Rainfall data analysis highway landslide early warning system based on intelligent AI

Abstract

The invention belongs to the technical field of intelligent AI for rainfall analysis and landslide early warning, and discloses an intelligent AI-based rainfall data analysis highway landslide early warning system, which comprises a slope hydrologic monitoring module, a slope damage detection module, an anti-slide pile dynamic monitoring module, a landslide risk analysis module and a risk early warning output module, a slope landslide risk prediction model is constructed by collecting rainfall, water seepage amount, slope structure damage state parameters, pressure difference value of two sides of an anti-slide pile and pile-soil gap variation key data in real time; and the pressure difference value of the two sides of the anti-slide pile, the structural damage state parameters and the rainfall capacity are subjected to cross dynamic association through time sequence characteristics, the water seepage amount and the pile-soil gap variation are fused, and the landslide risk level is output through the dynamic risk amplification factor and the soil saturation critical index. The real-time performance and accuracy of landslide early warning can be remarkably improved, and the method is suitable for highway slope safety monitoring under complex geological conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent AI for rainfall analysis and landslide early warning, and relates to a rainfall data analysis and highway landslide early warning system based on intelligent AI. Background Art

[0002] Slope stability is an increasingly prominent issue during the construction and operation of modern highways, especially in areas prone to heavy rainfall. Rainfall-induced landslides pose a serious threat to traffic safety, causing significant economic losses and potentially leading to significant casualties. Therefore, effectively monitoring and providing early warning of slope landslide risks is crucial to ensuring traffic safety.

[0003] Traditional landslide monitoring methods primarily rely on single rain gauge data to trigger early warning mechanisms. For example, the invention patent with publication number CN114093138A proposes an embedded rainfall monitoring terminal and monitoring method for landslide monitoring and early warning. This solution converts rainfall into electrical signals through a rainwater collection funnel, filter, and signal triggering device. The MCU control unit processes the signals to generate rainfall values ​​and warning levels, offering advantages such as flexible deployment and low cost. Although this solution can effectively reflect the occurrence of slope landslides under heavy rainfall conditions, long-term rainwater erosion can also lead to landslide risks due to cumulative effects or soil weakening, even with relatively low rainfall. Relying solely on single rainfall monitoring data lacks comprehensive consideration of key factors such as seepage processes, slope structural damage, and stress changes in support structures, making it difficult to fully reveal the true physical mechanisms of landslide occurrence.

[0004] Furthermore, with the development of deep learning technology, some research has attempted to enhance early warning capabilities through big data modeling. For example, the invention patent with publication number CN118982902A proposes a rainfall-induced landslide early warning system based on deep learning models. This system uses a recurrent neural network to process rainfall and monitoring time series data, and a convolutional neural network to process image data. Using a deep learning framework to construct a model structure, combined with GIS technology, it assesses landslide susceptibility and predicts landslide hazard. This system achieves dynamic assessment of landslide risk and multi-channel early warning issuance, significantly improving the accuracy and timeliness of predictions. This system establishes an early warning model based on statistical correlation, ignoring the actual physical mechanisms of the slope's internal mechanical response process. This makes it difficult to adapt to environmental changes, thus affecting the scientific nature and reliability of the early warning results. Summary of the Invention

[0005] In view of this, in order to solve the problems raised in the above background technology, a highway landslide early warning system based on rainfall data analysis of intelligent AI is proposed.

[0006] The purpose of the present invention can be achieved through the following technical solutions: a highway landslide early warning system based on intelligent AI rainfall data analysis, including the following modules: a slope hydrological monitoring module, which is used to deploy rain gauges on the slope surface to collect the rainfall of the target slope in a continuous period in real time, and to deploy piezometers at the foot of the slope to synchronously obtain the seepage amount.

[0007] The slope damage detection module is used to scan the target slope surface with a thermal infrared high-definition imager to obtain images, and analyze the porosity and crack distribution characteristics in the image to output the current structural damage status parameters of the slope.

[0008] The anti-slide pile dynamic monitoring module is used to detect the pressure difference on both sides of the anti-slide pile and the change in the pile-soil gap in real time during rainfall through patch pressure sensors and displacement sensors arranged on the soil-facing and soil-removing sides of the anti-slide pile.

[0009] The landslide risk analysis module is used to construct a slope landslide risk prediction model based on rainfall, seepage, structural damage status parameters, pressure difference on both sides of the anti-slide piles, and change in the pile-soil gap. The pressure difference on both sides of the anti-slide piles, structural damage status parameters and rainfall are cross-dynamically correlated through time series characteristics, and the seepage and pile-soil gap changes are integrated to output the landslide risk level.

[0010] The risk warning output module is used to generate and publish landslide warning information of corresponding levels according to the preset risk threshold matching the landslide risk level.

[0011] Compared with the existing technology, the beneficial effects of the present invention are as follows: (1) The present invention collects key data such as rainfall, seepage, slope structure damage state parameters, pressure difference on both sides of the anti-slide pile and pile-soil gap change in real time, reducing the risk of landslide caused by cumulative effect or soil weakening when rainfall is small, breaking through the limitations of traditional single rainfall monitoring, and completely covering the landslide chain driving factors of rainfall, seepage, damage and support response.

