module is used to generate corresponding disaster assessment credibility data based on each initial disaster assessment result data, and compare the credibility data with a first preset threshold to generate a first execution rule; the first execution rule is used to mark initial disaster assessment result data that are greater than the first preset threshold as assessment data to be confirmed.

[0017] The time-series construction module is used to collect the continuous assessment data to be confirmed in chronological order, taking the disaster spatial region as the dimension, to form regional assessment sequence data;

[0018] The second state transition module is used to measure the volatility of the disaster level of the regional assessment sequence data within a preset time window, generate assessment stability data that characterizes the smoothness of the change in disaster level, and compare the assessment stability data with a second preset threshold to generate a second execution rule; the second execution rule is used to convert the assessment data to be confirmed that is greater than the second preset threshold into confirmed assessment data.

[0019] The trend analysis and decision-making module is used to perform trend fitting based on the regional assessment sequence data corresponding to the confirmed assessment data, generate disaster evolution direction data, and make forward-looking adjustments based on the disaster evolution direction data to generate emergency level dynamic adjustment instructions.

[0020] The integrated status output module is used to generate the final emergency status data for the current moment based on the confirmed assessment data and the emergency level dynamic adjustment instructions.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] 1. This invention can pre-screen data used for emergency decision-making when disaster assessment results come from diverse sources and have uneven quality, making the data basis for subsequent emergency status judgments more stable and helping to reduce misjudgments of emergency status caused by occasional assessment biases.

[0023] 2. When the disaster assessment results fluctuate frequently in a short period of time, the present invention can suppress the emergency state switching caused by instantaneous changes, so that the final emergency state is more in line with the actual evolution characteristics of the disaster in the spatial area.

[0024] 3. This invention can dynamically update the emergency status in a reasonable manner at different stages of the continuous development or gradual mitigation of the disaster, so that the final emergency status is consistent in time sequence and sustainable at the decision-making level, which is conducive to improving the overall coordination and stability of the emergency response process. Attached Figure Description

[0025] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0026] Figure 1 This is a flowchart of the steps of the present invention;

[0027] Figure 2 This is a schematic diagram illustrating the working principle of the present invention;

[0028] Figure 3 This is a system module connection diagram of the present invention;

[0029] Figure 4 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0030] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0031] In existing earthquake emergency management technologies, disaster assessment results are often derived from rapid calculations at different times and from different data sources, resulting in phased and fluctuating outcomes. Current solutions typically trigger emergency responses directly based on single or the latest disaster assessment results. While this improves response speed, it also makes the assessment results susceptible to incomplete data coverage, quality differences, or short-term anomalies before disaster data has fully converged, leading to frequent changes in emergency levels. Especially in the initial stages of an earthquake or during periods of frequent aftershocks, disaster assessment results are repeatedly revised within a short period. Existing technologies struggle to distinguish which assessment results are sufficiently reliable for emergency decision-making, thus affecting the stability and rationality of emergency status generation.

[0032] Further analysis revealed that the reliability of disaster assessment results is not only related to the numerical values ​​themselves, but also closely related to the quality and spatial consistency of the data upon which the assessment results were generated. When assessment results from different time points and adjacent areas show high consistency in both space and time, their reliability for emergency situation judgment is relatively high; conversely, if the assessment results lack supporting data or differ significantly from assessments in surrounding areas, they are more prone to misjudgment. Therefore, relying solely on the instantaneous values ​​of disaster assessment results for judgment is insufficient to fully reflect their applicability in emergency decision-making.

[0033] The core of this application lies in constructing a "lightweight digital twin" for earthquake emergency response processes. This digital twin, within a virtual information space, defines multiple assessment states—including "initial state," "pending confirmation state," "confirmed state," and "final emergency state"—and their transition rules. It then performs real-time mapping, cleaning, simulation, and extrapolation of dynamic and uncertain disaster assessment data from the physical world. Through a cascaded triggering mechanism of "credibility threshold" and "stability threshold," it ensures that only high-fidelity mirror data can drive decision-making, ultimately achieving "virtual control of the real," generating stable, reliable, and forward-looking emergency state instructions, thereby effectively suppressing decision-making instability.

[0034] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0035] Example 1:

[0036] like Figure 1 As shown, the earthquake disaster data-driven method for generating multi-state emergency assessment states, by constructing a digital twin model of the earthquake emergency assessment process, realizes the transformation and decision simulation of multi-state assessment states in virtual space, including the following steps:

[0037] Obtain spatial data of the disaster area and at least one initial disaster assessment result;

[0038] Based on each initial disaster assessment result, a corresponding disaster assessment credibility data is generated, and the credibility data is compared with a first preset threshold to generate a first execution rule; the first execution rule is used to mark initial disaster assessment result data that are greater than the first preset threshold as assessment data to be confirmed.

[0039] Using the spatial region of the disaster as a dimension, continuous assessment data to be confirmed are collected in chronological order to form a regional assessment sequence data;

[0040] The volatility of the disaster level is measured in the regional assessment sequence data within a preset time window to generate assessment stability data that characterizes the smoothness of the change in disaster level. The assessment stability data is then compared with a second preset threshold to generate a second execution rule. The second execution rule is used to convert the unconfirmed assessment data that exceeds the second preset threshold into confirmed assessment data.

[0041] Based on the regional assessment sequence data corresponding to the confirmed assessment data, trend fitting is performed to generate disaster evolution direction data, and forward-looking adjustments are made according to the disaster evolution direction data to generate dynamic adjustment instructions for emergency level.

[0042] The final emergency status data for the current moment is generated by combining the confirmed assessment data and the dynamic adjustment instructions for the emergency level.

[0043] like Figure 2 The diagram illustrates the working principle of this application. In the virtual mapping layer of the digital twin, the system defines four core states: "Initial Evaluation State," "Pending Confirmation State," "Confirmed State," and "Final Emergency State." For data to advance from the "Initial State" to the "Pending Confirmation State," it must cross a "credibility threshold" based on data source quality and spatial consistency. This step addresses the mirror calibration problem of the reliability of the physical world data source within the digital thread. Subsequently, for data to advance from the "Pending Confirmation State" to the "Confirmed State," it must cross a "stability threshold" based on short-term series volatility. This step addresses the smoothing and confirmation problem of instantaneous changes in the physical world's disaster representation within the virtual space. Only data that successfully completes these two state advancements is considered high-fidelity mirror data and permitted for subsequent trend extrapolation and decision simulation.

[0044] This application first takes the spatial area of ​​disaster after an earthquake as the basic object of analysis. The spatial area of ​​disaster can be determined based on administrative divisions, grid divisions, or the impact range of the epicenter. On this basis, at least one initial disaster assessment result data is obtained. The initial disaster assessment result data can come from an automated assessment system, a rapid manual assessment, or other disaster assessment channels, and its content should at least contain information that can characterize the disaster level or severity.

