Subgrade engineering settlement deformation monitoring device and method

Abstract

The invention relates to the technical field of roadbed engineering monitoring, in particular to a roadbed engineering settlement deformation monitoring device and method, and the device comprises an optical fiber sensor network, a data collection module, a signal processing module, a model matching module, a frequency analysis module, a data fusion module, an early warning trigger module and a prediction report module. The optical fiber sensor network is deployed in a roadbed to collect an optical signal delay sequence, fluctuation characteristics are extracted through data collection and signal processing, settlement acceleration risks are judged through model matching, the speed trend is determined through frequency analysis, historical and real-time data are integrated through data fusion, an alarm is triggered when an early warning module exceeds a threshold value, and a risk assessment report is generated through a prediction module. According to the device, roadbed three-dimensional global monitoring and full-process automation are achieved, monitoring blind areas are eliminated, early warning lag is shortened, and accurate support is provided for roadbed stability operation and maintenance.

Description

Technical Field

[0001] This application relates to the field of roadbed engineering monitoring technology, and in particular to a roadbed engineering settlement and deformation monitoring device and method. Background Technology

[0002] Currently, settlement monitoring in roadbed engineering mostly uses single-point monitoring equipment (such as settlement nails, GPS receivers, and single-point strain sensors). Such equipment has significant technical limitations: due to the limited deployment method, single-point equipment can only monitor settlement data at specific points and cannot cover the entire cross-section of the roadbed (longitudinal, transverse, and depth), which easily creates monitoring blind spots. Especially in soft soil foundation roadbeds, local accelerated settlement often occurs in unmonitored areas, leading to missed risk assessments. When accelerated settlement occurs in the roadbed, the delayed early warning can easily cause roadbed cracking, pavement collapse, and other defects.

[0003] Therefore, a roadbed engineering settlement and deformation monitoring device and method are invented to solve the problems mentioned in the background art. Summary of the Invention

[0004] The purpose of this invention is to overcome the technical problems of existing roadbed settlement monitoring devices, such as "limited monitoring range and delayed early warning", and to provide a roadbed engineering settlement deformation monitoring device. Through distributed optical fiber three-dimensional full-domain perception and multi-module automated collaborative processing, it realizes real-time acquisition, automatic analysis, timely early warning and future trend prediction of roadbed settlement deformation, eliminates monitoring blind spots, shortens risk response time and provides accurate data support for roadbed stability management.

[0005] The roadbed engineering settlement and deformation monitoring device provided in this application adopts the following technical solution: It includes a fiber optic sensor network deployed inside the roadbed structure to collect real-time optical signal propagation data corresponding to roadbed deformation and output an optical signal delay sequence; a data acquisition module connected to the fiber optic sensor network to receive the optical signal delay sequence and form an initial delay distribution; a signal processing module connected to the data acquisition module to process weak fluctuations in the initial delay distribution, determine the fluctuation characteristic sequence, and output deformation reflection indicators; and a model matching module electrically connected to the signal processing module, which has a pre-stored preset deformation model and matches the fluctuation characteristic sequence with the model. If the matching degree exceeds a preset threshold... The system outputs acceleration risk labels; a frequency analysis module, connected to the model matching module, extracts frequency components based on the acceleration risk labels, determines the dominant mode, and outputs the settlement velocity change trend; a data fusion module, connected to both the frequency analysis module and the historical database, fuses the change trend with historical data, updates the trend model using filtering methods, and outputs dynamic characteristic estimates; an early warning trigger module, electrically connected to the data fusion module, extracts acceleration change components, triggers an alarm and outputs an early warning signal sequence if a preset threshold is exceeded; and a prediction report module, connected to the early warning trigger module, applies prediction algorithms to optimize future trends, determines the settlement evolution path, and generates a risk assessment report.

[0006] Optionally, the fiber optic sensor network is a distributed fiber optic sensor array, which is deployed in layers along the longitudinal, transverse and depth directions of the roadbed. The distributed fiber optic sensor array includes several fiber optic sensing units, and each fiber optic sensing unit is evenly embedded in different stress and strain sensitive areas of the roadbed structure.

[0007] Optionally, the data acquisition module includes an optical signal receiver, an analog-to-digital converter, and a data buffer unit. The optical signal receiver receives the optical signal delay sequence output by the optical fiber sensor network, converts it into a digital signal by the analog-to-digital converter, stores it through the data buffer unit, and outputs it to the signal processing module.

[0008] Optionally, a model storage unit is also included, connected to the model matching module, for storing the preset deformation model, which is a settlement acceleration feature model library constructed based on different roadbed types and geological conditions.

[0009] Optionally, the filtering method in the data fusion module is Kalman filtering or particle filtering. The data fusion module also includes a trend model update unit, which is used to dynamically adjust the trend model parameters based on the fused real-time data and historical data.

[0010] Optionally, the early warning triggering module includes a threshold comparison unit and an alarm execution unit. The threshold comparison unit is used to compare the extracted acceleration change component with a preset threshold. The alarm execution unit includes an audible and visual alarm and a remote communication module, which are used to simultaneously output local audible and visual alarms and remote early warning information.

