An integrated coastal slope monitoring method based on multi-parameter collaborative recognition

Abstract

To significantly improve the prediction accuracy, response speed and management efficiency of large river bank slope disasters, an integrated bank slope monitoring method based on multi-parameter collaborative recognition is proposed. The solution includes step S1 of synchronously collecting data on bank slope displacement, pore water pressure, inclination angle, vibration frequency and environmental temperature and humidity to form an original monitoring dataset and construct a multi-parameter collaborative recognition network; step S2 of using a multi-modal data fusion algorithm to generate a fusion data matrix including spatiotemporal correlation features and perform spatiotemporal data alignment and outlier cleansing; step S3 of combining a geomechanical parameter library and a past disaster case library to output a risk level map and perform dynamic risk assessment model analysis; and step S4 of triggering a multi-level early warning mechanism and generating linked control commands including treatment suggestions to perform multi-level early warning and linked control.

Description

[Technical Field]

[0001] The present invention relates to the technical field of river bank slope monitoring, and particularly to an integrated river bank slope monitoring method based on multi-parameter collaborative recognition. [Background technology]

[0002] River bank slope collapse disasters, especially those of large rivers and streams, have had a serious impact on the lives and infrastructure of people living along the rivers. Currently, due to the combined influences of climate change and human activities, river bank slope stability issues have become even more complex. Traditional bank slope monitoring methods often rely on single sensors to collect localized data, which often makes it difficult to comprehensively and accurately monitor changes in river bank slopes. This is particularly true under the circumstances of variable climates, complex water flow conditions, and unique topography of different river bank slopes, which may result in deviations in risk assessment results. With the advancement of monitoring technology in the water conservancy field, how to achieve real-time monitoring and accurate assessment of river bank slopes has become an important issue in current research and water conservancy engineering.

[0003] Existing river bank slope monitoring technologies can provide stability assessments to some extent, but they suffer from several significant shortcomings. First, the commonly used monitoring methods rely on independent sensors or equipment, lacking collaborative recognition and multi-source data integration, limiting the accuracy and comprehensiveness of the monitoring data. Second, traditional risk assessment methods typically rely on static data or simple computational models, failing to fully utilize geomechanical and historical disaster data, preventing dynamic adjustment of risk assessment results and further impacting the effectiveness of disaster warnings and emergency responses. Finally, many current monitoring systems lack early warning and emergency control mechanisms for river bank slope collapses, making it difficult to respond in a timely manner in high-risk situations such as unstable collapses. Summary of the Invention [Problem to be solved by the invention]

[0004] This invention provides an integrated bank slope monitoring method based on multi-parameter collaborative recognition, which significantly improves the accuracy, real-timeness and response capability of river bank slope stability monitoring, and can effectively predict and perform emergency linked control processing against potential collapse disasters. [Means for solving the problem]

[0005] An integrated bank slope monitoring method based on multi-parameter collaborative recognition, comprising: Step S1 of constructing a multi-parameter collaborative recognition network, and synchronously collecting data on bank slope displacement, pore water pressure, inclination angle, vibration frequency, and environmental temperature and humidity through distributed intelligent sensor nodes to form an original monitoring data set; a step S2 of spatiotemporal data alignment and outlier cleansing, in which the original monitoring dataset is subjected to spatiotemporal alignment and outlier cleansing, and a multimodal data fusion algorithm is used to generate a fused data matrix containing spatiotemporal correlation features; Step S3 of performing dynamic risk assessment model analysis, which involves inputting the fusion data matrix into a dynamic risk assessment model, combining the geomechanical parameter library and the past disaster case library, and outputting a risk level map; and step S4 of performing multi-level early warning and linked control, which involves triggering a multi-level early warning mechanism based on the gradient change characteristics of the risk level map and generating linked control instructions including treatment suggestions.