[0045] After obtaining the initial disaster assessment results, corresponding disaster assessment credibility data is generated for each initial disaster assessment result. Credibility data reflects the reliability of that assessment result in generating an emergency response at the current moment. Subsequently, the credibility data is compared with a pre-set first threshold, and a first execution rule is generated accordingly. When the credibility data meets the threshold condition, the corresponding initial disaster assessment result is marked as assessment data to be confirmed.

[0046] After screening the data to be confirmed, the data belonging to the same region are grouped in chronological order based on the spatial region of the disaster, forming a regional assessment sequence data. This regional assessment sequence data reflects the continuous changes in disaster assessment results over time within the same spatial region.

[0047] The volatility of disaster severity levels in regional assessment sequence data is measured within a preset time window to generate assessment stability data characterizing the smoothness of changes in disaster severity levels. This volatility measurement determines whether the disaster assessment results are relatively stable within a certain timeframe. Subsequently, the assessment stability data is compared with a second preset threshold, and a second execution rule is generated accordingly. When the assessment stability data meets the threshold condition, the corresponding assessment data to be confirmed is converted into confirmed assessment data.

[0048] After obtaining the confirmed assessment data, trend fitting is performed based on the corresponding regional assessment sequence data to generate disaster evolution direction data. This disaster evolution direction data reflects the development trend of the disaster over time, including possible scenarios such as worsening, mitigation, or relative stability. Based on this disaster evolution direction data, forward-looking adjustments are made to generate dynamic adjustment instructions for the emergency response level, enabling the emergency response level to be adjusted accordingly with the development of the disaster, rather than remaining fixed.

[0049] Based on the confirmed assessment data and the dynamic adjustment instructions for the emergency level, the final emergency status data for the current moment is generated comprehensively. The final emergency status data is used to characterize the overall state of the current emergency response. It reflects both the confirmed severity of the disaster and the impact of the disaster's evolution trend on the response level, thereby forming an emergency decision-making result that matches the current disaster status.

[0050] The system defines four core states: "Initial Assessment State," "Pending Confirmation State," "Confirmed State," and "Final Emergency State." Data must cross a "reliability threshold" based on data source quality and spatial consistency to advance from the "Initial State" to the "Pending Confirmation State," addressing the issue of data source reliability. Subsequently, data must cross a "stability threshold" based on short-term series volatility to advance from the "Pending Confirmation State," addressing the issue of sudden changes in disaster characterization. Only data that successfully completes these two state advancements is permitted for subsequent trend extrapolation and emergency decision-making. This application establishes emergency state generation based on multi-time- and multi-state constraints, thereby improving the continuity and consistency of emergency state assessment, avoiding frequent changes in emergency levels due to fluctuations in single assessment results, and contributing to improved decision-making rationality and controllability during earthquake emergency response.

[0051] In one embodiment, to improve the stability and reliability of disaster identification results, this application constructs a disaster state sequence and introduces disaster state evolution constraint rules.

[0052] Specifically, the disaster severity levels at each moment within a continuous time window are used to form a disaster severity state sequence. The consistency of the state sequence is checked by a preset disaster severity state transition rule. When a state fluctuation that is significantly inconsistent with the historical trend is detected in a short period of time, the current disaster severity level is kept from changing through a state locking mechanism until the state transition conditions are met for multiple consecutive time windows before the disaster severity level is allowed to be updated. This reduces the impact of sensor noise or instantaneous abnormal data on the disaster severity identification results.

[0053] This application further proposes that the specific steps for generating corresponding disaster assessment credibility data based on each initial disaster assessment result include:

[0054] The system acquires the coverage and quality scores of multi-source data used to generate initial disaster assessment results. Specifically, in determining the reliability of the disaster assessment results, the system first performs in-depth analysis of the underlying multi-source data used to generate the initial disaster assessment results. Multi-source data includes, but is not limited to, seismic accelerometer records, synthetic aperture radar (SAR) satellite imagery, social media text streams, and mobile terminal sensor data. The system establishes basic reliability by calculating the coverage and quality scores of this multi-source data. Coverage reflects the geographical distribution balance of data collection points within the disaster area; for example, by dividing the disaster area into equal-area grids and statistically analyzing the percentage of grids containing active sensors or valid observations. The quality score is based on real-time scoring of signal-to-noise ratio, resolution, or sensor online status.

[0055] To correct for potential systematic biases from a single data source, this application further introduces a spatial consistency check. The system acquires other initial disaster assessment results generated concurrently within a preset range adjacent to the disaster spatial area (e.g., within a 50km radius of the epicenter). By calculating the offset between the current assessment value and the mean or median of surrounding assessment values, a consistency score is obtained for other initial disaster assessment results generated concurrently.

[0056] If the surrounding assessment results show a high degree of clustering, while the current data deviates significantly from the cluster center, its consistency score will be reduced.

[0057] Based on coverage, quality scores, and consistency scores, the system uses a weighted fusion algorithm to calculate and generate the final credibility data. The specific calculation formula is as follows:

[0058] ;

[0059] in, To correspond to the data coverage ( ), data quality score ( ) and spatial consistency score ( The preset weighting coefficients in the credibility fusion calculation, based on historical data fitting, have a value range of [0.1, 0.6], and ;

[0060] The data coverage score represents the area of ​​the data coverage grid. The sum and total area The ratio, ;

[0061] This represents the normalized data quality score. ; The original quality score is calculated in real time based on indicators such as signal-to-noise ratio, resolution, and sensor online status. and These are the minimum and maximum quality scores in the historical data, respectively, used to map the scores to the [0,1] interval;

[0062] The consistency score is characterized by an exponential decay function, where The disaster severity level value is based on the preliminary assessment data. This represents the average of the assessment results from adjacent areas during the same period. This is the spatial variance correction coefficient, with a value range of [value missing]. .

[0063] Calculated credibility data With the first preset threshold Comparison is performed, and the first preset threshold is used. Usually set at Within a certain range, such as in the extremely high noise environment during the initial stage of a strong earthquake, it can be set to... .

[0064] In the above calculation process, the dynamic optimization of weights and the determination of quality scores can be achieved by a pre-trained reliability-weighted evaluation model. The model adopts a deep neural network (DNN) architecture, and the training process is completed on a high-performance computing server. The training set contains 500 real earthquake case data covering magnitudes 4.0-8.0 published by the China Earthquake Administration and their corresponding disaster verification results. The training conditions are set as follows: the initial learning rate is set to 0.001 and the Adam optimizer is used; the batch size is set to 64; and the number of iterations is no less than 200. During the training process, the input vector contains the meta-attributes of multi-source data, spatiotemporal coordinates, and historical sensor failure rates. By minimizing the evaluation bias loss function, the model learns the optimal mapping relationship of the contribution of each data source to the real disaster situation under different earthquake conditions.

