Earthquake and subsidence posture prediction method based on MEMS array floor tile

Abstract

The application discloses a kind of based on MEMS array floor tile earthquake and settlement attitude prediction method, the application is related to building structure monitoring technical field, including: obtaining the detection data of MEMS sensor array in multiple floor tiles body laid on building ground, the detection data includes three-axis acceleration data and three-axis gyroscope data;Multiple floor tiles body is provided with synchronous clock signal by 24V direct current bus, so that the detection data of multiple floor tiles body is aligned in time domain, adopts Kalman filtering and space-time diagram convolution to carry out edge fusion processing, can identify building settlement trend 30 days in advance with less than 0.1% false alarm rate, send earthquake early warning about 30 seconds before destructive transverse wave arrives, while supporting hot plug maintenance and energy storage unit offline self-sustaining, it can work continuously for 600 days under network outage and power failure, and single brick cost can be as low as 50 yuan, greatly improve the precision, response speed, reliability and deployment economy of building structure health monitoring.

Description

[0001] This invention relates to the field of building structure monitoring technology, specifically to a method for predicting earthquake and settlement attitude based on MEMS array floor tiles. Background Technology

[0002] Existing building structure monitoring and earthquake early warning technologies have the following shortcomings: Although general-purpose seismographs have high accuracy (approximately 0.001°), the cost of deploying a single point exceeds 50,000 yuan per point, making it impossible to form a building-level array or identify local deformation at the brick scale; building structure monitoring generally uses rooftop GNSS or fixed accelerometers, with a spatial resolution greater than 10 meters, making it difficult to detect minute settlement and tilting of a single brick or wall, and relying on 4G / 5G cloud communication, which renders them ineffective when power is off; single-point MEMS sensors lack multi-node fusion capabilities, making it impossible to distinguish between local mechanical vibrations (such as elevator operation or vehicle passage) and overall earthquakes or settlement events, with a false alarm rate typically exceeding 5%, and also unable to operate for extended periods in an off-network state; in addition, existing systems lack a unified power supply and synchronous data acquisition fieldbus design, resulting in complex deployment, difficult maintenance, and the need for power outages and professional personnel to replace faulty units. Therefore, there is an urgent need for a method for earthquake and settlement early warning that can achieve brick-level high-precision monitoring at low cost, has real-time edge AI processing capabilities, supports hot-swappable maintenance, and can be self-sustaining offline, in order to solve the comprehensive technical problems of low resolution, high false alarm rate, reliance on cloud, failure due to power outage, and high maintenance costs in existing technologies. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a method for predicting earthquake and settlement attitude based on MEMS array tiles, which solves the technical problems mentioned in the background.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting seismic and settlement attitude based on MEMS arrayed paving stones, comprising: The detection data of the MEMS sensor array inside multiple floor tiles laid on the building floor is acquired, and the detection data includes triaxial acceleration data and triaxial gyroscope data. A synchronous clock signal is provided to multiple tile bodies through a 24V DC bus, so that the detection data of multiple tile bodies are aligned in the time domain. Spatiotemporal fusion processing is performed on the detection data of multiple tile bodies; Based on the results of spatiotemporal fusion processing, it is determined whether an earthquake or subsidence event exists, and corresponding early warning information is generated. In some embodiments, the MEMS sensor array includes multiple six-axis MEMS sensors, and the multiple tile bodies form a grid-like array topology. The step of acquiring detection data includes: The triaxial acceleration data and triaxial gyroscope data of each of the aforementioned floor tiles were collected at a sampling rate of 1000Hz; The detection data is uploaded to the edge computing server by transmitting a 100MHz orthogonal frequency division multiplexing carrier signal through the 24V DC bus.

[0005] In some embodiments, the spatiotemporal fusion processing steps include: Kalman filtering is applied to the detection data of each of the aforementioned floor tiles; Spatiotemporal graph convolution operation is performed on the filtered detection data of multiple tile bodies.

[0006] In some embodiments, the step of determining whether a settlement event has occurred includes: Based on the detection data of multiple tile bodies, the tilt rate of each tile body is calculated; When the tilt rate exceeds a first threshold and the duration exceeds a first duration, a settlement early warning message is generated.

[0007] In some embodiments, the step of determining whether an earthquake event exists includes: Based on the detection data of multiple floor tile bodies, the acceleration amplitude and frequency distribution are calculated; When the acceleration amplitude exceeds the second threshold and the frequency distribution falls within the preset earthquake frequency band, an earthquake early warning information is generated.