[0011] Optionally, the prediction algorithm in the prediction report module is a neural network prediction algorithm or a time series prediction algorithm. The prediction report module also includes a report output unit for outputting the risk assessment report in the form of visual charts and text descriptions.

[0012] Optionally, a power supply module may also be included, which may be a solar power supply unit or an AC power supply unit, for providing stable power to the fiber optic sensor network, data acquisition module, signal processing module, model matching module, frequency analysis module, data fusion module, early warning triggering module, and prediction report module.

[0013] A method for monitoring settlement and deformation in roadbed engineering, characterized by comprising: S1: Real-time optical signal propagation data is collected by deploying an optical fiber sensor network in the roadbed structure to obtain an optical signal delay sequence to form an initial delay distribution, which reflects the influence of roadbed deformation on signal propagation time; S2: Based on the initial delay distribution, a signal analysis method is used to process the weak fluctuations, determine the fluctuation characteristic sequence, and obtain the deformation reflection index. The deformation reflection index is used to quantify the degree of deformation of the roadbed structure. S3: Match the wave characteristic sequence with the preset deformation model. If the matching degree is higher than the preset threshold, it is judged as potential settlement acceleration, and an acceleration risk label is obtained. The acceleration risk label indicates the risk of aggravated roadbed settlement. S4: Based on the accelerated risk label, frequency components are extracted using a transformation method, and the dominant mode in the frequency components is determined to identify the settlement velocity change trend, which describes the evolution direction of the roadbed deformation rate; S5: Based on the settlement velocity change trend, historical monitoring data are integrated, and a filtering method is used to update the trend model to obtain dynamic characteristic estimates. The dynamic characteristic estimates integrate real-time and historical information to evaluate the roadbed stability. S6: Extract the acceleration change component from the dynamic characteristic estimate. If the acceleration change component exceeds a preset threshold, trigger the alarm mechanism to obtain an early warning signal sequence. The early warning signal sequence records roadbed risk warning information. S7: Based on the warning signal sequence, apply the prediction algorithm to optimize the future trend, determine the settlement evolution path and generate a risk assessment report, which summarizes the potential hazards and development predictions of the roadbed.

[0014] In summary, this application includes the following beneficial technical effects: 1. Distributed fiber optic sensors are deployed in layers along the longitudinal, transverse, and depth of the roadbed, covering the entire cross-section of the roadbed. Compared with single-point monitoring, they can capture minute deformations in any area and completely eliminate monitoring blind spots. 2. From data acquisition (optical signal → digital signal), processing (wavelet denoising, feature extraction), analysis (model matching, frequency analysis) to early warning (audio-visual + remote) and prediction, no human intervention is required throughout the entire process, reducing early warning response time. Compared with traditional manual processing, the lag time is reduced by more than 90%. 3. It adopts an IP66 protected monitoring box, waterproof fiber optic connectors, and dual-mode power supply (solar + mains power), which can adapt to complex environments such as high temperature, rain, and vibration on the roadbed, and improve the reliability of continuous operation; 4. By combining historical data with real-time trends, the LSTM neural network is used to predict the settlement path for the next 30 days with a prediction error of <0.5mm. The generated visualized risk assessment report (including reinforcement recommendations) can directly guide operation and maintenance decisions and reduce the cost of treating roadbed defects. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the control signal transmission of this device; The system comprises: 1. Fiber optic sensor network; 2. Data acquisition module; 3. Signal processing module; 4. Model matching module; 5. Frequency analysis module; 6. Data fusion module; 7. Early warning triggering module; and 8. Prediction report module. Detailed Implementation

[0016] The present application will be further described in detail below with reference to the accompanying drawings. In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting the present invention.