[0065] Through the aforementioned technical solution, this application establishes a quantitative review system based on data facts and spatial logic for handling highly uncertain information such as earthquake disaster situations. By decoupling and fusing multi-source data coverage, quality, and spatial consistency in multiple dimensions, the system can objectively exclude off-alert data caused by individual sensor failures, local communication interference, or algorithmic islanding effects. This not only provides refined judgment criteria for screening data to be confirmed and assessed but also endows the system with the ability to adapt to different observation environments through model training. From the underlying logic, this supports the accuracy of subsequent stability analysis and trend fitting, minimizing data blind spots and information noise in emergency response decision-making.

[0066] This application further proposes that, in the dynamic process of earthquake disaster assessment, not all initial data below the first preset threshold are invalid noise. Due to the instability of communication links in the early post-earthquake period, the brief self-checking of damaged sensors, or the asynchronous nature of multi-source data transmission, some disaster data with real reference value may be judged as low reliability in the initial calculation due to "insufficient coverage" or "lagging consistency score". After generating the first execution rule, it also includes:

[0067] When the system identifies data related to credibility Less than the first preset threshold When a certain initial disaster assessment result data is received, the system does not discard it directly. Instead, it triggers alarm suppression logic to generate corresponding state suppression data and store it in the system's circular cache queue. This cache queue adopts a first-in, first-out (FIFO) storage strategy and assigns an initial acquisition timestamp to each piece of data entering the queue. During storage, this initial disaster assessment result data is in a "suspended" state and does not affect the generation of the current emergency status, thus effectively suppressing potential false alarms.

[0068] To enable data self-healing and release, the system executes a periodic secondary verification process. The system operates according to a preset cycle. (For example, extract features from the state suppression data in the cache queue every 10 to 60 seconds) and combine them with the preset period. The system recalculates the reliability data based on the latest acquired surrounding observation data. Over time, previously missing correlation data (such as subsequently uploaded disaster data from adjacent areas) may be gradually supplemented, causing the spatial consistency score to dynamically increase. The system uses the following recalculation trigger judgment to determine the reliability data obtained from the recalculation. The calculation formula is:

[0069] ;

[0070] in, Score the data coverage of the current data;

[0071] This is a score for the normalized data quality of the current data.

[0072] Represents the cycle The set of spatial neighborhood information updated within the internal space;

[0073] Represents a set of information based on spatial neighborhood. The spatial consistency score is calculated using the following formula:

[0074] ;

[0075] in for The mean of the mid-level assessment values;

[0076] For recalculation, regarding data coverage ( ), data quality score ( ) and updated space consistency ( The adjusted weighting coefficients satisfy... The default values ​​are 0.4, 0.3, and 0.3.

[0077] Due to credibility data The recalculation focuses on examining the consistency of the data with the latest environment after the delay; therefore, the weight allocation is adjusted and a set of updated spatial neighborhood information is adopted. The consistency calculation function. When the recalculated confidence data... Reaching or exceeding the first preset threshold At this time, the system executes a release command. This release command removes the corresponding initial disaster assessment result data from the cache queue, removes the suppression flag, officially marks it as assessment data to be confirmed, and sends it into the subsequent regional assessment sequence construction process. If the data remains in the cache queue for more than the lifecycle threshold... If the reliability check still fails after 10 to 30 minutes (usually set), the system will classify it as permanent noise and perform physical deletion.

[0078] By implementing a flexible "suppression-observation-release" processing logic, the robustness and comprehensiveness of disaster assessment are balanced. This allows the system to logically supplement low-quality data caused by asynchronous data processing in the highly uncertain post-earthquake environment through a time-for-space approach, effectively avoiding the risk of missed reports due to incomplete initial data. Simultaneously, the alarm suppression state ensures the system remains silent until the data meets the confidence standard, preventing instantaneous fluctuations from triggering erroneous emergency level adjustments, thereby improving the accuracy of emergency assessment status generation and the reliability of disaster response.

[0079] This application further proposes to measure the volatility of disaster severity levels in regional assessment sequence data within a preset time window, with specific steps including:

[0080] After generating the regional assessment sequence data, the system enters the core stability determination stage, aiming to identify the true stable periods of disaster severity from the fluctuating disaster flow. Since the initial assessment data of an earthquake often exhibits strong non-stationary characteristics, the system uses a sliding time window to capture local temporal features. Continuous disaster severity values ​​within a preset time window are extracted from the regional assessment sequence data. Specifically, the system extracts disaster severity values ​​from the regional assessment sequence within a preset time window of [length missing]. (Threshold adjustment amount) Continuous disaster level value The preset time window length The value is usually set between 2 minutes and 10 minutes, and the step size is set between 30 seconds and 1 minute depending on the data update frequency.

[0081] To quantify this time window The system calculates the moving variance of disaster severity levels to assess the dispersion of disaster severity levels. The moving variance sensitively reflects the amplitude of fluctuations in the assessment results. A smaller variance indicates that assessment conclusions from different data sources or at different times tend to be consistent, and the disaster severity characterization has entered a reliable stage. The calculation formula is as follows:

[0082] ;

[0083] In the formula, This represents the sliding variance within the current window, i.e., the generated stability assessment data.

[0084] Let be the disaster severity level value at the i-th moment within the window;

[0085] This represents the arithmetic mean of the disaster severity levels within the window. In practical applications, the system can also use a moving range. As a supplement or alternative, to simplify computational overhead.

[0086] The system will compare the calculated sliding variance or sliding range with a second preset threshold. Perform real-time comparison. If the sliding variance or sliding range is below a second preset threshold... A continuous period of time is identified as a stable period, and stability assessment data is generated based on this stable period. A second preset threshold... As a criterion for determining a "stable period," its numerical range is usually set according to the earthquake intensity interval. Between. For example, when the sliding variance If the value is below 0.15 for several consecutive periods, the system determines that the consecutive period as a stable period. Once the determination is successful, the system triggers the second execution rule to convert all "pending evaluation data" within that period into confirmed evaluation data in batches.

[0087] Furthermore, this application introduces a threshold adaptive learning model, employing a Long Short-Term Memory (LSTM) network architecture. The training dataset is derived from historical publicly available earthquake databases (such as USGS or historical data from the China Earthquake Networks Center). The training conditions are set as follows: on a GPU cluster, the input consists of historical noise fluctuation sequences over the past 24 hours and expert-annotated stable nodes. The threshold adaptive learning model uses a single hidden layer (128 dimensions), with 150 training epochs and a learning rate of 5×10⁻⁶. -5 The packet loss rate is 0.2%. These parameters were determined through grid search optimization to ensure convergence on a GPU cluster (NVIDIA A100). The model input is a historical noise fluctuation sequence, and the output is a dynamic threshold adjustment.