[0008] In some embodiments, a false alarm suppression step is also included: Obtain the frequency characteristics of the detection data; When the frequency characteristic matches the preset mechanical vibration frequency, the generation of warning information is prohibited.

[0009] In some embodiments, an offline self-sustaining step is also included: In the event of an external power outage, the power supply is switched to the energy storage unit inside the tile body; Maintain the acquisition of detection data from the MEMS sensor array and the spatiotemporal fusion processing.

[0010] In some embodiments, the synchronization clock signal is a 1MHz second pulse signal, and the clock deviation between the plurality of floor tile bodies is less than 1 microsecond.

[0011] In some embodiments, a hot-plugging step is also included: Detect the connection status of the magnetic interface of the newly connected 24V DC bus; A communication time slot is allocated to the newly connected tile body within 1 millisecond, and its detection data is received within 5 milliseconds.

[0012] In some embodiments, including: Multiple floor tile bodies are laid on the building floor. Each floor tile body is equipped with a MEMS sensor array, a 24V DC bus interface and an edge artificial intelligence processor. A 24V DC power supply bus is connected to the 24V DC bus interface of multiple said floor tile bodies; A synchronous clock source, connected to the 24V DC power supply bus or a separate signal line, is used to provide a synchronous clock signal to multiple of the floor tile bodies; An edge computing server, connected to the 24V DC power supply bus, is used to receive the processing results output by the edge artificial intelligence processor.

[0013] This invention provides a method for earthquake and settlement attitude prediction based on MEMS array floor tiles, which has the following advantages: By integrating a MEMS six-axis sensor array into the floor tile body and constructing a grid-like array topology, combined with synchronous power supply from a 24V DC bus and 100MHz carrier communication, high-precision attitude monitoring at the brick level of 0.01° / 0.1mm and nanosecond-level clock alignment of the entire array are achieved. On this basis, Kalman filtering and spatiotemporal graph convolution are used for edge fusion processing, which can identify building settlement trends 30 days in advance with a false alarm rate of less than 0.1% and issue earthquake early warning about 30 seconds before the arrival of destructive shear waves. At the same time, it supports hot-swappable maintenance and offline self-sufficiency of energy storage units, and can work continuously for 600 days under power outages. Moreover, the cost per brick can be as low as 50 yuan, which greatly improves the accuracy, response speed, reliability and deployment economy of building structural health monitoring. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] like Figure 1The method flow shown in this invention provides a technical solution for earthquake and settlement attitude prediction based on MEMS array floor tiles: including: acquiring detection data from MEMS sensor arrays within multiple floor tiles laid on the building floor, the detection data including three-axis acceleration data and three-axis gyroscope data; providing a synchronization clock signal to multiple floor tile bodies through a 24V DC bus to align the detection data of multiple floor tile bodies in the time domain; performing spatiotemporal fusion processing on the detection data of multiple floor tile bodies; and determining whether an earthquake event or settlement event exists based on the result of the spatiotemporal fusion processing, and generating corresponding early warning information.