[0017] Reference Figure 1As shown: The settlement and deformation monitoring device for this roadbed project includes a fiber optic sensor network 1, a data acquisition module 2, a signal processing module 3, a model matching module 4, a frequency analysis module 5, a data fusion module 6, an early warning triggering module 7, a prediction report module 8, and a historical database. The fiber optic sensor network 1 is embedded in layers along the roadbed depth, making close contact with the roadbed soil. Its signal output end is connected to the optical signal input end of the data acquisition module 2 via armored optical fiber. The fiber optic connector uses an SC-type waterproof connector, which is tightened by screws and sealed with sealant to ensure airtightness in environments with high roadbed moisture content. The data acquisition module 2 and the signal processing module 3 are installed in a stainless steel monitoring box at the edge of the roadbed. The monitoring box has an IP66 protection rating. The data acquisition module 2 is located on the left, and the signal processing module 3 is located on the right. They are connected by a shielded copper core data cable. The two ends of the data cable are soldered to the signal interfaces of the two modules, and the solder joints are covered with insulating heat shrink tubing. The data cable is fixed along a PVC cable tray on the inner wall of the monitoring box, maintaining a 10cm distance from the power cable to avoid electromagnetic interference. The output end of the signal processing module 3 is connected to the input end of the model matching module 4 via a tinned copper wire with a cross-sectional area of ​​0.3mm². The model matching module 4 is fixed in the middle of the monitoring box by an insulating bracket, located 15cm directly below the signal processing module 3. The two are kept in a stable relative position by metal supports. The signal output of module 4 is connected to frequency analysis module 5 via an LC-type fiber optic patch cord with a bending radius of not less than 30mm. Frequency analysis module 5 is fixed on the right side of the monitoring box and arranged parallel to model matching module 4, with a spacing of 10cm. Frequency analysis module 5 is connected to the first input of data fusion module 6 via an RJ45 communication cable. The second input of data fusion module 6 is connected to the historical database via a Cat5e network cable. The historical database is an industrial-grade server located in the monitoring center computer room 500m away. The network cable is laid through a galvanized steel pipe with a diameter of 50mm, filled with fireproof cotton. Data fusion module 6 is fixed at the bottom of the monitoring box, and its output is connected to early warning trigger module 7 via a PVC insulated signal wire. Early warning trigger module 7 is installed on a slide rail on the top of the monitoring box and its position can be adjusted horizontally along the slide rail. The vertical distance between it and data fusion module 6 is 30cm. Early warning trigger module 7 communicates with prediction report module 8 via a LoRa wireless transmission module at a frequency of 433MHz. Prediction report module 8 is an industrial tablet PC placed on the monitoring center console and is damped by rubber pads.

[0018] The fiber optic sensor network 1 generates micro-strain due to roadbed deformation, causing changes in the optical signal propagation path and outputting an optical signal delay sequence (containing time-delay correspondence). The data acquisition module 2 maps this sequence to a two-dimensional initial delay distribution according to the fiber optic deployment location (horizontal axis is roadbed length, vertical axis is depth, and value is delay time), visually presenting the deformation differences in different areas. The signal processing module 3 uses a wavelet transform algorithm to process the initial delay distribution: a db4 wavelet basis is selected, decomposed into 5 layers, and environmental noise is removed by threshold denoising (soft threshold, threshold = 0.6745 × median(|coefficient|)). The detail coefficients of layers 3-5 are extracted as the wave feature sequence (containing amplitude, frequency, and phase parameters). Then, the feature sequence is converted into deformation reflection indicators (such as cumulative strain value and strain rate) through a strain-delay conversion model (Δε = k × Δt, k is a calibration coefficient determined by the fiber optic factory parameters, with a value of 0.01 με / ns). Model matching module 4 has a pre-stored deformation model library. The models are constructed based on settlement data from over 100 similar roadbed projects. Each model includes the range of fluctuation characteristic parameters for the acceleration phase (e.g., amplitude growth rate > 5% / d, dominant frequency 0.2-0.5Hz). The matching degree between the fluctuation characteristic sequence and the model is calculated using a cosine similarity algorithm (similarity = dot product of feature vectors / product of modulus). When the matching degree is > 85%, an acceleration risk label is output. Frequency analysis module 5 performs Fourier transform on the fluctuation characteristic sequence with the acceleration risk label to obtain the frequency-energy spectrum. The frequency range with an energy proportion > 60% is selected as the dominant mode. The settlement velocity change trend (increasing / decreasing / stable) is determined by the rate of change of the dominant frequency over time (e.g., increased frequency corresponds to accelerated deformation). The data fusion module 6 uses Kalman filtering to fuse real-time trend and historical data: the state equation is x(k)=A×x(k-1)+w(k) (A is the state transition matrix, taken as 0.99; w is the process noise, variance 0.01), and the observation equation is z(k)=H×x(k)+v(k) (H is the observation matrix, taken as 1; v is the observation noise, variance 0.1). The fused data is obtained through a prediction-update loop, and then the trend model is updated using the least squares method (e.g., S(t)=a×t²+b×t+c, with real-time correction of the coefficients a, b, and c), outputting dynamic characteristic estimates (e.g., current acceleration a, stability coefficient 1 / (1+|a|)). The early warning trigger module 7 extracts the acceleration change component Δa and compares it with a preset threshold (set according to the "Highway Subgrade Design Specification", with the earthen subgrade taken as 0.5mm / d²). When the threshold is exceeded, an early warning signal sequence is generated (format: time + location + Δa + risk level), triggering an alarm.The prediction report module 8 uses an LSTM neural network to predict the trend for the next 30 days: the input layer is the settlement data of the past 90 days, there are 2 hidden layers (64 neurons each), and the output layer is the daily settlement amount. It is trained by the Adam optimizer (learning rate 0.001, 1000 iterations) and the prediction results are combined to generate a risk assessment report (including settlement curve, risk level classification and reinforcement suggestions).