[0088] Through this threshold adaptive learning model, the system can dynamically adjust the second preset threshold according to the current background noise level of the earthquake. The values ​​are carefully selected to ensure accurate identification of stability windows even in extreme environments with frequent strong earthquakes and aftershocks. The introduction of sliding statistical indicators provides an objective physical criterion for the state advancement of earthquake assessment data, effectively solving the problem of "false upgrades" caused by excessively large peak values ​​in single assessments and reducing decision-making redundancy during emergency response.

[0089] This application further proposes to make forward-looking adjustments based on disaster evolution data to generate dynamic adjustment instructions for emergency levels. Specific steps include:

[0090] After acquiring confirmed assessment data, the system not only focuses on the current absolute values ​​of the disaster situation, but its core logic lies in revealing the rate of evolution of the disaster through time-series modeling. To quantify the data on the direction of disaster evolution, the system performs trend fitting on regional assessment sequence data and processes the confirmed disaster level values ​​using the least squares method or a first-order linear regression model, thereby extracting the slope of disaster change that reflects the development trend of the disaster. As a quantitative indicator of the direction of disaster evolution, the slope of disaster change directly represents the speed at which the disaster worsens or alleviates within a unit of time.

[0091] Specifically, the system establishes a linear regression model within a preset sliding window, the sliding window length being... The value range is 2-10 minutes, with a step size of 30 seconds-1 minute. The formula for calculating the slope of disaster change is as follows:

[0092] ;

[0093] In the formula, k represents the slope of the disaster situation change;

[0094] n is the number of fitted points (usually the 10 to 30 most recent data points are selected);

[0095] Let be the timestamp corresponding to the i-th data point;

[0096] This corresponds to the confirmed disaster severity level value.

[0097] The calculated slope k of the disaster situation change can directly reflect the direction of the disaster's evolution: if This indicates that the disaster is in an escalating and expanding phase; if This indicates that the disaster situation has entered a stable period; if This indicates that the disaster situation is beginning to gradually ease.

[0098] The quantified slope k of the disaster change is compared with a set of predefined level adjustment thresholds. Interval matching is performed. These thresholds are preset based on the evolution of historical strong earthquakes, for example:

[0099] when When the match is "slowly rising interval", a preparatory instruction to finely adjust upward by 1 level is generated;

[0100] when When the match is in the "rapid deterioration zone", a powerful instruction to adjust upwards across levels is generated;

[0101] when When the match is "relief zone", a suggested instruction to adjust the response level downwards is generated.

[0102] Based on the matching results, a dynamic adjustment instruction for the emergency level is generated, containing specific adjustment ranges and directions. For example, the instruction could be expressed as "Based on the current slope of the disaster situation change..." "It is recommended to raise the emergency response level by 2 levels, effective immediately." To improve the accuracy of matching, the threshold range for level adjustment is usually set between [-0.5, 0.5], and dynamically adjusted based on real-time feedback from the post-earthquake rescue response.

[0103] Through the aforementioned technical solution, this application enables the system to detect subtle signals of a rapidly deteriorating disaster situation by quantitatively analyzing the slope of the disaster's evolution. This allows the system to issue "pre-adjustment instructions" before the actual disaster peak arrives. This dynamic adjustment logic based on trend extrapolation provides the emergency command center with invaluable response redundancy time, ensuring that the switching of rescue resource levels closely matches the actual evolution trajectory of the disaster, and preventing the expansion of disaster losses due to delayed response level adjustments.

[0104] This application further proposes to generate the final emergency status data for the current moment based on the confirmed assessment data and the dynamic adjustment instructions for the emergency level. The specific steps include:

[0105] After completing the disaster trend prediction, the system enters the final stage of decision output. This process aims to transform the abstract disaster level into specific, actionable instructions. The current disaster level contained in the confirmed assessment data is used as the baseline response level. This baseline response level represents the objective status quo of damage already caused in the disaster area and is typically divided into four benchmark dimensions: extremely severe, severe, relatively severe, and general.

[0106] Subsequently, the system performs an overlay correction calculation, applying the adjustment range included in the emergency level dynamic adjustment command. (e.g., +1 or -1) and direction, acting on the base response level. The target response level at the current moment is obtained through discretization mapping calculation. The computational logic follows the following mapping relationship:

[0107] ;

[0108] in, This is the direction coefficient (1 for upward and -1 for downward).

[0109] This is a level smoothing function used to handle non-integer level rounding logic;

[0110] This is a limiting function to ensure that the calculated level is within the preset upper limit of the emergency response level. and lower limit Within the range. For example, if the current basic response level is "major (Level II)", and the dynamic adjustment instruction suggests that it needs to be upgraded by 1 level due to the rapid deterioration of the disaster, then the calculated target response level is "extremely serious (Level I)".

[0111] After obtaining the target response level, the system invokes the core emergency response plan library. This library is not simply a collection of static documents, but rather a set of digitized action plans composed of structured data. The system matches the target response level against the preset emergency response plan library, using the target response level as the primary key and combining it with specific attributes of the disaster area (such as geographical environment, population density, and accessibility) to retrieve the optimal matching template from the library.

[0112] During the matching process, the system employs semantic association and attribute projection technologies to automatically extract specific action items (such as dispatching helicopters for search and rescue, activating power supply guarantees in the earthquake zone, and issuing evacuation guidance) that are appropriate for the current level, responsible entities (such as provincial emergency management departments, armed police forces, and power departments), and strict time nodes (such as responding within 2 hours and completing the first round of life screening within 12 hours). These elements are logically reorganized to comprehensively generate final emergency status data that includes specific action items, responsible entities, and time nodes, and are pushed to the joint command terminals at all levels in a standard format.

[0113] To ensure the scientific rigor of contingency plan matching, the system's built-in matching algorithm underwent extensive simulated stress training. Training parameters included: inputting over 1000 sets of extreme disaster scenarios in an offline environment; setting a matching loss function to optimize the fit between actions and disaster situations; and employing a reinforcement learning framework, using simulated rescue efficiency as a reward value, and training through no fewer than 50,000 rounds of agent-based game theory. This enabled the system to select the optimal solution from tens of thousands of action rules within milliseconds.

[0114] Through the aforementioned technical solution, this application combines static basic response levels with dynamic trend adjustment instructions to generate final emergency status data that not only addresses the current disaster situation but also anticipates future risks. This refined output, including tasks, responsible persons, and timelines, significantly reduces the information gap between the command and execution levels, ensuring that emergency actions remain efficient and precise even in the highly stressful post-earthquake environment, truly achieving a closed-loop multi-state emergency assessment driven by the earthquake disaster.