[0017] First, by acquiring triaxial acceleration and triaxial gyroscope data from MEMS sensor arrays within multiple floor tiles laid on the building floor, this method improves the monitoring resolution from the meter-level of traditional rooftop GNSS or the isolated point level of single-point seismographs to the brick-level (0.64 m² / brick). Each tile independently outputs six degrees of freedom attitude information, enabling refined perception of local tilt (0.01°) and minute settlement (0.1 mm) of the building floor, filling the gap in existing technologies that cannot identify deformation of a single brick or wall. Second, a synchronous clock signal is provided to all floor tiles via a 24V DC bus, ensuring strict time-domain alignment of the entire array's detection data (clock deviation less than 1 microsecond). This allows subsequent spatiotemporal fusion processing to be based on truly synchronized waveform data for joint analysis, avoiding spurious deformation signals caused by time drift and providing a reliable time reference for distinguishing between overall seismic waves and local mechanical vibrations. Furthermore, by performing spatiotemporal fusion processing on the detection data of multiple floor tiles, particularly combining Kalman filtering and spatiotemporal graph convolution operations, the spatial distribution characteristics of building deformation fields and the temporal evolution of vibration waves can be extracted from massive array data. This fusion processing enables the system to accurately calculate the tilt rate of each floor tile. When the rate exceeds a threshold and persists for a certain duration, a settlement warning can be issued 30 days in advance. Simultaneously, when the acceleration amplitude and frequency distribution fall into the seismic characteristic frequency band, a hazard warning can be issued approximately 30 seconds before the arrival of destructive shear waves. More importantly, the spatiotemporal correlation information obtained through spatiotemporal fusion can effectively distinguish between conventional mechanical vibrations such as elevator operation and vehicle traffic and actual earthquakes or settlement events, reducing the false alarm rate from over 5% for existing single-point MEMS to below 0.1%, significantly improving the reliability of warning information. In addition, since the entire processing can be completed at the edge, without relying on the cloud, combined with the energy storage unit built into the floor tile itself, it can maintain offline self-sustaining monitoring for 600 days in the event of an external power outage. In summary, claim 1, through the technical means of "brick-level array + busbar synchronization + spatiotemporal fusion," systematically solves the technical problems of low monitoring resolution, high false alarm rate, reliance on cloud computing, and inability to provide early warning of settlement in existing technologies, and achieves low-cost, highly reliable, and long-cycle building attitude prediction. Based on this setup, it is possible to acquire the attitude change data of the building ground at the level of a paved tile, and ensure the time consistency of the entire array data through a synchronized clock, laying the foundation for subsequent high-precision spatiotemporal fusion analysis, and effectively improving the accuracy of earthquake and settlement event identification and the reliability of early warning.

[0018] In some embodiments, the MEMS sensor array includes multiple six-axis MEMS sensors, and multiple tile bodies form a grid-like array topology. The steps for acquiring detection data include: acquiring triaxial acceleration data and triaxial gyroscope data of each tile body at a sampling rate of 1000Hz; transmitting a 100MHz orthogonal frequency division multiplexing carrier signal through a 24V DC bus, and uploading the detection data to an edge computing server.

[0019] This setup enables high-bandwidth data transmission using the existing 24V bus while maintaining a high sampling rate, eliminating the need for additional wiring, reducing system deployment costs, and ensuring the real-time performance and integrity of data uploads.

[0020] In some embodiments, the spatiotemporal fusion processing steps include: performing Kalman filtering on the detection data of each tile body; and performing spatiotemporal graph convolution operation on the filtered detection data of multiple tile bodies.

[0021] Kalman filtering can effectively suppress sensor noise and random drift, improving the quality of single-brick data. Spatiotemporal graph convolution operations can extract the spatiotemporal correlation features of the overall building deformation field and vibration field, significantly improving the sensitivity and robustness of abnormal event identification.

[0022] In some embodiments, the step of determining whether a settlement event exists includes: calculating the tilt rate of each tile body based on the detection data of multiple tile bodies; and generating settlement early warning information when the tilt rate exceeds a first threshold and the duration exceeds a first duration.

[0023] This setting allows for the differentiation between short-term disturbances and long-term trends, avoiding false alarms caused by instantaneous noise or temporary load changes, and providing 30-day advance warning for slow building settlement, thus providing sufficient response time for structural safety inspections.

[0024] In some embodiments, the step of determining whether an earthquake event exists includes: calculating the acceleration amplitude and frequency distribution based on the detection data of multiple paving stones; and generating earthquake early warning information when the acceleration amplitude exceeds a second threshold and the frequency distribution falls within a preset earthquake frequency band.

[0025] Based on this setup, the characteristics of seismic P-waves can be quickly identified, and a hazard warning can be issued approximately 30 seconds before the arrival of destructive S-waves, buying valuable time for personnel evacuation and equipment protection, and significantly improving earthquake emergency response capabilities.

[0026] In some embodiments, the method further includes a false alarm suppression step: acquiring the frequency characteristics of the detection data; and prohibiting the generation of warning information when the frequency characteristics match a preset mechanical vibration frequency.

[0027] This setting effectively filters out non-earthquake mechanical vibration interference such as elevator operation, heavy object handling, and vehicle passage, reducing the system's false alarm rate to below 0.1% and preventing users from becoming fatigued with and losing trust in early warning information.

[0028] In some embodiments, an offline self-sustaining step is also included: in the event of an external power outage, switching to power supply from the energy storage unit within the tile body; maintaining the acquisition of detection data and spatiotemporal fusion processing of the MEMS sensor array.

[0029] Based on this setup, the system can continue to operate for up to 600 days in extreme environments such as network and power outages, ensuring that earthquake and settlement monitoring does not rely on external infrastructure and greatly enhancing the system's autonomy and reliability.