[0006] Optionally, constructing the multi-parameter collaborative recognition network according to S1 includes: Step S11 of distributing intelligent sensor nodes, which includes a displacement sensor, a pore water pressure sensor, a tilt angle sensor, a vibration frequency sensor, and an environmental temperature and humidity sensor, and distributing the sensors in the bank slope area; Step S12 of synchronously collecting sensor data, in which each sensor node synchronously collects data on the bank slope displacement, pore water pressure, inclination angle, vibration frequency, and environmental temperature and humidity via a wireless communication network; and a data integration step S13 of receiving sensor data from each intelligent sensor node and generating the original monitoring data set D.

[0007] Optionally, performing the spatiotemporal data alignment and outlier cleansing in step S2 may include: timestamp t aligned a spatiotemporal alignment step S21, in which spatiotemporal alignment is performed on each sensor data in the original monitoring data set through Each sensor data is adjusted using a linear interpolation algorithm, and the threshold δ is set when the sensor data deviates from the reference range. i If the difference exceeds the threshold, the sensor data is regarded as an abnormal value and marked or removed in step S22, which is an abnormal value cleansing step. A multimodal data fusion algorithm is used to fuse different sensor data and generate a fused data matrix D containing spatiotemporal correlation features. fusion and a step S23 of performing multimodal data fusion to generate a

[0008] Optionally, performing the dynamic risk assessment model analysis described in step S3 includes: A step S31 of performing spatiotemporal feature analysis, in which the generated fusion data matrix is ​​input into a dynamic risk assessment model to perform spatiotemporal feature analysis and generate a risk score; Step S32 of performing geological and historical data analysis, combining the geomechanical parameter library and the historical disaster case library, generating risk adjustment factors, modifying the risk score, and generating a final area risk score; and outputting the risk level map S33, which generates a risk level map based on the final area risk scores.

[0009] Optionally, the dynamic risk assessment model in step S31 employs a graph convolutional network model (GCN), and constructing the graph convolutional network model (GCN) includes: Step S311 of constructing a graph structure, in which each sensor data represents one node, the spatial relationship between every two sensors is represented by an edge, and the relationship between the nodes is represented by an adjacency matrix A; Generated fusion data matrix D fusion Data X at time t of each sensor in i Step S312 of initializing node features by generating an initial feature matrix H0 using (t) as an initial input; Step S313 performs a graph convolution operation, which involves transmitting information through an adjacency matrix A, obtaining a weighted average between the feature vector of each node and the features of its adjacent nodes, and transmitting the information to its adjacent nodes through a convolutional layer; By convolving multiple graph convolution layers, the node features are convolved with the adjacent matrix and the features of the previous layer in each layer to produce the final node feature H (L) Step S314 of performing a multi-layer graph folding operation to gradually extract and step 315 of outputting a list evaluation by obtaining a feature representation of each node through multiple graph convolution calculations and outputting a risk score R of the node using a softmax function.

[0010] Optionally, performing the geological and historical data analysis described in step S32 may include: a step S321 of generating risk correction coefficients, which involves obtaining geological characteristics (soil type, rock strength, groundwater flow) and past disaster cases (past landslides, earthquakes, etc.) related to the bank slope by examining a geomechanical parameter library, and generating a corresponding risk correction coefficient Q for each area; and a step S322 of correcting the list score by correcting the risk score of each area based on the generated risk correction coefficient Q to obtain a final area risk score R'.

[0011] Optionally, outputting the risk level map in step S33 may include: Based on the final area risk score R' of each area, the risk level of each area is determined by comparing it with the preset risk level threshold, which is shown by the following formula:

[0012]

number

[0013] where Risk Level(i) is the risk level of the i-th area, and T low is the low-risk threshold, and T high a step S331 of generating a list level, where is a high risk threshold; and step S332 of generating a risk level map, which finally generates a risk level map by assigning a corresponding color to the risk level of each area (for example, green for low risk, yellow for medium risk, and red for high risk).