[0115] This application further proposes matching the target response level with a pre-set emergency response plan database, specifically including:

[0116] The system first retrieves current emergency resource status data through a real-time interface. This data includes the location coordinates of rescue teams, the available quantity of large equipment (such as life detectors and demolition tools), and the remaining supplies in surrounding depots. Simultaneously, the system extracts characteristic data of the disaster area, including terrain complexity (such as mountainous or plain areas), population and building density, and current real-time weather constraints (such as rainfall, strong winds, and other factors that limit air rescue).

[0117] The system performs multi-constraint coupled calculations on the target response level, the characteristic data of the disaster spatial area, and the current emergency resource status data;

[0118] When a perfectly matching plan exists in the emergency response plan library (i.e. a plan that is specific to a particular level, specific terrain, and whose resource requirements are fully met), the system executes the direct matching path and directly generates the action item corresponding to the plan.

[0119] When no perfectly matching contingency plan exists, the system initiates a "disassemble-reassemble" logic. Based on the coupled calculation results, multiple similar contingency plans are decomposed into elements, breaking them down into the smallest granularity atomic action items (e.g., road clearing, medical transport, tent erection). Subsequently, the system weights and merges the decomposed elements according to the priority of multiple constraints, constructing a multi-constraint evaluation model. Weight matching calculations are then performed on these atomic elements to generate specific action items adapted to the current multi-constraint conditions. The specific calculation formula is as follows:

[0120] ;

[0121] in, The fitting score for a given atomic action item;

[0122] m is the number of constraints;

[0123] The weighting factor for the j-th constraint (e.g., the weight for life search and rescue is set to 0.6, and the weight for material distribution is set to 0.4). The value range is usually in Between, and satisfy ;

[0124] The fitness function takes available resources as its input variable. and environmental characteristics .

[0125] Based on calculations The ranking system ranks and weights the extracted elements according to preset priorities (such as the "life safety first" principle). If the calculation results show that air transport scores too low due to weather constraints, while land transport scores higher, the system will automatically extract elements from the land rescue plan to replace the original air rescue part, thereby generating specific action items adapted to the current multiple constraints. This reorganization logic is supported by a generative adversarial network (GAN) algorithm. The model was trained using no fewer than 10,000 non-standard disaster scenarios in a computing environment with two NVIDIA A100 GPUs, with an initial learning rate of... The discriminator continuously verifies the resource feasibility and logical coherence of the reorganization plan until the action items output by the generator satisfy all current boundary constraints.

[0126] Through the aforementioned steps in generating final emergency status data, this application, by atomically decomposing plan elements and weighted merging based on multiple constraints, enables the system to automatically "tailor" the execution plan to best suit current combat conditions in a very short time in response to sudden resource shortages or environmental changes. This not only reduces the decision-making burden on commanders under extreme pressure but also ensures a high degree of coupling between rescue action items and existing resources through algorithms, achieving a precise transformation from earthquake emergency assessment status to efficient rescue operations from a technical perspective.

[0127] This application further proposes that the method also includes continuous monitoring and state rollback, with specific steps including:

[0128] Emergency management does not end after the issuance of instructions. After outputting the final emergency status data, the system immediately enters a closed-loop continuous monitoring phase. The system utilizes a real-time data stream interface to continuously acquire newly generated initial disaster assessment results within the disaster area. This process aims to identify whether the disaster has passed its peak and entered a receding phase, thus providing a basis for the orderly withdrawal of rescue forces or the scientific downgrading of the response level.

[0129] The state rollback logic is triggered based on a dual assessment of "stability" and "trend". First, the system recalculates the stability assessment data (such as sliding variance) based on the newly acquired data. If the stability assessment data generated based on the new initial disaster assessment results remains below a third preset threshold... This indicates that the energy of the disaster fluctuations has significantly decreased, and the assessment results have entered a period of low-level stability. If this stability data continues to be below the third preset threshold, the third preset threshold... Usually set at Within a certain range, for example, a value of 0.05, the standard is stricter than the second preset threshold. This is to ensure that the rollback operation is based on the premise that the disaster situation is extremely stable.

[0130] Simultaneously, the system combines a trend fitting model to obtain data on the current direction of disaster evolution. If the slope k of the disaster change remains negative (e.g.) And it meets the preset duration window. If the duration is more than 30 minutes, it is determined that the disaster situation is showing a clear trend of easing.

[0131] When the assessment stability data generated based on the new initial disaster assessment results remains below the third preset threshold Furthermore, when the disaster evolution data indicates that the disaster is showing a mitigating trend, the system automatically generates a status rollback command.

[0132] State rollback is not instantaneous, but follows a controlled, step-by-step path. Based on the state rollback instruction, the system calculates the new level using the following logic:

[0133] ;

[0134] in, The next response level after the rollback;

[0135] The current response level;

[0136] The rollback step size is usually set to the increment of the level value (e.g., rollback from level I to level II).

[0137] This is the security target level fitted based on the current real-time data.

[0138] Based on the status rollback instructions, the emergency status data will gradually decrease from the current high response level. At each step of the decrease, the system will re-execute a 10- to 20-minute observation period to confirm that the disaster situation has not rebounded (i.e., there is no sudden increase in secondary disasters caused by aftershocks) before continuing to decrease the level until the final "relief" status is achieved.

[0139] By introducing a third preset threshold The stringent steady-state determination solves the problems of "easy to escalate but difficult to de-escalate" or "blind withdrawal" in emergency response. It ensures that the rollback operation is based on objective feedback from technical facts, rather than subjective assumptions. Through gradient-based downsizing logic, the system can effectively prevent secondary damage that may be caused by earthquake aftershocks, realizing the optimal exit strategy for emergency response resources in the time dimension, which not only ensures the safety of the disaster area, but also avoids the ineffective waste of social resources.

[0140] This application further proposes that, in order to enable the system to cope with fluctuations in data source quality under different geographical locations and seismic environments during medium- and long-term operation, an adaptive threshold adjustment mechanism based on feedback loops is introduced. The adaptive threshold adjustment steps specifically include:

[0141] The system establishes a continuously running monitoring thread to record credibility data, final emergency status data, and effectiveness assessment results provided by experts or on-site verification within a historical period (e.g., the past 30 days or the past 10 earthquake events).

[0142] The system assesses the rationality of the current threshold setting by monitoring the correlation between the reliability data and the validity feedback of the final emergency status data generated in the historical period.

[0143] Based on the correlation, the values ​​of the first and second preset thresholds are dynamically adjusted so that the strictness of the state transition is adapted to the overall quality of the data source being evaluated.

[0144] Specifically, the system uses the Pearson correlation coefficient or mutual information entropy to quantify the degree of matching between the reliability prediction and the actual disaster situation. If monitoring detects that the false alarm rate of the final emergency state increases during a period of declining quality of a specific data source, it indicates that the current first preset threshold is being met. Or the second preset threshold It was too loose and failed to effectively block noise.