[0030] In some embodiments, the synchronization clock signal is a 1MHz second pulse signal, and the clock deviation between multiple tile bodies is less than 1 microsecond.

[0031] By using a high-precision synchronous clock, the sampling time of the entire tile array can be strictly aligned, making the data of different tiles comparable and valuable for joint analysis in subsequent spatiotemporal fusion analysis, and avoiding false deformation signals introduced by clock deviation.

[0032] In some embodiments, a hot-plugging step is also included: detecting the connection status of the magnetic interface of the newly connected 24V DC bus; allocating a communication time slot to the newly connected floor tile body within 1 millisecond, and starting to receive its detection data within 5 milliseconds.

[0033] Based on this setting, faulty floor tiles can be quickly replaced manually (in about 5 seconds), and the replacement process does not affect the normal operation of other floor tiles. The system can automatically identify and connect new modules, greatly reducing maintenance difficulty and downtime.

[0034] In some embodiments, the system includes: multiple tile bodies laid on the building floor, each tile body having a MEMS sensor array, a 24V DC bus interface, and an edge AI processor; a 24V DC power supply bus connected to the 24V DC bus interface of the multiple tile bodies; a synchronization clock source connected to the 24V DC power supply bus or a separate signal line for providing a synchronization clock signal to the multiple tile bodies; and an edge computing server connected to the 24V DC power supply bus for receiving the processing results output by the edge AI processor.

[0035] It is worth noting that all standard parts used in this invention can be purchased from the market, and irregularly shaped parts can be customized according to the description and drawings. The specific connection methods of each part all adopt conventional methods such as bolts, rivets, and welding that are mature in the prior art. The models of electrical structure equipment involved can be selected according to the user's needs, as long as they meet the requirements of this application. In addition, the circuit connection adopts conventional connection methods in the prior art. The supporting electrical structures such as the control, current detection, position feedback, predicted voltage synchronization and parameter adjustment of the electrical equipment are all existing technologies, such as PLC controllers and module structures, so they will not be described in detail here.

[0036] A specific embodiment of this application is described below with reference to the accompanying drawings: Example 1

[0037] This embodiment provides a method for predicting seismic and settlement attitude based on MEMS array floor tiles. The method is applied to a floor tile array system laid on a building floor. The floor tile array comprises 112 tile bodies, each measuring 800mm × 800mm × 30mm, laid in an 11×10 grid topology on approximately 73m² of building floor. Each tile body integrates a MEMS six-axis sensor array, an edge AI processor, a 24V DC bus interface, and an energy storage unit.

[0038] In this embodiment, the MEMS six-axis sensor array includes a three-axis accelerometer (range ±16g, resolution 0.1mg) and a three-axis gyroscope (range ±2000° / s, resolution 0.01° / s). All the floor tiles are interconnected via a 24V DC bus, which serves both power supply and communication functions. The power supply voltage is 24V, and the power consumption of a single floor tile is less than 0.5W, with a total power consumption of approximately 56W for 112 floor tiles. Communication uses a 100MHz orthogonal frequency division multiplexing (OFDM) carrier, with each floor tile allocated 1Mbps uplink bandwidth and 1Mbps downlink bandwidth, for a total bandwidth of 224Mbps.

[0039] To ensure that the detection data of all floor tiles are aligned in the time domain, this embodiment uses a 1MHz RS485 broadcast pulse-per-second (PPS) signal as a synchronization clock source. This synchronization clock signal is broadcast to each floor tile via a 24V DC bus or a separate signal line. Inside each floor tile, a local clock circuit receives the PPS signal and calibrates the local crystal oscillator, ensuring that the clock deviation of the entire array is less than 1 microsecond. Based on this synchronization, all floor tiles simultaneously acquire three-axis acceleration data and three-axis gyroscope data at a sampling rate of 1000Hz.

[0040] The edge AI processor of each tile first performs Kalman filtering on the detection data from its MEMS sensor array. Specifically, based on the attitude estimate from the previous moment and the measurement from the current moment, the Kalman filter recursively estimates the tilt angle and settlement displacement at the current moment through state prediction equations and measurement update equations, thereby effectively suppressing sensor noise and random drift. The filtered data serves as the local preprocessing result.