[0014] Optionally, performing the multi-level early warning and linked control described in step S4 is A step S41 of performing gradient change characteristic analysis, in which a gradient change characteristic analysis is performed on the generated risk level map, and a gradient change value is calculated by calculating the change gradient between the risk level of each zone and the risk level of an adjacent zone; A step S42 of triggering a multi-level early warning mechanism, which is to compare the calculated gradient change value with a corresponding early warning threshold value and trigger a corresponding multi-level early warning mechanism, including a low-risk early warning, a medium-risk early warning, and a high-risk early warning; and generating an interlocking control command S43, which generates a corresponding interlocking control command based on the corresponding early warning mechanism and attaches a treatment suggestion.

[0015] Optionally, triggering the multi-level early warning mechanism in step S42 comprises:

[0016]

number

[0017] If T, it indicates that the early warning level of the area is low risk early warning, where T low a step S421 of issuing a low-risk early warning, where ' is a low-risk early warning threshold;

[0018]

number

[0019] If T, it indicates that the early warning level of the area is medium risk early warning, where T high step S422 of issuing a medium risk early warning, where ' is a high risk early warning threshold;

[0020]

number

[0021] If so, issuing a high-risk early warning, that is, the early warning level of the area is a high-risk early warning.

[0022] Optionally, generating the interlocking control command described in S43 includes: Step S431 of issuing a low-risk warning command to the low-risk area, maintaining normal monitoring and updating risk assessment data periodically while maintaining normal operation of the early warning system; Step S432 of issuing a medium-risk early warning command for the medium-risk area, strengthening real-time monitoring, issuing warning information, strengthening defenses in the disaster-hit area, and adjusting the operation and maintenance strategies of related facilities; Step S433 includes issuing a high-risk early warning command to activate an emergency response program for high-risk areas, including scheduling emergency rescue teams, issuing area closure or personnel evacuation orders, and increasing monitoring frequency. [Effects of the Invention]

[0023] Compared with the prior art, the present invention has the following beneficial effects.

[0024] This invention builds a multi-parameter cooperative sensing network and combines multiple intelligent sensors to synchronously collect multidimensional data such as bank slope displacement, pore water pressure, inclination angle, vibration frequency, and environmental temperature and humidity data, thereby overcoming the shortcomings of traditional monitoring methods that rely on a single data source and enabling real-time, comprehensive monitoring of bank slope conditions. The spatiotemporal alignment and outlier cleansing technologies ensure the accuracy and reliability of the data, significantly improving the precision and real-time nature of monitoring.

[0025] This invention adopts a dynamic risk assessment model based on a graph convolutional network model, and combines a geomechanical parameter library and a past disaster case library to accurately assess the risk levels of different areas. It then triggers a multi-level early warning mechanism based on the gradient change characteristics of the risk level map, ensuring that high-risk areas receive high-priority early warnings immediately and enabling decision-makers to take scientific response measures based on different risk levels. This intelligent assessment and early warning capability effectively improves the accuracy and response speed of coastal slope disaster prediction.

[0026] The present invention makes emergency response more efficient by generating linked control commands and treatment suggestions. Based on the risk level map, the system can generate corresponding control commands based on different early warning mechanisms, for example, low-risk areas should maintain normal monitoring, medium-risk areas should strengthen monitoring and defense, and high-risk areas should launch emergency response programs including scheduling emergency resources and issuing orders to seal off areas or evacuate personnel. Such a linked control mechanism based on real-time risk assessment can ensure that effective response measures can be taken quickly and effectively when a disaster occurs, minimizing losses. [Brief explanation of the drawings]

[0027] In order to more clearly describe the technical aspects of the present invention or the prior art, the following briefly describes the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only for the present invention, and those skilled in the art can also obtain other drawings based on these drawings without paying creative labor.

[0028] [Figure 1] 1 is a flowchart of a monitoring method of the present invention. [Figure 2] 1 is a schematic diagram of a dynamic risk assessment model analysis according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0029] The present invention will be described in detail below with reference to the drawings and specific embodiments. At the same time, it is to be noted that in order to provide more detailed examples, the following examples are the most suitable and preferred examples, and those skilled in the art may also adopt other alternative methods to carry out the present invention. In addition, the drawings are only for the purpose of more specifically describing the embodiments, and are not intended to specifically limit the present invention.