[0012] (2) The present invention dynamically correlates the pressure difference on both sides of the anti-slide pile, the structural damage state parameters and the rainfall through time series characteristics, integrates the seepage volume and the change in the pile-soil gap, and establishes an interpretable physical-data dual-driven decision model through the dynamic risk amplification factor and the soil saturation critical index. This overcomes the one-sided attribution problem caused by simple statistical association in pure data-driven models, and provides a transparent mechanism for risk assessment of highway landslides. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 This is a schematic diagram of the connection of various modules of the system of the present invention.

[0015] Figure 2 This is a diagram showing the implementation scenario of landslide monitoring on highway slopes according to the present invention.

[0016] Figure 3 This is a diagram of the coupling effect between α and γ in the model of the present invention.

[0017] Figure numerals: 1—pressure sensor, 2—anti-slide pile, 3—displacement sensor, 4—piezometer, 5—thermal infrared high-definition imager, 6—rain gauge, 7—slope, 8—slope foot, 9—highway. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] See also Figure 1 As shown, the present invention provides a highway landslide warning system based on intelligent AI-based rainfall data analysis. The system includes a slope hydrological monitoring module, a slope damage detection module, a dynamic monitoring module for anti-slide piles, a landslide risk analysis module, and a risk warning output module. The slope hydrological monitoring module, slope damage detection module, and dynamic monitoring module for anti-slide piles simultaneously collect real-time data, transmit the collected data to the landslide risk analysis module, which then processes the data and sends it to the risk warning output module.

[0020] The slope hydrological monitoring module collects the rainfall of the target slope in a continuous period in real time by laying rain gauges on the slope surface, and synchronously obtains the seepage amount by laying piezometers at the slope foot.

[0021] In order to accurately understand the impact of rainfall on slope stability, it is first necessary to reasonably deploy rain gauges on the slope surface to collect real-time rainfall data on the target slope over a continuous period of time.

[0022] The target slope surface is divided into equal intervals along the slope length and slope width to form regularly distributed rectangular grids.

[0023] Specifically, the slope length direction is adjusted according to the slope height. The high slope takes the largest value and the equal spacing can be selected in The width of the slope is the same as the length of the slope, forming a square grid.

[0024] It should be noted that by dividing the slope into regularly distributed square grids with equal spacing, the entire slope surface can be evenly covered, avoiding the neglect of certain areas, thereby improving the comprehensiveness and accuracy of data collection.

[0025] Rain gauges are placed at the intersection of each rectangular grid.

[0026] It should be noted that the rain gauge can directly monitor the rainfall on the target slope in real time and continuously over a continuous period of time, and calculate the rainfall intensity every hour, thereby providing more accurate and timely data.

[0027] When constructing a slope landslide risk prediction model, gridded rain gauges can help to perform spatial interpolation and trend analysis more accurately, thereby improving the model's prediction ability.

[0028] The rainfall amount of each rain gauge is continuously collected in real time every minute, and the rainfall intensity is accumulated and calculated every hour.

[0029] It should be noted that by collecting data once a minute, it is possible to accurately record rainfall peaks and fluctuations, and quickly capture rapid changes in rainfall. In the face of extreme weather events such as local heavy rainfall or short-term rainstorms, high-frequency data collection can more realistically reflect complex and changeable meteorological conditions, providing a reliable basis for evaluating their impact on slope stability.

[0030] Multiple piezometer monitoring points are pre-buried at equal intervals along the slope foot to obtain the pore water pressure values ​​of each piezometer monitoring point in real time and continuously.

[0031] It should be noted that the piezometer can more directly reflect the changes in groundwater level caused by rainfall. Through multi-point measurement, it can ensure that the seepage monitoring points evenly cover the entire slope foot, avoid missing local seepage anomalies, and obtain more accurate and representative pore water pressure data.

[0032] The choice of spacing will take into account factors such as the length of the slope, the complexity of the geological conditions, and the expected monitoring accuracy.

[0033] According to the saturated bulk density of the soil at the location where the piezometer is installed, the ratio of the measured pore water pressure to the saturated bulk density of the soil at the location is used as the equivalent head height.

[0034] It's important to note that saturated soil bulk density refers to the mass of soil per unit volume when it is completely saturated with water. It reflects the density or weight of soil in its saturated state, meaning that all pores are filled with water. Saturated soil bulk density is typically determined through laboratory testing, but can also be estimated based on soil type and moisture content.

[0035] The incremental change of the water head under rainfall is obtained by subtracting the equivalent water head height from the water head height in the initial stable state.

[0036] The cumulative seepage volume in the corresponding period is obtained by multiplying the effective infiltration area of ​​the monitoring area, the incremental change value of the water head under rainfall, and the saturated bulk density of the soil.

[0037] It should be noted that the effective permeability area of ​​the monitoring area is the area where all piezometers are installed as its effective permeability area.

[0038] Calculating water seepage in this way enables data from different locations to be compared and analyzed under the same standard, which is conducive to establishing a unified risk assessment model and early warning mechanism.

[0039] Since there is often a toe structure between the slope and the highway, continuous rainfall will cause rainwater to seep down along the slope surface and accumulate at the toe, thereby affecting the stability of the toe. Therefore, the detection of seepage volume is particularly critical.