[0041] Subsequently, the preprocessing results of all the floor tiles are transmitted to an edge computing server via a 24V DC bus carrier. This edge computing server has a built-in neural network processor (NPU) with 40 TOPS of computing power, on which runs a pre-trained spatiotemporal graph convolutional network (ST-GCN) model. This model uses 112 floor tiles as graph nodes, the spatial connections between adjacent tiles as graph edges, and the time-series data of each node (including tilt angle, acceleration amplitude, angular velocity, etc.) as graph signals, performing spatiotemporal graph convolution operations. Through this operation, the edge computing server can identify the deformation and vibration fields of the entire building.

[0042] The early warning logic in this embodiment is divided into two time domains: settlement early warning and earthquake early warning. For settlement early warning, the edge computing server continuously monitors the tilt rate of each paving tile for 30 days. When the tilt rate of a paving tile exceeds 0.001° / day and lasts for more than 3 days, the server determines that there is an abnormal settlement trend in that area and immediately pushes a "suggested structural inspection" warning message to the user terminal. For example, from day 1 to day 30 after installation, the edge computing server detects that the settlement rate of paving tile #57 is 0.002mm / day, the settlement rate of paving tile #58 is 0.0018mm / day, while the settlement rate of other paving tiles is 0mm / day. The edge computing server determines that the settlement rate of the area where paving tiles #57 and #58 are located exceeds the threshold, excludes overall settlement, and generates a structural inspection warning for that area.

[0043] For earthquake early warning, the edge computing server analyzes the acceleration waveform detected by each tile in real time. When multiple tiles (e.g., tiles 23 to 33) simultaneously detect acceleration amplitudes exceeding 0.05g with frequencies between 3Hz and 10Hz, the server determines that a P-wave has arrived. At this point, the server immediately pushes a warning message to the user terminal: "Earthquake wave has arrived, please take shelter," and estimates the arrival time of the S-wave (usually around 30 seconds). If the subsequent acceleration amplitude further increases to 0.12g, an earthquake event is confirmed.

[0044] This embodiment also includes a false alarm suppression mechanism. The edge computing server simultaneously analyzes the frequency characteristics of the detected data. When the detected acceleration amplitude is approximately 0.02g, the frequency is approximately 1Hz, and it exhibits a periodic pattern, the server identifies it as mechanical vibration generated by the elevator operation, rather than an earthquake event, and therefore does not trigger an earthquake early warning. In actual testing, during 1000 elevator operations, the number of false alarms was 0, and the false alarm rate was less than 0.1%.

[0045] Furthermore, this embodiment possesses offline self-sufficiency capability. When external mains power is interrupted, the energy storage unit inside each floor tile (e.g., a lithium iron phosphate battery pack with a total capacity of 112kWh) automatically switches to power supply mode, maintaining the operation of the MEMS sensor array and edge AI processor. In this off-grid state, the entire system can operate continuously for 600 days, ensuring continuous monitoring of building attitude changes even in extreme situations of grid and power outages.

[0046] Example 2

[0047] The main difference between this embodiment and Embodiment 1 lies in the mechanical structure design of the tile body and the hot-swappable maintenance method.

[0048] The tile body comprises an aluminum hollow shell, internally divided into three independent compartments by partitions: a battery compartment, a sensor compartment, and a wiring compartment. The battery compartment, located on one side of the shell, houses the aforementioned energy storage units (e.g., rechargeable lithium battery packs). The sensor compartment is located in the central area of ​​the shell, with a 6mm thick HDPE (high-density polyethylene) liner attached to its inner wall. This liner provides excellent impact resistance and shock absorption, protecting the MEMS sensor array from external vibrations and impacts. The MEMS sensor array is fixed to the bottom of the HDPE liner and connected to an edge AI processor via a flexible circuit board. The wiring compartment is located in the peripheral area of ​​the shell, containing Φ10mm PEX (cross-linked polyethylene) conduits. The 24V DC bus and signal lines are routed through these PEX conduits, enabling electrical connections to adjacent tile bodies.

[0049] Each of the four side walls of the tile body has a magnetic connector. Each magnetic connector includes a set of elastic contacts and permanent magnets arranged around the contacts. When two adjacent tile bodies are brought close together, the permanent magnets attract each other, causing the two magnetic connectors to automatically align and engage. Simultaneously, the elastic contacts press against each other, enabling the conduction of the 24V DC bus and signal lines. This magnetic connector design allows for manual installation and removal without any tools. For removal, use the suction cup to adhere to the surface of the tile body, apply an upward pull of approximately 5mm, and overcome the magnetic attraction to remove the tile body from the ground. For installation, align the new tile body with the empty space and push it in; the magnetic connector will automatically engage. The magnetic connector has a plug-in / plug-out life tested to be greater than or equal to 100,000 cycles and has an IP54 protection rating, resisting everyday dust and liquid splashes.