[0030] It should be noted that references in the specification to "one embodiment," "an embodiment," "an exemplary embodiment," "some embodiments," etc., indicate that the described embodiments may include a particular feature, structure, or characteristic, but do not necessarily indicate that each embodiment includes that particular feature, structure, or characteristic. Furthermore, when a particular feature, structure, or characteristic is described in connection with an embodiment, such feature, structure, or characteristic should be realized within the knowledge of one skilled in the art in connection with other embodiments (whether or not explicitly described).

[0031] Generally, terms can be understood, at least in part, from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Furthermore, the term "based" is not necessarily intended to convey an exclusive set of elements, but instead, depending at least in part on the context, can allow for the presence of other elements not necessarily explicitly described.

[0032] As shown in Fig. 1-2, an integrated bank slope monitoring method based on multi-parameter collaborative recognition, Step S1 of constructing a multi-parameter collaborative recognition network, and synchronously collecting data on bank slope displacement, pore water pressure, inclination angle, vibration frequency, and environmental temperature and humidity through distributed intelligent sensor nodes to form an original monitoring data set; a step S2 of spatiotemporal data alignment and outlier cleansing, in which the original monitoring dataset is subjected to spatiotemporal alignment and outlier cleansing, and a multimodal data fusion algorithm is used to generate a fused data matrix containing spatiotemporal correlation features; Step S3 of performing dynamic risk assessment model analysis, which involves inputting the fusion data matrix into a dynamic risk assessment model, combining the geomechanical parameter library and the past disaster case library, and outputting a risk level map; and step S4 of performing multi-level early warning and linked control, which involves triggering a multi-level early warning mechanism based on the gradient change characteristics of the risk level map and generating linked control instructions including treatment suggestions.

[0033] Constructing a multi-parameter collaborative recognition network as described in S1 includes the following steps:

[0034] Step S11: Deploying intelligent sensor nodes: Select various types of intelligent sensors, including displacement sensors, pore water pressure sensors, tilt angle sensors, vibration frequency sensors, and environmental temperature and humidity sensors, to ensure that important monitoring points are covered and distributed across the bank slope area. Here, the displacement sensors, pore water pressure sensors, tilt angle sensors, vibration frequency sensors, and environmental temperature and humidity sensors are integrated into a monitoring rod, which is attached to the bank slope of the riverbank and buried in a pre-drilled hole.

[0035] Step S12: Synchronously collect sensor data: Each sensor node synchronously collects the bank slope displacement, pore water pressure, inclination angle, vibration frequency and environmental temperature and humidity data via the wireless communication network. Specifically, Measuring the bank slope displacement △X(t) using a displacement sensor; Measuring pore water pressure P(t) using a pore water pressure sensor; measuring the tilt angle θ(t) with a tilt angle sensor; measuring a vibration frequency f(t) with a vibration frequency sensor; This includes measuring the temperature T(t) and humidity H(t) measured by the environmental sensors.

[0036] Step S13: Perform data integration: receive the sensor data from each intelligent sensor node and generate the original monitoring data set D, which is represented by the following equation:

[0037]

number

[0038] Performing data spatio-temporal alignment and outlier cleansing in step S2 includes the following steps:

[0039] Step S21: Perform space-time alignment: unified timestamp t aligned The space-time alignment is performed for each sensor data in the original monitoring dataset through

[0040]

number

[0041] where X i (t aligned ) is the unified timestamp t of the i-th sensor aligned This is the data in.

[0042] Step S22, outlier cleansing: adjust each sensor data by a linear interpolation algorithm, and when the sensor data deviates from the reference range, the set threshold δ i If the value exceeds the threshold, the sensor data is regarded as an abnormal value and marked or removed.

[0043]

number

[0044] where X i (t) is the data of the i-th sensor at time t, and μ i is the mean value of the i-th sensor, NaN is an outlier mark, Threshold δi is set based on past data and is given by the following formula:

[0045]

number

[0046] where σ i is the standard deviation of the i-th sensor, and k is an empirical coefficient whose value ranges from 2 to 3.