[0019] Reference Figure 1 As shown: Fiber optic sensor network 1 is a distributed fiber optic sensor array containing 30 fiber optic sensing units. Each unit is a G.652D single-mode fiber with an outer layer of 0.5mm polyimide coating and a 2mm stainless steel armor layer. The tensile strength is ≥1000MPa, and the temperature resistance is -40℃ to 80℃. Along the longitudinal direction of the roadbed (road extension direction), 10 units are arranged parallel to the centerline, spaced 5m apart, located 0.5m below the top surface of the roadbed. The fiber axis deviation from the longitudinal direction is ≤1°, and they are fixed to the roadbed soil with plastic clips. Along the transverse direction (vertical longitudinal direction), 10 units are arranged vertically, spaced 3m apart, covering the entire width of the roadbed (4 rows for 12m roadbeds), located 2m below the top surface. The units are fixed with insulating clamps at intersections with the longitudinal fibers to prevent compression and wear. Along the depth direction, 10 units are arranged in 4 layers (0.5m, 2m, 4m, 6m), vertically embedded into the foundation bearing layer. Each layer of units is staggered in a quincunx pattern, spaced 4m apart, and fixed at the bottom with concrete anchors. All units are connected in series via fiber optic fusion splicing, with a splicing loss ≤0.1dB. Heat shrink tubing is used to seal the splices, forming a mesh structure. The main input / output terminals are led out from the left edge of the roadbed (0.5m from the shoulder) and connected to a waterproof junction box. The input terminal connects to the laser transmitter of data acquisition module 2 via an FC / APC connector, and the output terminal connects to the optical signal receiver via a similar connector. Stress-strain sensitive areas include: the interface between the roadbed and the base course (horizontal unit), the weak compaction zone in the middle of the roadbed (longitudinal unit), the roadbed-subgrade transition zone (depth unit), and the slope toe (end of the transverse unit).

[0020] The implementation principle is as follows: A distributed fiber optic sensor array achieves full-domain monitoring through three-dimensional deployment, utilizing the Rayleigh scattering effect to sense deformation: when the roadbed settles, the optical fiber undergoes strain along with the soil, causing a frequency shift in the scattered light (Δf=2×n×ε×f0 / c, where n is the refractive index of the optical fiber (1.468), ε is the strain, f0 is the incident light frequency (193 THz), and c is the speed of light). The frequency shift is linearly related to the strain. The deformation location is located using optical time-domain reflectometry (OTDR): a laser pulse propagates in the optical fiber, and the scattered light returns along the original path. The distance to the deformation point is calculated based on the round-trip time difference (positioning accuracy ±0.5m). The mesh-like serial structure ensures that the optical signal covers all units. Data acquisition module 2 analyzes the signal delay differences at different locations (Δt=ΔL / c, where ΔL is the change in optical fiber length) and converts it into a strain distribution (ε=ΔL / L0, where L0 is the initial length), thereby reflecting the degree of settlement in different areas of the roadbed. The elements in sensitive areas are prioritized to capture minute deformations (minimum monitored strain 0.1 με), providing high-resolution raw data for subsequent analysis and ensuring that early settlement trends are accurately identified.

[0021] Reference Figure 1 As shown: Data acquisition module 2 includes an optical signal receiver, an analog-to-digital converter (ADC), and a data buffer unit. The optical signal receiver is an InGaAs PIN photodiode with a response wavelength of 1310nm and a receiving sensitivity of -45dBm. Its input is connected to the output of the fiber optic network via an FC / PC fiber optic connector. The connector is encased in a metal protective shell and secured to the left panel of the monitoring box with screws. The receiver itself is mounted on an aluminum alloy bracket with a heat sink, 5cm away from the panel to ensure heat dissipation. The ADC is a 16-bit ADS8364 chip with a sampling rate of 2MHz. Its analog input is connected to the receiver output via a 50Ω coaxial cable (BNC connector), 30cm in length. The chip is soldered to the left side of the main circuit board, 10cm vertically from the receiver. The grounding terminal is directly connected to the circuit board's ground plane to reduce noise. The data buffer unit is a 256GB industrial-grade SATA solid-state drive with a read / write speed of 300MB / s. It is connected to the ADC output via a SATAIII data cable with a metal clip. The hard drive is secured to the right side of the circuit board using anti-vibration feet, 15cm horizontally from the ADC. The output of the buffer unit is connected to the signal processing module 3 via a USB 3.0 data cable. The data cable is laid along the cable groove on the inner wall of the monitoring box, and the cable groove is filled with insulating cotton.

[0022] The implementation principle is as follows: The optical signal receiver converts the delayed sequence of optical signals transmitted through the optical fiber (an analog signal where light intensity changes over time) into a 0-5V voltage signal. This weak signal (corresponding to 1mV voltage for a -40dBm optical signal) is amplified by an internal low-noise amplifier (1000x gain) to a range recognizable by the analog-to-digital converter (ADC). The ADC quantizes the voltage signal at a 2MHz sampling rate: each sampling point corresponds to a 1 / 2MHz = 500ns time interval, converting the continuous voltage signal into a digital quantity of 0-65535 (16-bit resolution), with a quantization error ≤0.0015%. The data buffer unit employs a cyclic overwrite storage mechanism: a 1-hour buffer period is set (2MHz × 3600s = 7.2 × 10). 9 The system (approximately 28GB) automatically overwrites the oldest data once full, while simultaneously transmitting data to the signal processing module 3 at a rate of 100MB / s via a USB 3.0 interface to ensure no data loss. The total latency of the entire process is <10ms, meeting real-time requirements and providing a precise digital signal foundation for constructing the initial latency distribution.