[0145] Based on this correlation, the system corrects the threshold by executing a dynamic gain adjustment algorithm. Its adjustment logic follows these rules:

[0146] ;

[0147] in, The correction increment for the preset threshold (which can be applied to the first preset threshold) Or the second preset threshold );

[0148] This is the learning rate factor, and its value typically ranges from [value range missing]. Used to control the smoothness of the adjustment;

[0149] This is the actual disaster effectiveness value verified on-site;

[0150] This is the performance value predicted by the system based on the current threshold;

[0151] This represents the maximum theoretical performance value set by the system, used to normalize performance differences, and is usually taken as... (Corresponding to the ideal state of a perfect match);

[0152] This is the overall quality evaluation coefficient for the data source, calculated by combining the signal-to-noise ratio and completeness.

[0153] This formula enables the system to achieve self-adaptation of the threshold: when evaluating the overall quality of the data source. When the signal weakens (e.g., due to large-scale damage to base stations in the earthquake zone causing signal instability), the system will automatically raise the first preset threshold. (For example, adjusting from 0.70 to 0.80) to enhance anti-interference capabilities. Conversely, when the data source quality is robust and highly consistent, the system can appropriately lower the threshold to improve response sensitivity.

[0154] To ensure the scientific validity of this adaptive adjustment, the system incorporates a threshold decision model based on reinforcement learning. During the training phase, this model utilizes historical disaster samples under multivariate random noise perturbations, and the training environment is set up in a distributed cluster with high concurrency processing capabilities. Model training conditions include: setting the state space (including the current threshold, data quality distribution, and feedback error), the action space (adjusting, lowering, or maintaining the threshold), and the reward function (using a combined score of response time and accuracy as the reward function). Through at least [number missing]... The sampling training allows the model to automatically optimize in complex perceptual environments and find the threshold balance point that maximizes the overall system performance.

[0155] To address real-time data latency or communication interruptions, the system incorporates a fault-tolerance mechanism:

[0156] Data timeout retransmission: If the initial disaster assessment results are received within the preset time window... If the data is not fully received within 30 seconds, a retransmission request is triggered, and some data is temporarily stored in the cache queue.

[0157] Degradation handling: When communication interruption lasts for more than 5 minutes, the system automatically switches to local cached data mode, uses the most recently confirmed assessment data to generate an emergency status, and issues an alarm.

[0158] Consistency check: Retransmitted data must be made complete by adding a checksum (such as CRC check), otherwise it will be discarded.

[0159] Through the above technical solution, this application breaks through the limitations of traditional systems that rely on manual experience to set fixed parameters. By using data-driven closed-loop feedback, the system can perceive the quality of the evaluation environment and automatically adjust the "review scale." This adaptive capability enhances the system's survivability in extremely complex environments, ensuring that the final emergency status data generated is always within the optimal reliability range under the current data conditions, regardless of whether the data is abundant or scarce.

[0160] The following is a complete and practical example of a data-driven method for generating multi-state emergency assessment states:

[0161] A magnitude 6.5 earthquake struck a region, with its epicenter located at 31.2°N, 103.5°E, and a focal depth of 15 kilometers. The system received real-time access to multi-source disaster data streams, including P-wave triggered data from seismic networks, synthetic aperture radar (SAR) satellite imagery, disaster-related text streams from social media, and peak acceleration data uploaded by mobile terminal sensor clusters in the earthquake zone. Based on an attenuation model and a population density grid, the system generated the first batch of 15 initial disaster assessment results within 30 seconds of the earthquake, with assessment levels ranging from VI to IX.

[0162] The system initiates the credibility calculation process. For one data point with an evaluation result of "VIII", its underlying multi-source data features are extracted: SAR image coverage is 85% (total area 1200 km²). 2 The effective grid is 1020km. 2 The grid was divided with a precision of 1km × 1km, and the quality score was Q = 0.78 (based on image sharpness and distortion correction results). Simultaneously, seven evaluation results within a 50km radius were acquired, with a consistency score of 0.82 (the current value of VIII deviates little from the neighborhood mean of VII.5 degrees). A weighted fusion formula was used to calculate the reliability. Set the first preset threshold , This data point was marked as "data to be confirmed and evaluated". Three other data points had their reliability reduced due to insufficient SAR image coverage (<60%) or significant differences from neighboring assessments (|Δ|>1.5). The values ​​were 0.52, 0.48, and 0.61 respectively, all lower than... It is stored in the cache queue and its state is suppressed.

[0163] Using a 50km radius around the epicenter as the disaster spatial area, the system aggregates all "pending confirmation assessment data" within a consecutive 5-minute period using second-level timestamps, forming a regional assessment sequence data, containing a total of 42 assessment records. The disaster level sequence is: VII, VII, VIII, VIII, VIII, VIII, ... (time window) (Step length 30s).

[0164] The volatility of the disaster severity level was measured in the above sequence, and the severity level values ​​at six consecutive time points were extracted: (Numerized, degree VIII corresponds to 8). Calculate the sliding variance. Sliding variance Below the second preset threshold The system determines that the time period is a "stable period" and converts all "data to be confirmed" within that time period into "data to be confirmed" in batches.

[0165] Based on the regional assessment sequence corresponding to the confirmed data, the system performs trend fitting (linear regression, most recent 20 points) and calculates the slope of disaster change. , The indication is that the disaster situation is still showing a slow upward trend. The threshold range is adjusted according to the preset level. In response to the “slightly rising range”, the system makes a forward-looking adjustment and generates a dynamic adjustment instruction for the emergency level: “It is recommended to adjust the emergency response level from the current basic level VIII, which corresponds to level II response, upward by 0.5 levels, and prepare to activate the cross-regional support mechanism.”

[0166] The system acquires current emergency resource status data: 8 available rescue teams, 25 life detectors, and 12 drones in the earthquake zone. Combined with spatial regional characteristics data of the disaster (70% mountainous area, moderate population density), it performs multi-constraint coupled calculations. No fully matching "Level II+0.5, Mountainous Area, Road Disruption" plan was found in the emergency response plan database. The system then initiates plan decomposition and reorganization logic: extracting "helicopter delivery of supplies" and "foot search and rescue team organization" from the "Level II Mountain Rescue Plan," and "inter-provincial medical team dispatch" from the "Level II+ Urban Plan." The system then... Scoring is used to select the highest-scoring atomic action items and combine them to generate specific action items that fit the current constraints, including:

[0167] ①Use helicopters to deliver demolition tools to the northeast area (arriving within 2 hours);

[0168] ② Form 3 foot search and rescue teams, each equipped with 2 life detectors (to be assembled within 1.5 hours);

[0169] ③ Coordinate with medical teams from neighboring provinces to stand by, and decide whether to dispatch them depending on the development of the disaster situation.