[0047] Step S23, perform multimodal data fusion: use a multimodal data fusion algorithm to fuse different sensor data to generate a fused data matrix D containing spatiotemporal correlation features. fusion , which is shown by the following equation:

[0048]

number

[0049] where D fusion (t aligned ) is the fused data matrix after spatiotemporal alignment and outlier cleansing, and w i is the weight of the i-th sensor data, n is the number of sensors, and X i (t aligned ) is the i-th sensor data after alignment and cleansing.

[0050] Performing the dynamic risk assessment model analysis described in step S3 includes the following steps:

[0051] Step S31, perform spatiotemporal feature analysis: input the generated fusion data matrix into the dynamic risk assessment model to perform spatiotemporal feature analysis and generate a risk score.

[0052] Step S32, geological and historical data analysis: combine the geomechanical parameter library and the past disaster case library to generate risk correction coefficients, modify the risk score, and generate the final area risk score.

[0053] Step S33, output risk level map: generate a risk level map based on the final area risk score.

[0054] The dynamic risk assessment model described in step S31 adopts a graph convolutional network model (GCN), and constructing the graph convolutional network model (GCN) includes the following steps:

[0055] Step S311: construct a graph structure: in the graph structure, each sensor data represents one node, the spatial relationship between every two sensors is represented by an edge, and the relationship between nodes is represented by an adjacency matrix A, which is shown in the following formula:

[0056]

number

[0057] where A ij is the connection relationship between node i and node j.

[0058] Step S312: Initialize the node features: the generated fusion data matrix D fusion Data X at time t of each sensor in i (t) is the initial input and the initial feature matrix H0 , which is shown by the following equation:

[0059]

number

[0060] where H0 is the initial feature matrix input to the graph convolutional network model.

[0061] Step S313, perform graph convolution operation: propagate information through the adjacency matrix A, obtain the weighted average between the feature vector of each node and the features of its adjacent nodes, and propagate the information to its adjacent nodes through the convolution layer. For each convolution layer, the node update is expressed by the following formula:

[0062]

number

[0063] where H (l+1) is the feature vector of the node in the l+1th layer, σ is the ReLU activation function, (the following equation) is the normalized matrix of the adjacent matrix A, and H (l) is the feature vector of the node in the lth layer, and W (l) is the weight matrix of the lth layer.

[0064]

number

[0065] Step S314: Perform multi-layer graph convolution: By convolving multiple graph convolution layers, the node features are convolved with the adjacent matrix and the features of the previous layer in each layer to obtain the final node features H (L) is extracted stepwise and is shown by the following formula:

[0066]

number

[0067] where H (L) is the final node feature, and H (L-1) is the node feature matrix of the L-1th layer, and W (L-1) is the weight matrix of the L-1th layer.

[0068] Step 315, outputting a list evaluation: through multiple graph convolution calculations, obtain the feature representation of each node, and use the softmax function to output the risk score R of the node, which is given by the following formula:

[0069]

number

[0070] Here, Softmax(H i (L) ) k is the probability value of the i-th node at risk level k, n is the total number of nodes, and W k is the weight for risk level k (low risk has a weight of 0.2, medium risk has a weight of 0.5, and high risk has a weight of 0.8).

[0071] Performing geological and historical data analysis as described in step S32 includes the following steps.

[0072] Step S321, generate risk correction coefficient: by examining the geomechanical parameter library, obtain the geological characteristics (soil type, rock strength, groundwater flow) and past disaster cases (past landslides, earthquakes, etc.) of the bank slope, and generate the corresponding risk correction coefficient Q for each area, which is shown by the following formula:

[0073]

number

[0074] where Q(i) is the risk correction factor for the i-th zone, and W j is the weight of the jth past disaster case, Disaster Similarity(i,j) is the similarity between the ith and jth past disaster cases, and α k is the weight of the kth geological feature, Geological Factor(i,k) is the kth geological feature in the ith area, m is the total number of past disaster cases, and p is the total number of geological features.