[0040] Water seepage is a key indicator of water saturation within a slope. As rainfall increases, pore water pressure rises, potentially reducing the effective stress between soil particles and increasing the risk of landslides. Simultaneously monitoring these two factors allows for more precise analysis of the relationship between rainfall and slope stability, providing key input parameters for developing landslide risk prediction models and improving the accuracy and timeliness of early warning systems.

[0041] The slope damage detection module scans the target slope surface with a thermal infrared high-definition imager to obtain an image, analyzes the porosity and crack distribution characteristics in the image, and outputs the current structural damage state parameters of the slope.

[0042] It should be noted that thermal infrared imaging is a non-contact detection technology, which means it can quickly scan large areas without disturbing the slope structure. It is suitable for large-scale monitoring of slopes along highways and can capture the distribution of thermal radiation on the slope surface, thereby reflecting the temperature differences between different areas. Structural damage such as cracks and pores usually leads to changes in local thermal conductivity, which will appear as different temperature patterns on thermal images.

[0043] The thermal infrared high-definition imager is used to scan the target slope surface at multiple angles to obtain a thermal infrared image sequence of the temperature distribution characteristics of the target slope surface.

[0044] It should be noted that a single angle of scanning may not cover all areas, especially when the slope terrain is complex or there are obstructions. By scanning from multiple angles, these visual blind spots can be reduced or even eliminated, ensuring comprehensive monitoring of the entire slope surface. Scanning at different angles can obtain the temperature distribution of the slope surface at different viewing angles, providing more abundant information for subsequent analysis.

[0045] The acquired thermal infrared image is preprocessed, and the temperature values of each pixel point of the preprocessed thermal infrared image are compared with the preset reference temperature field point by point, and the pixels with a difference exceeding a preset temperature difference threshold are marked.

[0046] It should be noted that the preprocessing includes radiation correction, geometric correction, noise removal, contrast enhancement, and edge detection. The image after preprocessing has higher quality and accuracy, and can more accurately reflect the real temperature distribution of the slope surface, thereby improving the accuracy of the structural damage state parameter calculation.

[0047] The temperature values of each pixel point are compared with the preset reference temperature field point by point, and the pixels with a difference exceeding a preset temperature difference threshold are marked, mainly to highlight the areas showing abnormal temperature changes. These areas may be due to structural damage such as cracks, pores, etc. leading to differences in local heat conduction performance, providing basic data for further connected component analysis, morphological skeleton extraction, etc.

[0048] The spatially continuous marked pixels are subjected to connected component analysis, and a closed boundary is generated and extracted as a pore feature area.

[0049] Specifically, the connected component analysis has the following specific steps: first, the image is converted into a binary image, where the pixels with abnormal temperature difference are marked as foreground, which is usually white and has a value of 1, and the rest is background, which is usually black and has a value of 0. Then it is determined that the pixels are considered to be connected to each other, four-connectedness considers the adjacent pixels in the up, down, left and right directions, while eight-connectedness also includes eight adjacent pixels in the diagonal direction. A connected component labeling algorithm using depth-first search is used to traverse each pixel in the image, and each independent connected component is assigned a unique identifier.

[0050] More specifically, after all the connected components are labeled, a contour tracing algorithm can be used to move along the edge of the connected domain and record the position of the boundary points, thereby forming a closed contour.

[0051] For each identified pore feature area, its geometric properties such as area, perimeter, shape factor, etc. are further calculated, and certain rules are set according to application requirements, such as minimum or maximum size limits, to remove small noise points or non-target objects that do not meet the conditions, and to retain truly meaningful pore feature areas.

[0052] The total area of ​​the pore characteristic region is counted, and its proportion to the total area of ​​the scanning region is calculated as the porosity.

[0053] Specifically, for each identified pore feature, its area is determined by counting the number of pixels within the region. If the image resolution is known, the number of pixels can be converted to actual area units. The areas of all pore features are summed to obtain the total area of ​​the pore features within the entire scanned area.

[0054] Porosity is a key indicator of the void space within a slope. A higher porosity typically indicates more voids or cracks within the slope, which can reduce its overall stability. Therefore, calculating porosity not only accurately assesses the current condition of a slope but also provides an important scientific basis for subsequent risk management, the establishment of early warning mechanisms, and specific repair work. It can also directly reflect the current structural health of the slope.

[0055] The areas in the pore feature area with an aspect ratio greater than or equal to a preset ratio and a spatial overlap rate greater than or equal to a preset threshold in several consecutive frames of images are screened as continuous temperature anomaly zones.

[0056] It should be noted that the preset ratio, or aspect ratio threshold, for the porosity characteristic zone in slope monitoring can be set based on historical data. Generally, a larger aspect ratio may indicate potential cracks or linear structural anomalies, which may be early signs of increased landslide risk.

[0057] Specifically, the spatial overlap ratio refers to the degree of overlap between specific pore feature regions in several consecutive image frames. This is calculated as follows: For the same pore feature region in two or more consecutive image frames, first align them to ensure they are in the same coordinate system. Then, the ratio of the intersection area to the union area of ​​the pore feature regions in each pair of adjacent frames is calculated. This ratio is the spatial overlap ratio. If multiple frames are involved, the average of the spatial overlap ratios of all pairs of adjacent frames is used as the spatial overlap ratio for that region.