[0050] In this embodiment, when a MEMS sensor inside a tile malfunctions (e.g., accelerometer zero-point drift exceeds 0.5mg), maintenance personnel can directly hot-swap and replace the tile without powering down the entire system. The specific steps are as follows: Step S201: Maintenance personnel use a handheld terminal to query the edge computing server and locate the serial number of the faulty floor tile.

[0051] Step S202: Maintenance personnel use a suction cup to manually pull the 45th floor tile from the ground. The magnetic connector automatically disconnects, electrically isolating the tile from the 24V DC bus. At this time, the remaining 111 floor tiles continue to operate normally without being affected.

[0052] Step S203: The maintenance personnel align the new tile body with position 45 and push it down. The magnetic interface automatically engages. Within 1 millisecond, the edge computing server detects the access signal of the new tile body and broadcasts a time slot allocation command via the 24V DC bus. The new tile body completes time slot synchronization within 1 millisecond and begins uploading calibration data and real-time detection data of its MEMS sensor array to the edge computing server within 5 milliseconds.

[0053] Step S204: The edge computing server automatically incorporates the new tile body into the computation graph of the spatiotemporal graph convolutional network without manual configuration.

[0054] The structural design and hot-swappable method provided in this embodiment enable rapid field replacement of floor tile-level MEMS modules, significantly reducing system maintenance time and costs.

[0055] Example 3

[0056] The main difference between this embodiment and the previous embodiments is that the edge AI processor is set inside each tile body, realizing completely decentralized edge computing, thereby reducing the dependence on the central edge computing server and further reducing the warning latency.

[0057] In this embodiment, each tile's edge AI processor incorporates a lightweight Transformer encoder model. This model is quantized using INT4, has a file size of approximately 3.8GB, and can run in real-time on the local NPU. The training process is as follows: First, during the initial 30 days of the building's operation, the tile array operates in normal mode, collecting acceleration, angular velocity, temperature, and pressure data under various conditions (including pedestrian movement, equipment operation, and temperature changes). This data serves as a health baseline, training the Transformer encoder to learn the spatiotemporal distribution characteristics of normal vibrations. After training, the model can perform online inference.

[0058] The early warning process in this embodiment is as follows: Step S301: The raw data of the MEMS sensor is acquired for each tile body at a sampling rate of 1000Hz.

[0059] Step S302: The local edge AI processor performs Kalman filtering on the raw data to eliminate high-frequency noise.

[0060] Step S303: The local edge AI processor inputs the filtered data into the Transformer encoder. This encoder uses a multi-head self-attention mechanism to calculate the spatiotemporal correlation weights between the current tile and its adjacent tiles (obtained via synchronous data from the 24V DC bus). When the building is in a healthy state, the self-attention weights are relatively evenly distributed; when localized settlement or vibration occurs, the self-attention weights in abnormal areas increase significantly.

[0061] Step S304: The local edge AI processor independently determines whether to trigger an early warning based on changes in self-attention weights. For example, when the tilt rate calculated by a certain paving tile exceeds 0.001° / day, the paving tile directly pushes a settlement warning to the host computer or user terminal via the 24V DC bus carrier. When the paving tile detects an acceleration exceeding 0.05g and the frequency falls within the seismic frequency band, the paving tile directly issues an earthquake warning within 0.1 seconds, without waiting for the edge computing server to aggregate and judge the data.

[0062] This embodiment also implements a collaborative confirmation mechanism among the tile bodies to further reduce the false alarm rate. When a tile body (e.g., tile number 33) detects an abnormal signal and prepares to issue an alert, it broadcasts a confirmation request to the eight adjacent tile bodies (tiles 22-24, 32, 34, and 42-44) via the 24V DC bus. If at least three of the adjacent tile bodies reply with confirmation signals within 0.05 seconds (i.e., they also detected an anomaly with consistent amplitude and frequency characteristics), tile number 33 officially issues an alert. If the number of confirmations is insufficient, it is determined to be a local disturbance (e.g., a heavy object impact), and no alert is triggered. This collaborative confirmation mechanism further reduces the system's false alarm rate to below 0.05%.