[0075] Disaster Similarity(i,j) is calculated based on factors such as geographic location, disaster type, and disaster scale, and is expressed by the following formula:

[0076]

number

[0077] where α1 and α2 are weighting factors related to the similarity of geographical location and the similarity of disaster type, Location Similarity(i,j) is the similarity based on geographical location, and Disaster Type Similarity(i,j) is the similarity based on disaster type, which is expressed by the following formula:

[0078]

number

[0079] Here, Distance(i,j) is the geographic distance between the i-th area and the j-th past disaster case, and is expressed by the following formula:

[0080]

number

[0081] Here, Disaster Type(i) and Disaster Type(j) are the disaster types of the i-th area and the j-th past disaster case, respectively. Geological Factor (i, k) evaluates the contribution of geological factors to risk through weighting analysis of each geological feature, and is expressed by the following formula:

[0082]

number

[0083] Here, β kis the weighting coefficient associated with geological feature k, and Geological Attribute(i,k) is the attribute value of the kth geological feature in the ith area, which is given by the following formula:

[0084]

number

[0085] where Measured Value(i,k) is the measured value of the kth geological feature in the ith area, and μ k is the average value of the geological features in the entire area, and σ k is the standard deviation of the geological feature across the entire area, and standardization allows different types of geological features (e.g., rock strength, soil type, etc.) to be converted into comparative measurements.

[0086] Step S322, correct the list score: based on the generated risk correction coefficient Q, correct the risk score of each area to obtain the final area risk score R', which is shown by the following formula:

[0087]

number

[0088] Outputting the risk level map in step S33 includes the following steps.

[0089] Step S331, generate a list level: based on the final area risk score R' of each area, compare it with the preset risk level threshold to determine the risk level of each area, which is shown by the following formula:

[0090]

number

[0091] where Risk Level(i) is the risk level of the i-th area, and T lowis the low-risk threshold, and T high is the high-risk threshold, T low、 T high is the average risk score for all areas μ R’ and standard deviation σ R’ and is calculated based on the following formula:

[0092]

number

[0093]

number

[0094] Step S332: Generate a risk level map: By assigning a color to the risk level of each area (for example, low risk is green, medium risk is yellow, and high risk is red), a risk level map Risk Map is finally generated, which is shown by the following formula:

[0095]

number

[0096] where RiskLevel(1), RiskLevel(2), ..., RiskLevel(s) are the risk levels of the 1st, 2nd, ..., sth zones, respectively, and s is the total number of zones.

[0097] The multi-level early warning and interlocking control described in step S4 includes the following steps:

[0098] Step S41: Perform gradient change feature analysis: perform gradient change feature analysis on the generated risk level map, and calculate the gradient of change between the risk level of each area and the risk level of the adjacent area, thereby calculating the gradient change value, which is shown by the following formula:

[0099]

number

[0100] where Gradient(i) is the gradient change value of the i-th area, N(i) is the set of areas adjacent to the i-th area, Risk Level(i) is the risk level of the i-th area, and Risk Level(j) is the risk level of the j-th area.

[0101] Step S42, trigger a multi-level early warning mechanism: based on the calculated gradient change value, compare it with the corresponding early warning threshold value, and trigger a corresponding multi-level early warning mechanism, including low-risk early warning, medium-risk early warning, and high-risk early warning.

[0102] Step S43, generate interlocking control instructions: generate corresponding interlocking control instructions according to the corresponding early warning mechanism, and attach treatment suggestions.

[0103] Triggering the multi-level early warning mechanism described in step S42 includes the following steps:

[0104] Step S421, issue low-risk early warning:

[0105]

number

[0106] If T, it indicates that the early warning level of the area is low risk early warning, where T low ' is the low-risk early warning threshold.

[0107] Step S422, medium risk early warning:

[0108]

number

[0109] If T, it indicates that the early warning level of the area is medium risk early warning, where T high ' is the high risk early warning threshold.

[0110] Step S423, issue high risk early warning:

[0111]

number

[0112] If so, the early warning level of the area is a high risk early warning.