[0023] Reference Figure 1 As shown: The model storage unit is a 128GB industrial-grade TF card (U3 level), which is connected to the ARM processor main control board of the model matching module 4 via a surface-mount card slot. The card slot is soldered to the upper right corner of the motherboard. After the TF card is inserted, it is locked by a clip, with a distance of <2mm from the motherboard. The model matching module 4 is an STM32H743 processor (1GHz), which is fixed to a phenolic resin bracket inside the monitoring box by an insulating support. The processor and the TF card are connected by a 5cm long PCB differential trace to reduce signal interference. The model library is stored according to a two-level classification: the first-level classification is roadbed type (earth roadbed, crushed stone roadbed, etc., 4 types), based on the "Highway Engineering Quality Inspection and Evaluation Standard"; the second-level classification is geological conditions (soft soil foundation, etc., 4 sub-categories), classified according to the foundation bearing capacity (<100kPa is soft soil). Each subclass contains 8 models, corresponding to 4 stages of settlement (initial / stable / accelerated / dangerous). Each model includes characteristic parameters: fluctuation frequency range (e.g., 0.2-0.5Hz in the acceleration stage), amplitude growth rate (>5% / d), and delay change rate (>2ns / m·d). The parameters are obtained through statistical analysis of settlement data from 100+ projects (95% confidence interval). The model files are in binary format (5MB / file), and the storage path is classified according to "subgrade type / geological conditions / stage".

[0024] The implementation principle is as follows: the model storage unit provides a stable model carrier, and the industrial-grade TF card is adaptable to environments ranging from -40℃ to 85℃. During model matching, the processor first determines the current type (e.g., earthen roadbed) and geological conditions (e.g., soft soil foundation) through the roadbed design file, and indexes the corresponding sub-type model; then it extracts the parameters of the fluctuation feature sequence (frequency, amplitude growth rate, etc.) and calculates the Euclidean distance with the model parameters (distance = √Σ(eigenvalue - model value)²). The smaller the distance, the higher the matching degree (matching degree = 1 - distance / maximum distance); when the matching degree > 85%, it is determined that the settlement has entered the corresponding stage (e.g., acceleration stage), and an acceleration risk label is output. Close-range differential connections ensure a model read rate of 50MB / s and an indexing time of <100ms, achieving real-time matching and providing a standardized basis for risk identification.

[0025] Reference Figure 1 As shown: Data fusion module 6 includes an FPGA filtering unit (Xilinx Spartan-6) and an STM32H743 trend model update unit. Both are soldered onto the same PCB board, 8cm apart, with a grounding isolation strip in between, and connected via a 20-bit parallel bus (equal-length traces to ensure synchronization). The first input of the filtering unit is connected to the frequency analysis module 5 via a shielded Ethernet cable, and the second input is connected to the historical database via a single-mode optical fiber (the fiber is run through a metal corrugated pipe). The output of the update unit is connected to the early warning trigger module 7 via a 26-core flat cable, laid along a guide groove, 10cm away from the power line. Kalman filter parameters: process noise covariance Q = 0.01 (reflecting model uncertainty), measurement noise covariance R = 0.1 (reflecting sensor noise), initial state X0 = 0, covariance P0 = 1. The trend model is a quadratic polynomial S(t) = a × t² + b × t + c (a is acceleration, b is initial velocity, and c is initial settlement). The update unit corrects the parameters every hour using the least squares method: based on the residual between the fused data and the model prediction (e = measured - predicted), the Σe² is minimized. When the residual is greater than 0.05 mm / d, parameter adjustment is triggered.

[0026] The implementation principle is as follows: Kalman filtering fuses data through a "prediction-update" cycle: in the prediction stage, the current state is estimated based on historical data (x(k|k-1)=A×x(k-1|k-1)), and in the update stage, the estimate is corrected by combining real-time data (x(k|k)=x(k|k-1)+K(k)×(z(k)-H×x(k|k-1)), where K is the Kalman gain), filtering out noise (improving the signal-to-noise ratio by 30%). The trend model update unit dynamically adjusts parameters based on the fused data: if recent settlement accelerates, the value of 'a' increases; if it tends to stabilize, 'a' approaches 0. By correcting the model in real time, the estimated values ​​of dynamic characteristics (such as the current acceleration 'a' and the stability coefficient 1 / (1+|a|)) accurately reflect the roadbed state, providing a reliable data foundation for early warning.