[0058] More specifically, a higher spatial overlap rate indicates that the area is relatively stable in the time series and is more likely to represent a real physical phenomenon rather than accidental noise. By analyzing historical data, we can understand the distribution of spatial overlap rates under normal circumstances and set an appropriate threshold based on actual conditions to distinguish meaningful continuous temperature anomaly bands from random fluctuations.

[0059] Larger aspect ratios often point to stress concentrations in specific directions, such as cracks, which can be precursors to landslides. Therefore, screening for pore features with large aspect ratios helps focus on areas most likely to cause problems.

[0060] Regions with high spatial overlap across consecutive frames indicate persistent changes rather than transient anomalies. This persistence is more likely to reflect actual geological processes rather than temporary measurement errors or environmental noise. Therefore, using spatial overlap as a screening criterion can help improve the accuracy of prediction models.

[0061] Morphological skeleton extraction is performed on the continuous temperature anomaly zone to generate a crack skeleton trajectory with a single pixel width, and the total length along the crack skeleton trajectory is counted.

[0062] The width of each position on the trajectory is measured and the maximum value is recorded. Cracks with a width exceeding the width threshold are marked as high-risk cracks.

[0063] Specifically, the width threshold can analyze the data of crack width in past landslide events to find the crack width range that ultimately led to landslides, and use this as a reference to set the threshold.

[0064] It's important to note that wider cracks generally indicate greater structural damage, suggesting potential stress concentration or material degradation within the slope material, all of which can accelerate landslides. Identifying high-risk cracks by setting a width threshold can help establish a more sensitive and effective early warning system. Promptly detecting and addressing cracks exceeding this width threshold can significantly reduce landslide risk and protect infrastructure safety.

[0065] The total length of high-risk cracks per unit area is calculated as the high-risk crack density, which is the crack distribution characteristic.

[0066] Specifically, the actual area of ​​the slope surface was recorded, and for all cracks marked as high-risk, the lengths were accumulated section by section along their skeleton trajectories using image processing software.

[0067] It should be noted that the high-risk crack density directly reflects the number and distribution density of high-risk cracks on the slope surface. A higher density generally indicates a higher landslide risk, indicating that there are more concentrated areas of structural damage within the slope.

[0068] The porosity and high-risk crack density are respectively calculated by ratioing the porosity and the high-risk crack density with the preset benchmark porosity and the preset benchmark crack density to obtain the dimensionless porosity index and the high-risk crack density index.

[0069] It should be noted that the preset baseline porosity and preset baseline crack density are usually determined by analyzing historical monitoring data of the slope under stable conditions to determine a representative porosity and crack density value. These values ​​are usually recorded during a period without significant geological changes or landslide events.

[0070] For some types of soil slopes, the preset base porosity can be set as The preset reference crack density can be set according to specific conditions, such as that the total length of high-risk cracks in each square meter is not more than 0.5 meters. These values are only examples, and in actual application, they should be adjusted according to detailed site investigation and professional evaluation.

[0071] The dimensionless index can simplify complex physical parameters, facilitate quick understanding and information transmission, eliminate the influence of different units, make data from different slopes or different periods of the same slope comparable, and when constructing a slope landslide risk prediction model, the dimensionless index can be used as one of the input features, which helps to improve the prediction ability and robustness of the model.

[0072] The porosity index and the high-risk crack density index are fused by weighting to generate a slope current structure damage state parameter.

[0073] Specifically, the weighting fusion generates a slope current structure damage state parameter , wherein represents a porosity influence weight coefficient, represents a high-risk crack density influence weight coefficient, represents a porosity index, represents a high-risk crack density index.

[0074] It should be noted that the porosity influence weight reflects the overall strength attenuation degree of the soil body caused by the development of pores, and the high porosity influence weight value is suitable for porous loose soil, and the porosity has a greater impact on stability. The high-risk crack density influence weight represents the direct triggering risk of crack propagation to landslides, and the high-risk crack density influence weight value is high, which is suitable for brittle rock slope, and cracks are easy to cause sudden collapse. Generally, the following condition is met , to ensure that the damage parameter is within a reasonable range.

[0075] For example, the clay has high porosity but strong ductility, and the crack propagation is slow, For example, it is detected that the porosity index is 1.2, indicating that the current porosity exceeds the reference value by 20%, the high-risk crack density index is 0.8, indicating that the crack density is lower than the reference value, and the damage parameter is 1.08, indicating that the damage state exceeds the reference value, but is mainly dominated by porosity.

[0076] For example, the rock has low porosity, but the crack is easy to cause chain damage, For example, it is detected that the porosity index is 0.5, indicating that the current porosity is normal, and the high-risk crack density index is 1.5, indicating that the high-risk crack density exceeds the reference value by 50%, and the damage parameter is 1.2, and the crack contribution ratio is 87.5%, so the crack needs to be handled first.