[0113] Low-risk early warning threshold T low ', high risk early warning threshold T high ' is set based on the quantile, Low-risk early warning threshold: Select the 25th percentile (Q1) of slope change values ​​to represent areas with low slope change; High risk early warning threshold: Select the 75th percentile (Q3) of slope change values ​​to represent areas with high slope change.

[0114] The generation of the interlocking control command described in S43 includes the following steps.

[0115] Step S431, issuing a low-risk warning command: for low-risk areas, maintaining normal monitoring and maintaining the normal operation of the early warning system, while periodically updating risk assessment data.

[0116] Step S432, issue a medium-risk early warning command: for medium-risk areas, strengthen real-time monitoring, issue warning information, strengthen defenses in disaster-hit areas, and adjust operation and maintenance strategies for relevant facilities.

[0117] Step S433, issue a high-risk early warning command: for high-risk areas, activate an emergency response program, including scheduling an emergency rescue team, issuing an order to close off areas or evacuate personnel, and increasing the monitoring frequency.

[0118] The present invention encompasses alternatives, modifications, equivalent methods and schemes within the spirit and scope of the present invention. Generally, in order to provide a thorough understanding of the present invention, specific details are described in the following preferred embodiments of the present invention, and those skilled in the art can fully understand the present invention without these details. In addition, well-known methods, processes, flows, elements, circuits, etc. are not described in detail to avoid unnecessary confusion regarding the essence of the present invention.

[0119] The above is only a preferred embodiment of the present invention, and those skilled in the art can make some improvements and refinements without departing from the principle of the present invention, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. Step S1 of constructing a multi-parameter collaborative recognition network, and synchronously collecting data on bank slope displacement, pore water pressure, inclination angle, vibration frequency and environmental temperature and humidity through distributed intelligent sensor nodes to form an original monitoring data set; a step S2 of spatiotemporal data alignment and outlier cleansing, in which the original monitoring dataset is subjected to spatiotemporal alignment and outlier cleansing, and a multimodal data fusion algorithm is used to generate a fused data matrix containing spatiotemporal correlation features; Step S3: performing a dynamic risk assessment model analysis, inputting the fusion data matrix into a dynamic risk assessment model, combining the geomechanical parameter library and the past disaster case library, and outputting a risk level map; and a multi-level early warning and linked control step S4, which triggers a multi-level early warning mechanism based on the gradient change characteristics of the risk level map and generates linked control instructions including treatment suggestions; The construction of the multi-parameter collaborative recognition network according to S1 includes: Step S11 of distributing intelligent sensor nodes, which includes a displacement sensor, a pore water pressure sensor, a tilt angle sensor, a vibration frequency sensor, and an environmental temperature and humidity sensor, and distributing and distributing the intelligent sensor nodes in the bank slope area; a step S12 of synchronously collecting sensor data, in which each sensor node synchronously collects data on the displacement of the bank slope, pore water pressure, inclination angle, vibration frequency, and environmental temperature and humidity via a wireless communication network; and performing a data integration step S13 of receiving sensor data from each intelligent sensor node and generating an original monitoring data set D; The step S2 of performing the spatiotemporal data alignment and outlier cleansing includes: a spatiotemporal alignment step S21, in which spatiotemporal alignment is performed on each sensor data in the original monitoring dataset through a timestamp t aligned; a step S22 of performing outlier cleansing, in which each sensor data is adjusted using a linear interpolation algorithm, and if the sensor data falls outside the reference range and exceeds a set threshold δ i , the sensor data is regarded as an outlier and marked or removed; performing a multimodal data fusion step S23, fusing the different sensor data using a multimodal data fusion algorithm to generate a fused data matrix D fusion containing spatiotemporal correlation features; The dynamic risk assessment model analysis in step S3 is performed by: A step S31 of performing spatiotemporal feature analysis, in which the generated fusion data matrix is ​​input into a dynamic risk assessment model to perform spatiotemporal feature analysis and generate a risk score; Step S32 of performing geological and historical data analysis, combining the geomechanical parameter library and the historical disaster case library, generating risk adjustment factors, modifying the risk score, and generating a final area risk score; and outputting a risk level map (S33) based on the final area risk score, generating a risk level map; The dynamic risk assessment model described in step S31 adopts a graph convolutional network model, and constructing the graph convolutional network model includes: Step S311 of constructing a graph structure, in which each sensor data represents one node, the spatial relationship between every two sensors is represented by an edge, and the relationship between the nodes is represented by an adjacency matrix A; Step S312 initializes node features by generating an initial feature matrix H 0 using data X i (t) at time t of each sensor in the generated fusion data matrix D fusion as an initial input; Step S313 performs a graph convolution operation, which involves transmitting information through an adjacency matrix A, obtaining a weighted average between the feature vector of each node and the features of its adjacent nodes, and transmitting the information to its adjacent nodes through a convolutional layer; Step S314 of performing a multi-layer graph convolution operation, in which a node feature is convolved with an adjacent matrix and a feature of a previous layer in each layer by convolving multiple graph convolution layers, and a final node feature H (L) is extracted step by step; and outputting a list evaluation in step 315, by performing multiple graph convolution calculations to obtain a feature representation of each node and outputting a risk score R of the node using a softmax function; The multi-level early warning and linked control described in step S4 is A step S41 of performing a gradient change feature analysis on the generated risk level map to calculate a gradient change value (Gradient) by performing a gradient change feature analysis on the generated risk level map and calculating a gradient change between the risk level of each zone and the risk level of an adjacent zone; triggering a multi-level early warning mechanism in step S42, which is based on the calculated gradient change value and compares it with a corresponding early warning threshold value, and triggers a corresponding multi-level early warning mechanism, including a low-risk early warning, a medium-risk early warning, and a high-risk early warning; and generating a linked control command (S43), based on the corresponding early warning mechanism, generating a corresponding linked control command and attaching treatment suggestions.