[0170] Based on the confirmed assessment data (current disaster level VIII) and the emergency level dynamic adjustment instruction (upgraded by 0.5 levels), the system generates the final emergency status data, with the following output: Response level: II+ (ready to upgrade); list of action items; responsible entity; time nodes: deployment 2h, assembly 1.5h, medical standby in real time.

[0171] The system continues to monitor new data. Two hours after the earthquake, newly added assessment data showed that the disaster level had stabilized at degree VIII, with a moving variance of [missing data]. Continuously below the third preset threshold Furthermore, the trend slope k turned -0.03, indicating that the disaster had entered a mitigation phase. The system generated a status rollback instruction, gradually reducing the response level from Level II+ to Level II according to a tiered logic, while adjusting the action item to "gradually reduce airdrops and switch to ground supply distribution." After one week of system operation, the system automatically adjusted the first preset threshold based on feedback data. Adjusted from 0.70 to 0.73.

[0172] This embodiment demonstrates that in the very early stages after an earthquake, the system achieves a closed-loop processing chain, encompassing multi-source heterogeneous data access, dynamic credibility verification, temporal stability assessment, trend extrapolation and early warning, and structured generation and dynamic adjustment of emergency states. It effectively suppresses assessment jumps caused by local distortions in SAR images or misinformation on social media, avoids frequent switching of response levels before the disaster situation is clear through stability criteria, provides a critical time window for the deployment of rescue forces based on trend fitting, and ensures that executable action plans can still be generated under unforeseen constraints such as road closures. Practical application shows that the system can stably generate emergency states within 30 minutes after an earthquake, reducing the frequency of response level switching by approximately 60% compared to traditional systems.

[0173] Example 2:

[0174] like Figure 3 and Figure 4 As shown, the earthquake disaster data-driven multi-state emergency assessment state generation system uses the aforementioned earthquake disaster data-driven multi-state emergency assessment state generation method, including:

[0175] The data processing module is used to acquire data on the spatial area of ​​the disaster and the initial disaster assessment results;

[0176] The first state transition module is used to generate corresponding disaster assessment credibility data based on each initial disaster assessment result data, and compare the credibility data with a first preset threshold to generate a first execution rule; the first execution rule is used to mark initial disaster assessment result data that are greater than the first preset threshold as assessment data to be confirmed.

[0177] The time series construction module is used to collect continuous assessment data to be confirmed in chronological order, based on the spatial region of the disaster, to form regional assessment sequence data;

[0178] The second state transition module is used to measure the volatility of the disaster level of the regional assessment sequence data within a preset time window, generate assessment stability data that characterizes the smoothness of the change in disaster level, and compare the assessment stability data with a second preset threshold to generate a second execution rule; the second execution rule is used to convert the unconfirmed assessment data that is greater than the second preset threshold into confirmed assessment data.

[0179] The trend analysis and decision-making module is used to fit trends based on the regional assessment sequence data corresponding to the confirmed assessment data, generate disaster evolution direction data, and make forward-looking adjustments based on the disaster evolution direction data to generate dynamic adjustment instructions for emergency levels.

[0180] The integrated status output module is used to generate the final emergency status data for the current moment based on the confirmed assessment data and the emergency level dynamic adjustment instructions.

[0181] The hardware components of this application mainly include: a data acquisition and sensing array, consisting of strong seismometers, inclinometers, and meteorological monitoring stations deployed in various locations, responsible for capturing raw physical signals; a high-speed data transmission link, based on satellite communication and a 5G private network, to ensure data transmission under extreme post-earthquake conditions; a central processing server cluster, equipped with high-performance computing GPUs and high-bandwidth memory, used to run complex stability analysis algorithms and trend fitting models; and a command and dispatch terminal, used for the visualization of final emergency status data and the issuance of action instructions.

[0182] The various modules within the system interact logically via a standardized data bus. The data processing module normalizes the heterogeneous data transmitted from the acquisition array using a pre-defined data cleaning algorithm, eliminating obvious conflicts in physical dimensions. After obtaining the initial evaluation results, the first state transition module invokes the aforementioned reliability-weighted evaluation model. During execution, this module, through a hardware acceleration engine, completes the coverage and quality scoring of massive data sources within milliseconds and pushes qualified data into the time-series buffer according to the first execution rule.

[0183] The time-series construction module maintains a dynamically indexed geospatial database in memory, using the disaster-affected area as the key to concatenate the data to be confirmed at high frequency. The second state transition module periodically polls the buffer, utilizing the parallel computing power of the server cluster to perform sliding variance calculations; the resulting stability data determines the "promotion" of the data in memory. The trend analysis and decision-making module retrieves confirmed data from historical windows and performs multi-order polynomial fitting to quantify the slope characteristics of the disaster's evolution. Finally, the comprehensive state output module associates the calculated dynamic adjustment instructions with a structured emergency response plan library, sending the final emergency status, containing specific task packages, to rescue units at all levels through an output interface.

[0184] In one embodiment, taking a magnitude 6.8 earthquake in a certain region as an example, 20 seconds after the earthquake, the data processing module acquired 15 initial assessment data points from satellite remote sensing and ground monitoring stations. The first state transition module calculated and found that 3 of these data points had a low signal-to-noise ratio due to communication interference, thus affecting their reliability. It is only 0.45, which is lower than the first preset threshold. The first 12 data entries were then stored in the alarm suppression queue; the remaining 12 data entries were marked as "pending confirmation." The time series construction module arranged these 12 data entries at the second level, forming a 5-minute sequence. The second state transition module calculated the sliding variance of this sequence as follows: It is lower than the second preset threshold. Once the period is determined to be stable, the data is converted to "confirmed." Subsequently, the trend analysis and decision-making module fits the disaster slope as follows: The system indicated that the disaster was spreading rapidly and generated an instruction to "upgrade the response level by 1 level". Based on this, the integrated status output module adjusted the basic Level II response to Level I and reorganized the data from the emergency response plan library to generate the final emergency status data, including "dispatching the heavy air rescue team" and "opening the Level II shelter", successfully achieving closed-loop management of the disaster.

[0185] This system achieves automation and precision across the entire earthquake emergency assessment process through the synergistic coupling of hardware architecture and software logic. The distributed deployment of the hardware ensures the system's survivability and high-concurrency processing capabilities under strong earthquake impacts; the modular software architecture, through polymorphic transformation logic, effectively solves the false alarm and delayed reporting problems caused by inconsistent data quality and lagging assessment logic in traditional systems. The system can sift through massive amounts of fragmented post-earthquake data, converting dynamically changing disaster slopes into precise action instructions, improving the timeliness of emergency response and the adaptability of contingency plan execution, and providing a reliable system platform to minimize casualties and property losses.