[0077] Based on detailed structural damage information, slope stability can be assessed more accurately and risk factors that may cause landslides can be discovered in a timely manner, thereby improving the accuracy and response speed of the early warning system. The quantitative indicators provided, such as porosity index and high-risk crack density index, provide managers with intuitive and easy-to-understand decision support tools, helping them make faster and better decisions about slope management.

[0078] refer to Figure 2 As shown, the anti-slide pile dynamic monitoring module detects the pressure difference on both sides of the anti-slide pile and the change in the pile-soil gap in real time during rainfall through patch pressure sensors and displacement sensors arranged on the soil-facing and soil-removing sides of the anti-slide pile.

[0079] It should be noted that by installing pressure sensors on both sides of the anti-slide pile, the pressure difference between the two sides can be directly measured. This pressure difference reflects the impact of changes in pore water pressure caused by factors such as rainfall on slope stability. Displacement sensors can detect the relative displacement or deformation between the anti-slide pile and the surrounding soil.

[0080] Arrays of patch-type pressure sensors are installed at symmetrical positions on the soil-facing and soil-removing sides of the anti-slide piles to ensure one-to-one correspondence with the elevation sensors.

[0081] It should be noted that by installing sensors on both sides at the same height, the pressure difference on both sides of the same horizontal plane can be accurately measured. If the sensors on both sides are not at the same height, due to factors such as soil stratification and groundwater level changes, the pressure values ​​on both sides cannot be directly compared, making it difficult to obtain an accurate pressure difference.

[0082] A synchronous acquisition mechanism for double-sided pressure sensors is established to synchronously obtain the original pressure values ​​on the soil-facing side and the soil-receiving side at fixed time intervals.

[0083] It should be noted that the synchronous acquisition mechanism can ensure that the pressure values ​​on the soil-facing side and the soil-receiving side are obtained at the same time point, avoiding data errors caused by time difference.

[0084] Considering that rainfall intensity changes rapidly, a shorter time interval can be used to capture rapidly changing pressure conditions. For example, in the heavy rainfall season, it can be set to every 5 minutes, while in the dry season, it can be appropriately extended to every 30 minutes.

[0085] Only data obtained at the same time point can accurately reflect the actual pressure difference between the two sides. If the data on both sides are not collected synchronously, errors may occur due to pressure changes caused by the time difference, affecting the accuracy of the pressure difference.

[0086] The difference between the pressure values ​​of the soil-facing side sensor and the soil-facing side sensor paired at the same elevation is calculated in real time.

[0087] The pressure difference values at each elevation position are arithmetically averaged to output the overall net lateral pressure difference of the pile body.

[0088] The pressure difference values at different elevation positions reflect the changing trend of stress distribution within the slope. By analyzing these data, a more comprehensive understanding of the stability of the entire anti-slide pile and the surrounding soil can be obtained, providing a basis for subsequent risk assessment and maintenance plans.

[0089] Along the longitudinal side of the anti-slide pile, multiple sets of laser displacement sensor arrays are arranged, each set containing symmetric measurement points on the soil-facing side and the soil-backing side.

[0090] In the absence of rainfall interference, the reference values of each sensor are calibrated to establish a three-dimensional spatial coordinate reference for the initial position of the pile body.

[0091] It should be noted that establishing a three-dimensional spatial coordinate reference for the initial position of the pile body can accurately record any small displacement changes, ensuring the reliability and accuracy of the data.

[0092] During rainfall, the distance measurements from each symmetric measurement point sensor to the slope surface are obtained in real time.

[0093] The real-time distance measurements are differenced with the corresponding reference values to output the instantaneous gap changes between the pile and the soil at each measurement point.

[0094] It should be noted that by differencing the real-time distance measurements with the initial reference values, the relative displacement between the anti-slide pile and the surrounding soil at each measurement point can be accurately obtained. This "gap change" directly reflects whether the pile and soil have separated, slid, or deformed locally.

[0095] The gap changes of all measurement points on the soil-facing side and the soil-backing side are arithmetically averaged to generate the overall gap change of the pile body.

[0096] It should be noted that although the overall gap change is an average value, in combination with the instantaneous changes of each measurement point, it can identify which heights or positions have experienced significant displacement, which is helpful for subsequent key investigation and maintenance.

[0097] By real-time monitoring of the pressure difference on both sides of the anti-slide pile and the gap change between the pile and the soil, factors that may lead to slope instability can be detected early, providing detailed data support for subsequent risk assessment models. For example, if the pressure on one side significantly increases or the gap between the pile and the soil significantly expands, it may indicate an impending landslide risk, thereby issuing an early warning signal.

[0098] The landslide risk analysis module constructs a slope landslide risk prediction model based on rainfall, seepage, structural damage status parameters, pressure difference on both sides of the anti-slide piles, and change in the pile-soil gap. The pressure difference on both sides of the anti-slide piles, structural damage status parameters and rainfall are cross-dynamically correlated through time series characteristics, and the seepage and pile-soil gap changes are integrated to output the landslide risk level.

[0099] The real-time collected rainfall, seepage, structural damage parameters, pressure difference on both sides of the anti-slide pile and pile-soil gap change are normalized.