[0027] Reference Figure 1 As shown: The early warning triggering module 7 includes an LMV7219 threshold comparison unit and an alarm execution unit. The comparison unit is a high-speed voltage comparator (response time < 50ns). The signal input terminal is connected to the data fusion module 6 via a twisted-pair shielded cable (the shield is grounded). The reference voltage terminal is connected to a 3.3V reference source via an adjustable resistor. The preset threshold corresponds to a voltage of 1V (equivalent acceleration 0.5mm / d², set according to the "Highway Subgrade Design Specification"). The alarm execution unit: The audible and visual alarm is a 12V LED buzzer (800cd red light, 110dB@1m), which is fixed to the outer surface of the top of the monitoring box with screws. The wires are connected to the output terminal of the comparison unit along the cable tray on the top of the box. The 4G remote communication module (EC20) is fixed to the right side wall inside the monitoring box, 12cm away from the comparison unit, and connected via DuPont wires. The antenna extends vertically 1m outside the box (signal strength > -80dBm).

[0028] The implementation principle is as follows: The threshold comparison unit compares the acceleration change component (converted to a 0-3.3V voltage) with a 1V threshold; when the threshold is exceeded, a 3.3V trigger signal is output. After receiving the signal, the alarm execution unit activates the audible and visual alarm, which flashes and sounds at a frequency of 1Hz (local warning); the 4G module sends a JSON-formatted warning signal ({"Time":"2025-11-10 12:00:00","Location":"K1+200","Acceleration":"0.6mm / d²","Level":"High"}) to three preset terminals, retransmitting it every 1 minute until the warning is lifted. This dual alarm system ensures timely response to risks and shortens the handling time.

[0029] Reference Figure 1 As shown: The prediction report module 8 includes an NVIDIA Jetson Nano algorithm processing unit (1.43GHz, 4GB memory) and a report output unit. The processing unit receives warning signals via a 470MHz LoRa module and incorporates two algorithms: an LSTM neural network (10 neurons in the input layer, 2 hidden layers each with 32 neurons, and 1 neuron in the output layer) for non-linear trends, and ARIMA (p=3, d=1, q=2) for linear / periodic trends, automatically switching via R² test (LSTM is used when R² < 0.8). The output unit includes a 21.5-inch display (1920×1080) and a USB printer interface. The display is embedded in the front panel, and the interface is located on the right side (with a waterproof cover). LSTM training parameters: Adam optimizer (learning rate 0.001), 1000 iterations, input of nearly 90 days of data, output of predicted values ​​for the next 30 days; ARIMA eliminates non-stationarity through differencing and captures trends using autoregressive terms.

[0030] The implementation principle is as follows: The algorithm processing unit receives early warning signals and historical data (last 6 months), normalizes it, and inputs it into the algorithm: LSTM captures long-term dependencies (such as seasonal fluctuations) through the memory unit, while ARIMA is suitable for short-term linear prediction. After the prediction results are verified by RMSE (<0.5mm), they are converted into visual charts (settlement rate curve, cumulative settlement bar chart), and combined with rules to generate text descriptions (such as "30-day cumulative settlement 60mm, high risk, grouting reinforcement recommended"), which are displayed on a screen and can be printed, providing intuitive basis for decision-making.

[0031] Reference Figure 1 As shown: The power supply module adopts a dual-mode system with solar power as the primary source and AC power as a secondary source. Solar unit: A 200W monocrystalline silicon solar panel (conversion efficiency 23%) is mounted on the top of the monitoring box (facing south) via a 35° tilt bracket. It connects to a 12V / 20AMPPT charge / discharge controller (efficiency 95%) via an MC4 waterproof connector. The controller is connected to a 12V / 100Ah lead-acid battery (placed on an insulated tray at the bottom of the monitoring box, 20cm away from the controller) via a 2.5mm² wire. AC power unit: AC220V is supplied through an outdoor waterproof socket. After conversion by a 12V / 3A switching power supply, it is connected in parallel with the solar output via a diode (to prevent reverse power supply). The switching power supply is placed in a rainproof box outside the box. The controller output is distributed to each module (0.5A for fiber optic network, 22A for data acquisition module, etc.) via multiple 1.5mm² power cables. The cables are independently arranged along PVC cable trays, with a distance of ≥5cm from the data cables.

[0032] The implementation principle is as follows: the solar panels generate an average of ≥1kWh of electricity per day. The MPPT controller tracks the maximum power point, improving efficiency by 20%. The energy is stored in a battery (enough for 3 days of power supply during cloudy or rainy weather). The controller has overcharge (14.5V) and over-discharge (10.5V) protection, extending battery life (>3 years). Mains power automatically switches on when solar energy is insufficient (seamless switching via unidirectional diode conduction). Multiple power supplies avoid interference, and the wire diameter meets current requirements (e.g., 1.5mm² wire for a 2A module, with a current carrying capacity of 3A), ensuring continuous and stable operation of the device.