2. The geological and past data analysis described in step S32 is a step S321 of generating risk correction coefficients, which involves obtaining geological characteristics and past disaster cases related to the bank slope by examining a geomechanical parameter library, and generating a corresponding risk correction coefficient Q for each area; The integrated bank slope monitoring method based on multi-parameter collaborative recognition as claimed in claim 1, further comprising a step S322 of correcting the list score, that is, correcting the risk score of each zone based on the generated risk correction coefficient Q to obtain a final zone risk score R'.

3. The step S33 of outputting the risk level map includes: Final area risk score for each area Ｒ’ Based on this, the risk level of each area is determined by comparing with the preset risk level threshold, and is represented by the following formula: [Equation 31] where Risk Level(i) is the risk level of the i-th area, and T ｌｏｗ is the low-risk threshold, and T ｈｉｇｈ a step S331 of generating a list level, where is a high risk threshold; and a step S332 of generating a risk level map by assigning a color to the risk level of each area, thereby finally generating a risk level map.

4. Triggering the multi-level early warning mechanism in step S42 includes: [Equation 32] If T, it indicates that the early warning level of the area is low risk early warning, where T ｌｏｗ a step S421 of issuing a low-risk early warning, where ' is a low-risk early warning threshold; [Equation 33] If T, it indicates that the early warning level of the area is medium risk early warning, where T ｈｉｇｈ Step S422 of issuing a medium risk early warning, where ' is a high risk early warning threshold; [Equation 34] and if so, issuing a high-risk early warning, that is, if so, the early warning level of the area is a high-risk early warning.

5. The generation of the interlocking control command described in S43 includes: Step S431: issuing a low-risk warning command to the low-risk area, maintaining normal monitoring and updating the risk assessment data periodically while maintaining the normal operation of the early warning system; Step S432: issuing a medium-risk early warning command for the medium-risk area, strengthening real-time monitoring, issuing warning information, strengthening defenses in the disaster-hit area, and adjusting the operation and maintenance strategies of related facilities; and step S433 of issuing a high-risk early warning command for high-risk areas, to activate an emergency response program, including scheduling an emergency rescue team, issuing an order to close off the area or evacuate personnel, and increasing the monitoring frequency.

Images