[0100] It should be noted that the normalization of rainfall, seepage, structural damage parameters, pressure difference, and pile-soil gap change is to enable these data from different dimensions and sizes to be effectively compared, integrated, analyzed, and intelligently modeled on the same scale, thereby more accurately assessing the safety status of the slope and achieving scientific early warning.

[0101] The normalized rainfall, structural damage state parameters and pressure difference on both sides of the anti-slide piles are time-series aligned using a unified time reference, and the rainfall, structural damage state parameters and pressure difference on both sides of the anti-slide piles at the same moment are bound into a triplet according to the timestamp.

[0102] It should be noted that the purpose of using a unified time base for timing alignment is to ensure that data obtained from different sources can be accurately linked at the same time point, thereby improving the accuracy and reliability of data analysis.

[0103] Arrange the triplets in consecutive time periods in chronological order to form a time series correlation data set and construct a dynamic risk amplification factor ,in, represents the rainfall influence coefficient, represents the cumulative effect coefficient of pressure difference, Indicates rainfall, represents the structural damage state parameter, Indicates the pressure difference on both sides of the anti-slide pile, Represents a dynamic time point, represents the initial observation time.

[0104] It should be noted that Parameters representing rainfall intensity and structural damage state The product of represents the instantaneous destructive effect of rainstorm impact on the structure. express arrive The cumulative rainfall over a period of time and the pressure difference on both sides of the anti-slide pile The product of reflects the soil softening and lateral pressure accumulation caused by long-term seepage, and is composed of the pressure difference cumulative effect coefficient through the superposition of the pressure difference cumulative effect coefficient and the pressure difference cumulative effect coefficient.

[0105] When a sudden downpour occurs, the intensity of the rainfall drops sharply The short-term peak of the leading dynamic risk amplification factor, the cumulative rainfall when the rainfall continues The dominant dynamic risk amplification factor of the item is elevated for a long time. The higher the value of the dynamic risk amplification factor, the faster the structural damage will be. The time series of the structural damage state parameters in the next time period will increase, which will lead to further amplification of the dynamic risk amplification factor.

[0106] Combining the seepage volume and the change in the pile-soil gap, the soil saturation critical index is calculated. ,in, represents the seepage saturation weight, represents the weight of pile-soil gap change, Indicates the amount of water seepage, Indicates rainfall, Indicates the change in pile-soil gap.

[0107] The ratio of seepage volume to rainfall is the drainage efficiency term, which measures the soil's drainage capacity. A smaller value indicates a higher risk of waterlogging. The change in the pile-soil gap directly reflects the deterioration of the anti-slip structure caused by soil deformation. The combined seepage stability and the risk of structure-soil debonding quantitatively characterize the critical state of the soil approaching saturation failure.

[0108] When there is heavy rainfall, the amount of rainfall will increase, which will lead to a decrease in drainage efficiency. The critical saturation index of the dominant soil increases, and the change in the pile-soil gap increases, indicating that the contact surface is deteriorating. The critical saturation index of the dominant soil increases sharply. When the drain is stable, When represents the critical seepage region, It indicates soil saturation failure.

[0109] Integrate the above parameters to output landslide risk value ,in, represents the structural damage weight, represents the critical weight of soil saturation, Indicates the instantaneous pressure difference weight.

[0110] The weighted sum of structural damage and soil saturation critical index describes the internal causes of landslides, which refer to structural deterioration and soil instability. The pressure difference on both sides of the anti-slide pile acts independently, representing the immediate driving force of the external water pressure. The dynamic risk amplification factor is used to strengthen the internal cause item, reflecting the accelerating effect of rainfall on the internal cause of landslides. The pressure difference on both sides of the anti-slide pile is used as an independent superposition item to ensure that the sudden hydraulic driving force is not masked.

[0111] refer to Figure 3As shown in the figure, when the dynamic risk amplification factor increases, the structural damage intensifies. At this time, the pile-soil constraint weakens, resulting in an increase in the pile-soil gap change, which in turn leads to an increase in the soil saturation critical index. The increase in the soil saturation critical index will cause soil softening. At this time, the pressure difference fluctuation on both sides of the anti-slide pile increases, resulting in an amplified cumulative rainfall effect, which in turn leads to a further increase in the dynamic risk amplification factor.

[0112] Landslide risk levels are divided according to landslide risk values.

[0113] The monitoring time series data and corresponding landslide status labels in the historical landslide event database are used as the basis for model training.

[0114] It should be noted that using real historical landslide event data for training can enable the deep learning model to learn the actual patterns and characteristics that lead to landslides. These data contain the actual response conditions under different conditions, such as rainfall, seepage, structural damage status, etc., which helps to build a more accurate prediction model.

[0115] The time window length for real-time data collection is preset to several groups. Every time several groups of monitoring data are added, these data are added to the training set, and the rainfall influence coefficient, pressure difference cumulative effect coefficient, seepage saturation weight, pile-soil gap change weight, structural damage weight, soil saturation critical weight and instantaneous pressure difference weight are fine-tuned.

[0116] It should be noted that the length of the time window can be used to analyze the effective time span of historical landslide precursor data. For example, it can be determined that 80% of landslide precursors appear within 12 hours after the start of rainfall. By scientifically setting the time window, an optimal balance can be achieved between real-time performance and prediction accuracy.