[0033] A method for monitoring settlement and deformation in roadbed engineering, characterized by comprising: S1: Real-time optical signal propagation data is collected by deploying an optical fiber sensor network 1 in the roadbed structure, and an optical signal delay sequence is obtained to form an initial delay distribution, which reflects the influence of roadbed deformation on signal propagation time; S2: Based on the initial delay distribution, a signal analysis method is used to process the weak fluctuations, determine the fluctuation characteristic sequence, and obtain the deformation reflection index. The deformation reflection index is used to quantify the degree of deformation of the roadbed structure. S3: Match the wave characteristic sequence with the preset deformation model. If the matching degree is higher than the preset threshold, it is judged as potential settlement acceleration, and an acceleration risk label is obtained. The acceleration risk label indicates the risk of aggravated roadbed settlement. S4: Based on the accelerated risk label, frequency components are extracted using a transformation method, and the dominant mode in the frequency components is determined to identify the settlement velocity change trend, which describes the evolution direction of the roadbed deformation rate; S5: Based on the settlement velocity change trend, historical monitoring data are integrated, and a filtering method is used to update the trend model to obtain dynamic characteristic estimates. The dynamic characteristic estimates integrate real-time and historical information to evaluate the roadbed stability. S6: Extract the acceleration change component from the dynamic characteristic estimate. If the acceleration change component exceeds a preset threshold, trigger the alarm mechanism to obtain an early warning signal sequence. The early warning signal sequence records roadbed risk warning information. S7: Based on the warning signal sequence, apply the prediction algorithm to optimize the future trend, determine the settlement evolution path and generate a risk assessment report, which summarizes the potential hazards and development predictions of the roadbed.

[0034] In the above method, based on the roadbed design and geological data, distributed optical fiber sensing units are first embedded in the sensitive area in three dimensions: longitudinal (0.5m below the top surface, 5m interval), lateral (2m below the top surface, 3m interval), and depth (0.5-6m layered to the bearing layer). These units are thermally fused into a mesh and led out to a waterproof junction box to initialize the acquisition of reference optical signal delay data. Then, the photodiode of data acquisition module 2 captures the delay sequence of deformed optical signals, which is amplified, converted into a digital signal by a 16-bit 2MHz ADC, buffered, and transmitted to signal processing module 3. Signal processing module 3 constructs the initial delay distribution, uses a 5-layer soft thresholding method with a db4 wavelet base for noise reduction, extracts amplitude, frequency, and phase to form a wave feature sequence, and converts it into a deformation reflection index using Δε=0.01με / ns×Δt. Model matching module 4 calls a preset model library according to the roadbed type and geology, compares the feature sequences using cosine similarity (threshold 85%), and outputs an acceleration risk label. Frequency analysis module 5 performs FFT on the labeled sequence, selecting those with an energy percentage > 60%. The dominant frequency is used to determine the settlement velocity trend; the data fusion module 6 uses Kalman filtering (Q=0.01, R=0.1) to fuse real-time trends and historical data, and updates the dynamic characteristic estimate of the quadratic polynomial trend model output; the early warning trigger module 7 extracts the acceleration component, and triggers local audible and visual alarms and remote 4G early warning when the acceleration exceeds the 0.5mm / d² threshold; after receiving the signal, the prediction report module 8 selects LSTM (R²<0.8) or ARIMA (p=3, d=1, q=2) algorithm according to the R² value to predict the settlement path in the next 30 days and generate a risk assessment report with visualization charts and suggestions, realizing fully automated monitoring and risk management.

[0035] The working principle of this device is as follows: Fiber optic sensor network 1 generates micro-strain as the roadbed deforms, outputting a delayed optical signal sequence containing deformation information; data acquisition module 2 receives the sequence and converts it into a digital signal, constructing an initial delay distribution. Signal processing module 3 uses wavelet transform for noise reduction, extracts the fluctuation feature sequence, and converts it into a deformation reflection index; model matching module 4 compares the feature sequence with a preset deformation model, and outputs an acceleration risk label if the matching degree exceeds a threshold. Frequency analysis module 5 performs Fourier transform on the labeled sequence, determining the settlement velocity trend based on the dominant frequency; data fusion module 6 uses Kalman filtering to fuse the trend with historical data, updating the trend model and outputting dynamic characteristic estimates. Early warning triggering module 7 extracts the acceleration component, triggering local audible and visual warnings and remote early warnings if the acceleration exceeds a threshold; the prediction module uses LSTM or ARIMA algorithms to predict the future settlement path and generate a risk assessment report.