[0186] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A method for generating multi-state emergency assessment states driven by earthquake disaster data, characterized by: Includes the following steps: Obtain spatial data of the disaster area and at least one initial disaster assessment result; Based on each initial disaster assessment result data, a corresponding disaster assessment credibility data is generated, and the credibility data is compared with a first preset threshold to generate a first execution rule; the first execution rule is used to mark initial disaster assessment result data that are greater than the first preset threshold as assessment data to be confirmed. Using the disaster-affected spatial region as a dimension, the continuous assessment data to be confirmed are collected in chronological order to form a regional assessment sequence data. The volatility of the disaster level is measured in the regional assessment sequence data within a preset time window to generate assessment stability data that characterizes the smoothness of the change in disaster level. The assessment stability data is then compared with a second preset threshold to generate a second execution rule. The second execution rule is used to convert the assessment data to be confirmed that is greater than the second preset threshold into confirmed assessment data. Based on the regional assessment sequence data corresponding to the confirmed assessment data, trend fitting is performed to generate disaster evolution direction data, and forward-looking adjustments are made according to the disaster evolution direction data to generate emergency level dynamic adjustment instructions. The final emergency status data for the current moment is generated by combining the confirmed assessment data and the emergency level dynamic adjustment instructions.

2. The earthquake disaster data-driven multi-state emergency assessment state generation method according to claim 1, characterized in that: The specific steps for generating corresponding disaster assessment reliability data based on each initial disaster assessment result include: Obtain the coverage and quality scores of the multi-source data used to generate the initial disaster assessment results; Obtain a consistency score for other initial disaster assessment results data within a preset range adjacent to the disaster spatial area during the same period; The credibility data is generated by weighted fusion calculation based on the coverage, the quality score, and the consistency score.

3. The earthquake disaster data-driven multi-state emergency assessment state generation method according to claim 2, characterized in that: After generating the first execution rule, the process also includes: For the initial disaster assessment result data whose credibility data is less than the first preset threshold, state suppression data is generated and stored in the cache queue; The credibility data of the state suppression data in the cache queue is recalculated according to a preset period. When the recalculated credibility data reaches the first preset threshold, the corresponding evaluation result data is released and marked as the evaluation data to be confirmed.

4. The earthquake disaster data-driven multi-state emergency assessment state generation method according to claim 1, characterized in that: The fluctuation of the disaster level in the regional assessment sequence data is measured within a preset time window. Specific steps include: Extract continuous disaster severity levels within a preset time window from the regional assessment sequence data; Calculate the sliding variance of the disaster severity level value; The continuous periods in which the sliding variance is lower than the second preset threshold are determined as stable periods, and the evaluation stability data is generated based on the stable periods.

5. The earthquake disaster data-driven multi-state emergency assessment state generation method according to claim 1, characterized in that: Based on the disaster evolution data, a forward-looking adjustment is made to generate a dynamic adjustment instruction for the emergency level. Specific steps include: The disaster evolution direction data is quantified to obtain the disaster change slope; The slope of the disaster change is matched against a set of predefined level adjustment thresholds; Based on the matching results, a dynamic adjustment instruction for the emergency level is generated, which includes the specific adjustment range and direction.

6. The earthquake disaster data-driven multi-state emergency assessment state generation method according to claim 1, characterized in that: The final emergency status data for the current moment is generated by comprehensively considering the confirmed assessment data and the emergency level dynamic adjustment instructions. Specific steps include: The current disaster level contained in the confirmed assessment data will be used as the base response level; The adjustment range and direction contained in the emergency level dynamic adjustment instruction are applied to the basic response level, and the target response level is calculated. The target response level is matched with a preset emergency response plan database to generate the final emergency status data, which includes specific action items, responsible parties, and time points.

7. The earthquake disaster data-driven multi-state emergency assessment state generation method according to claim 6, characterized in that: Matching the target response level with a pre-set emergency response plan database, specifically including: Obtain current emergency resource status data; The target response level, the characteristic data of the disaster spatial area, and the current emergency resource status data are coupled and calculated under multiple constraints. When a perfectly matching plan exists in the emergency response plan library, the corresponding action item is directly generated. When no perfectly matching plan exists, multiple similar plans are decomposed based on the coupling calculation results, and the decomposed elements are weighted and merged according to the priority of the multiple constraints to generate the specific action item that is adapted to the current multiple constraints.

8. The earthquake disaster data-driven multi-state emergency assessment state generation method according to claim 1, characterized in that: The method also includes continuous monitoring and state rollback, with specific steps including: After generating the final emergency status data, new initial disaster assessment results data for the disaster spatial area are continuously acquired; If the assessment stability data generated based on the new initial disaster assessment results is consistently lower than the third preset threshold, and the disaster evolution direction data indicates that the disaster is showing a mitigation trend, then a state rollback instruction is generated. Based on the state rollback instruction, the final emergency status data will be gradually downgraded from the current level until it is lifted.

9. The earthquake disaster data-driven multi-state emergency assessment state generation method according to claim 1, characterized in that: The method further includes an adaptive threshold adjustment step, specifically comprising: The correlation between the credibility data and the validity feedback of the final generated emergency status data within the monitoring historical period; Based on the aforementioned correlation, the values ​​of the first preset threshold and the second preset threshold are dynamically adjusted.

10. A multi-state emergency assessment status generation system driven by earthquake disaster data, characterized in that: The earthquake disaster data-driven multi-state emergency assessment state generation method as described in any one of claims 1 to 9 includes: The data processing module is used to acquire data on the spatial area of ​​the disaster and the initial disaster assessment results; The first state transition module is used to generate corresponding disaster assessment credibility data based on each initial disaster assessment result data, and compare the credibility data with a first preset threshold to generate a first execution rule; the first execution rule is used to mark initial disaster assessment result data that are greater than the first preset threshold as assessment data to be confirmed. The time-series construction module is used to collect the continuous assessment data to be confirmed in chronological order, taking the disaster spatial region as the dimension, to form regional assessment sequence data; The second state transition module is used to measure the volatility of the disaster level of the regional assessment sequence data within a preset time window, generate assessment stability data that characterizes the smoothness of the change in disaster level, and compare the assessment stability data with a second preset threshold to generate a second execution rule; the second execution rule is used to convert the assessment data to be confirmed that is greater than the second preset threshold into confirmed assessment data. The trend analysis and decision-making module is used to perform trend fitting based on the regional assessment sequence data corresponding to the confirmed assessment data, generate disaster evolution direction data, and make forward-looking adjustments based on the disaster evolution direction data to generate emergency level dynamic adjustment instructions. The integrated status output module is used to generate the final emergency status data for the current moment based on the confirmed assessment data and the emergency level dynamic adjustment instructions.

Images