[0117] Continuously monitor the deviation between the model prediction results and the actual landslide risk status. When the prediction error continuously exceeds the preset threshold, immediately start the model retraining process and quickly update the coefficients and weights.

[0118] It should be noted that by continuously monitoring the prediction error and quickly retraining the model when necessary, it can be ensured that the landslide early warning system always maintains high accuracy, high adaptability and high reliability, which is a key link in building an intelligent slope safety monitoring system.

[0119] The landslide risk analysis module integrates various monitoring data and utilizes intelligent AI technology to build a landslide risk prediction model. This not only significantly improves the accuracy and reliability of the landslide warning system but also provides strong technical support for slope management. Building an intelligent warning system automates the entire process from data collection and analysis to warning issuance, reducing manual intervention and improving work efficiency.

[0120] The risk warning output module generates and publishes landslide warning information of corresponding levels according to the landslide risk level matching the preset risk threshold.

[0121] According to the landslide risk level, the preset landslide risk level threshold range is matched to determine the corresponding warning level.

[0122] Specifically, low risk: Medium risk: ; High risk: ; Very high risk: .

[0123] Based on the warning level, the preset warning information template is called to dynamically generate landslide warning information including the warning level, risk section location information, real-time landslide risk value and recommended disposal measures for the corresponding risk level.

[0124] It should be noted that by dynamically generating detailed landslide warning information based on the warning level, not only can the practicality and operability of the warning system be significantly improved, but it can also provide strong support for management departments at all levels, ensuring that they can respond quickly and effectively when faced with landslide threats.

[0125] The parameters involved in the above formula are all dimensionless and calculated numerically. The formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0126] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0127] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0128] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0129] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0130] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A highway landslide early warning system based on rainfall data analysis using intelligent AI, characterized by: include: Rain gauges are placed on the slope surface to collect real-time rainfall on the target slope over a continuous period of time, and piezometers are placed at the slope foot to obtain water seepage simultaneously. Scan the target slope surface with a thermal infrared high-definition imager to obtain images, and analyze the current structural damage parameters of the slope based on the porosity and crack distribution characteristics in the images; By deploying patch-type pressure sensors and displacement sensors on the soil-facing and soil-removing sides of the anti-slide piles, the pressure difference on both sides of the anti-slide piles and the change in the pile-soil gap during rainfall are detected in real time. A slope landslide risk prediction model is constructed based on rainfall, water seepage, structural damage state parameters, pressure difference on both sides of the anti-slide pile, and pile-soil gap change. The pressure difference on both sides of the anti-slide pile, structural damage state parameters, and rainfall are cross-dynamically correlated through time series characteristics, and the water seepage and pile-soil gap change are integrated to output the landslide risk level. According to the landslide risk level and the preset risk threshold, landslide warning information of the corresponding level is generated and released.

2. The intelligent AI-based rainfall data analysis highway landslide early warning system according to claim 1 is characterized by: The specific contents of deploying rain gauges on the slope surface to collect the rainfall of the target slope in a continuous period in real time are as follows: The target slope surface is divided into equal intervals along the slope length and slope width to form regularly distributed rectangular grids. Rain gauges are placed at the intersection of each rectangular grid; The rainfall amount of each rain gauge is continuously collected in real time every minute, and the rainfall intensity is accumulated and calculated every hour.

3. The intelligent AI-based rainfall data analysis highway landslide early warning system according to claim 1 is characterized by: The specific contents of deploying piezometers at the foot of the slope to synchronously obtain water seepage are as follows: Multiple piezometer monitoring points are pre-buried at equal intervals along the slope foot to continuously obtain the pore water pressure values ​​at each piezometer monitoring point in real time; According to the saturated bulk density of the soil at the location where the piezometer is installed, the ratio of the measured pore water pressure to the saturated bulk density of the soil at the location is used as the equivalent water head height; The incremental change of water head under rainfall is obtained by taking the difference between the equivalent water head height and the water head height in the initial stable state; The cumulative seepage volume in the corresponding period is calculated by multiplying the effective infiltration area of ​​the monitoring area, the incremental change in water head under rainfall, and the saturated bulk density of the soil.

4. The intelligent AI-based rainfall data analysis highway landslide early warning system according to claim 1 is characterized by: The specific steps for obtaining the porosity and crack distribution characteristics in the image are as follows: The high-definition thermal infrared imager is used to scan the target slope surface at multiple angles to obtain a thermal infrared image sequence of the target slope surface temperature distribution characteristics. Preprocess the acquired thermal infrared image, compare the temperature value of each pixel in the preprocessed thermal infrared image with the preset reference temperature field point by point, and mark the pixels whose difference exceeds the preset temperature difference threshold; Perform connected domain analysis on spatially continuous marked pixels to generate closed boundaries and extract them as pore feature areas; The total area of ​​the pore characteristic region is counted, and its proportion to the total area of ​​the scanning area is calculated as the porosity; Screening the areas in the pore feature area whose aspect ratio is greater than or equal to a preset ratio and whose spatial overlap rate in a number of consecutive frames of images is greater than or equal to a preset threshold as continuous temperature anomaly zones; Perform morphological skeleton extraction on the continuous temperature anomaly zone to generate a crack skeleton trajectory with a single pixel width, and calculate the total length along the crack skeleton trajectory; Measure the width of each position on the trajectory and record the maximum value. Mark the cracks whose width exceeds the width threshold as high-risk cracks. The total length of high-risk cracks per unit area is calculated as the high-risk crack density, which is the crack distribution characteristic.