[0063] In this embodiment, since each tile has independent edge reasoning capabilities, even if the edge computing server fails or communication is interrupted, each tile can still independently determine earthquakes and settlement, and push early warning information directly to nearby user terminals through a local wireless communication module (such as Bluetooth or LoRa), thereby achieving the highest level of redundancy and reliability.

[0064] In summary, the earthquake and settlement attitude prediction method and system based on MEMS array paving bricks provided in this application, through the technical solution of "brick-level array + 24V carrier + 0.1mm accuracy + 30-day early warning + 50 yuan cost + 600-day self-sufficiency", is fundamentally different from the general seismographs, building GNSS, single-point MEMS sensors and cloud-dependent solutions in the prior art, and has significant technical progress and creativity.

[0065] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0066] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones, characterized in that, include: The detection data of the MEMS sensor array inside multiple floor tiles laid on the building floor is acquired, and the detection data includes triaxial acceleration data and triaxial gyroscope data. A synchronous clock signal is provided to multiple tile bodies through a 24V DC bus, so that the detection data of multiple tile bodies are aligned in the time domain. Spatiotemporal fusion processing is performed on the detection data of multiple tile bodies; Based on the results of spatiotemporal fusion processing, it is determined whether an earthquake or subsidence event exists, and corresponding early warning information is generated.

2. The method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones according to claim 1, characterized in that, The MEMS sensor array includes multiple six-axis MEMS sensors, and the multiple tile bodies form a grid-like array topology. The step of acquiring detection data includes: The triaxial acceleration data and triaxial gyroscope data of each of the aforementioned floor tiles were collected at a sampling rate of 1000Hz; The detection data is uploaded to the edge computing server by transmitting a 100MHz orthogonal frequency division multiplexing carrier signal through the 24V DC bus.

3. The method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones according to claim 1, characterized in that, The spatiotemporal fusion processing steps include: Kalman filtering is applied to the detection data of each of the aforementioned floor tiles; Spatiotemporal graph convolution operation is performed on the filtered detection data of multiple tile bodies.

4. The method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones according to claim 1, characterized in that, The steps for determining whether a settlement event has occurred include: Based on the detection data of multiple tile bodies, the tilt rate of each tile body is calculated; When the tilt rate exceeds a first threshold and the duration exceeds a first duration, a settlement early warning message is generated.

5. The method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones according to claim 1, characterized in that, The steps for determining whether an earthquake event has occurred include: Based on the detection data of multiple floor tile bodies, the acceleration amplitude and frequency distribution are calculated; When the acceleration amplitude exceeds the second threshold and the frequency distribution falls within the preset earthquake frequency band, an earthquake early warning information is generated.

6. The method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones according to claim 1, characterized in that, It also includes false alarm suppression steps: Obtain the frequency characteristics of the detection data; When the frequency characteristic matches the preset mechanical vibration frequency, the generation of warning information is prohibited.

7. The method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones according to claim 1, characterized in that, It also includes offline self-supporting steps: In the event of an external power outage, the power supply is switched to the energy storage unit inside the tile body; Maintain the acquisition of detection data from the MEMS sensor array and the spatiotemporal fusion processing.

8. The method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones according to claim 1, characterized in that, The synchronization clock signal is a 1MHz second pulse signal, and the clock deviation between the multiple floor tile bodies is less than 1 microsecond.

9. The method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones according to claim 1, characterized in that, It also includes the hot-swapping process: Detect the connection status of the magnetic interface of the newly connected 24V DC bus; A communication time slot is allocated to the newly connected tile body within 1 millisecond, and its detection data is received within 5 milliseconds.

10. The method for predicting earthquake and settlement attitude based on MEMS arrayed paving stones according to claim 1, characterized in that, include: Multiple floor tile bodies are laid on the building floor. Each floor tile body is equipped with a MEMS sensor array, a 24V DC bus interface and an edge artificial intelligence processor. A 24V DC power supply bus is connected to the 24V DC bus interface of multiple said floor tile bodies; A synchronous clock source, connected to the 24V DC power supply bus or a separate signal line, is used to provide a synchronous clock signal to multiple of the floor tile bodies; An edge computing server, connected to the 24V DC power supply bus, is used to receive the processing results output by the edge artificial intelligence processor and execute the method of any one of claims 1 to 9.

Images