[0036] The working principle of this device has been explained through the above embodiments. These embodiments merely illustrate several implementation methods of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A roadbed engineering settlement and deformation monitoring device, characterized in that: The system includes a fiber optic sensor network (1), deployed inside the roadbed structure, which collects real-time optical signal propagation data corresponding to roadbed deformation and outputs optical signal delay sequence; a data acquisition module (2), which is signal-connected to the fiber optic sensor network (1), receives the optical signal delay sequence and forms an initial delay distribution; a signal processing module (3), which is communicatively connected to the data acquisition module (2), processes the weak fluctuations in the initial delay distribution, determines the fluctuation characteristic sequence and outputs deformation reflection indicators; a model matching module (4), which is electrically connected to the signal processing module (3), has a preset deformation model, matches the fluctuation characteristic sequence with the model, and outputs an acceleration risk label if the matching degree exceeds a preset threshold; and a frequency analysis module (5), which is signal-connected to the model matching module (4), extracts frequency components based on the acceleration risk label, determines the dominant mode and outputs the settlement velocity change trend. The data fusion module (6) is connected to the frequency analysis module (5) and the historical database respectively. It integrates the changing trend and historical data, updates the trend model using a filtering method, and outputs the dynamic characteristic estimate. The early warning trigger module (7) is electrically connected to the data fusion module (6). It extracts the acceleration change component. If the value exceeds the preset threshold, it triggers an alarm and outputs an early warning signal sequence. The prediction report module (8) is connected to the early warning trigger module (7). It applies a prediction algorithm to optimize the future trend, determines the settlement evolution path, and generates a risk assessment report.

2. The roadbed engineering settlement and deformation monitoring device according to claim 1, characterized in that: The optical fiber sensor network (1) is a distributed optical fiber sensor array, which is deployed in layers along the longitudinal, transverse and depth directions of the roadbed. The distributed optical fiber sensor array includes several optical fiber sensing units, and each optical fiber sensing unit is evenly embedded in different stress and strain sensitive areas of the roadbed structure.

3. The roadbed engineering settlement and deformation monitoring device according to claim 1, characterized in that: The data acquisition module (2) includes an optical signal receiver, an analog-to-digital converter and a data buffer unit. The optical signal receiver receives the optical signal delay sequence output by the optical fiber sensor network (1), converts it into a digital signal by the analog-to-digital converter, stores it through the data buffer unit and outputs it to the signal processing module (3).

4. The roadbed engineering settlement and deformation monitoring device according to claim 1, characterized in that: It also includes a model storage unit, which is connected to the model matching module (4) and is used to store the preset deformation model. The preset deformation model is a settlement acceleration feature model library constructed based on different roadbed types and geological conditions.

5. The roadbed engineering settlement and deformation monitoring device according to claim 1, characterized in that: The filtering method in the data fusion module (6) is Kalman filtering or particle filtering. The data fusion module (6) also includes a trend model update unit, which is used to dynamically adjust the trend model parameters based on the fused real-time data and historical data.

6. The roadbed engineering settlement and deformation monitoring device according to claim 1, characterized in that: The early warning triggering module (7) includes a threshold comparison unit and an alarm execution unit. The threshold comparison unit is used to compare the extracted acceleration change component with a preset threshold. The alarm execution unit includes an audible and visual alarm and a remote communication module, which are used to synchronously output local audible and visual alarms and remote early warning information.

7. The roadbed engineering settlement and deformation monitoring device according to claim 1, characterized in that: The prediction algorithm in the prediction report module (8) is a neural network prediction algorithm or a time series prediction algorithm. The prediction report module (8) also includes a report output unit, which outputs the risk assessment report in the form of visual charts and text descriptions.

8. The roadbed engineering settlement and deformation monitoring device according to claim 1, characterized in that: It also includes a power supply module, which is a solar power supply unit or a mains power supply unit, used to provide stable power to the fiber optic sensor network (1), data acquisition module (2), signal processing module (3), model matching module (4), frequency analysis module (5), data fusion module (6), early warning triggering module (7) and prediction report module (8).

9. The method for monitoring settlement and deformation of roadbed engineering according to any one of claims 1-8, characterized in that: Includes the following steps: S1: Real-time optical signal propagation data is collected by deploying an optical fiber sensor network (1) in the roadbed structure, and an optical signal delay sequence is obtained to form an initial delay distribution, which reflects the influence of roadbed deformation on signal propagation time; S2: Based on the initial delay distribution, a signal analysis method is used to process the weak fluctuations, determine the fluctuation characteristic sequence, and obtain the deformation reflection index. The deformation reflection index is used to quantify the degree of deformation of the roadbed structure. S3: Match the wave characteristic sequence with the preset deformation model. If the matching degree is higher than the preset threshold, it is judged as potential settlement acceleration, and an acceleration risk label is obtained. The acceleration risk label indicates the risk of aggravated roadbed settlement. S4: Based on the accelerated risk label, frequency components are extracted using a transformation method, and the dominant mode in the frequency components is determined to identify the settlement velocity change trend, which describes the evolution direction of the roadbed deformation rate; S5: Based on the settlement velocity change trend, historical monitoring data are integrated, and a filtering method is used to update the trend model to obtain dynamic characteristic estimates. The dynamic characteristic estimates integrate real-time and historical information to evaluate the roadbed stability. S6: Extract the acceleration change component from the dynamic characteristic estimate. If the acceleration change component exceeds a preset threshold, trigger the alarm mechanism to obtain an early warning signal sequence. The early warning signal sequence records roadbed risk warning information. S7: Based on the warning signal sequence, apply the prediction algorithm to optimize the future trend, determine the settlement evolution path and generate a risk assessment report, which summarizes the potential hazards and development predictions of the roadbed.

Images