5. The intelligent AI-based rainfall data analysis highway landslide early warning system according to claim 4 is characterized by: The specific contents of the output slope current structural damage state parameters are as follows: The porosity and high-risk fracture density are respectively calculated by comparing the porosity and high-risk fracture density with the preset benchmark porosity and preset benchmark fracture density to obtain the dimensionless porosity index and high-risk fracture density index; The porosity index and high-risk crack density index are weightedly fused to generate the current structural damage state parameters of the slope.

6. The intelligent AI-based rainfall data analysis highway landslide early warning system according to claim 1, characterized in that: The specific steps for obtaining the pressure difference between the two sides of the anti-slide pile are as follows: Install patch pressure sensor arrays at symmetrical positions on the soil-facing and soil-removing sides of the anti-slide piles to ensure a one-to-one correspondence with the elevation sensors; Establish a synchronous acquisition mechanism for double-sided pressure sensors to synchronously obtain the original pressure values ​​on the soil-facing and soil-receiving sides at fixed time intervals; The difference between the pressure values ​​of the sensors on the soil facing side and the pressure values ​​of the sensors on the soil facing side that are paired at the same elevation is calculated in real time; The pressure differences at each elevation are arithmetically averaged to output the overall net lateral pressure difference of the pile.

7. The intelligent AI-based rainfall data analysis highway landslide early warning system according to claim 1 is characterized by: The specific steps for obtaining the pile-soil gap variation are as follows: Multiple laser displacement sensor arrays are arranged longitudinally along the side of the anti-slide pile, with each array containing symmetrical measuring points on the soil-facing side and the soil-removing side. Calibrate the reference values ​​of each sensor in a state without rainfall interference and establish the three-dimensional spatial coordinate reference of the pile's initial position; During rainfall, the distance measurement value between each measuring point sensor and the slope surface is obtained in real time; Perform differential calculation between the real-time distance measurement value and the corresponding reference value, and output the instantaneous pile-soil gap change at each measuring point; The arithmetic average of the gap changes of all measuring points on the soil side and the soil side is taken to generate the overall gap change of the pile.

8. The intelligent AI-based rainfall data analysis highway landslide early warning system according to claim 1 is characterized by: The specific contents of the landslide risk analysis module are as follows: Normalize the real-time collected rainfall, seepage, structural damage parameters, pressure difference between the two sides of the anti-slide pile, and pile-soil gap changes; The normalized rainfall, structural damage state parameters, and pressure difference on both sides of the anti-slide pile are time-series aligned using a unified time base. The rainfall, structural damage state parameters, and pressure difference on both sides of the anti-slide pile at the same moment are bound into a triplet according to the timestamp. Arrange the triplets in consecutive time periods in chronological order to form a time series correlation data set and construct a dynamic risk amplification factor ,in, represents the rainfall influence coefficient, represents the pressure difference cumulative effect coefficient, Indicates rainfall, represents the structural damage state parameter, Indicates the pressure difference on both sides of the anti-slide pile, Represents a dynamic time point, represents the initial observation time; Calculation of soil saturation critical index by combining seepage volume and pile-soil gap change ,in, represents the seepage saturation weight, represents the weight of pile-soil gap change, Indicates the amount of water seepage, Indicates rainfall, Indicates the change in pile-soil gap; Integrate the above parameters to output landslide risk value ,in, represents the structural damage weight, represents the critical weight of soil saturation, Indicates the instantaneous pressure difference weight; Landslide risk levels are divided according to landslide risk values.

9. The intelligent AI-based rainfall data analysis highway landslide early warning system according to claim 8, characterized in that: The specific steps for obtaining the rainfall influence coefficient, pressure difference cumulative effect coefficient, seepage saturation weight, pile-soil gap change weight, structural damage weight, soil saturation critical weight and instantaneous pressure difference weight are as follows: The monitoring time series data and corresponding landslide status labels in the historical landslide event database are used as the basis for model training; The time window length for real-time data collection is preset to several groups. Every time a number of new monitoring data groups are added, these data are added to the training set, and the rainfall influence coefficient, pressure difference cumulative effect coefficient, seepage saturation weight, pile-soil gap change weight, structural damage weight, soil saturation critical weight and instantaneous pressure difference weight are fine-tuned; Continuously monitor the deviation between the model prediction results and the actual landslide risk status. When the prediction error continuously exceeds the preset threshold, immediately start the model retraining process and quickly update the coefficients and weights.

10. The intelligent AI-based rainfall data analysis highway landslide early warning system according to claim 1, characterized in that: The specific contents of the risk warning output module are as follows: According to the landslide risk level, the preset landslide risk level threshold range is matched to determine the corresponding warning level; Based on the warning level, the preset warning information template is called to dynamically generate landslide warning information including the warning level, risk section location information, real-time landslide risk value and recommended disposal measures for the corresponding risk level.

Images