An embankment piping monitoring management system based on artificial intelligence

Abstract

This invention discloses an AI-based levee piping monitoring and management system, which relates to the field of levee safety monitoring technology. This AI-based levee piping monitoring and management system identifies piping anomaly characteristics and then assesses the impact of various factors based on these characteristics. Different levels of parameters are generated based on these impacts, weighted, and weighted to produce an early warning assessment value. Based on the early warning assessment value, different levels of early warning commands are generated and transmitted. A piping prediction model is established to analyze and process the collected data, accurately determining whether there is piping risk in the levee and the severity of the piping. It can analyze historical data to uncover the patterns and influencing factors of piping occurrence, predict the future state of the levee, and issue corresponding early warning signals according to different warning levels, improving the accuracy and timeliness of early warnings and achieving adaptive piping monitoring and management operations.

Description

Technical Field

[0001] This invention relates to the field of dike safety monitoring technology, specifically to an artificial intelligence-based dike piping monitoring and management system. Background Technology

[0002] Dike engineering is an important component of the flood control system, and it plays a vital role in protecting people's lives and property and maintaining social stability. However, during long-term operation, dikes are susceptible to various factors such as water erosion, geological changes, and construction quality, which can lead to problems such as piping. Piping refers to the phenomenon where fine particles in the soil are carried away through the pores between coarse particles by seepage, resulting in the destruction of the soil structure. Once piping occurs, if it is not detected and dealt with in time, it may expand rapidly and eventually lead to dike breaches and flood disasters.

[0003] The reference patent is titled: "A Monitoring and Early Warning System and Method for Water Inflow in Underground Utility Tunnels" (Patent Publication No.: CN114252128A, Patent Publication Date: 2022-03-29). It includes: a wireless automatic rain gauge for real-time monitoring of rainfall; a piezometer for real-time monitoring of pore water pressure; a weir level gauge for real-time monitoring of drainage ditch water inflow; and a central control center, including data processing software, which calculates the acceleration of rainfall increase and decrease, the acceleration of groundwater level rise and fall, and the acceleration of water level rise and fall in drainage ditches. By integrating rainfall, groundwater level, and drainage ditch water inflow data, it predicts water inflow data within the utility tunnel under different rainfall amounts and extreme rainfall conditions, and proposes processing measures. It actively captures meteorological information outside the utility tunnel and water inflow information within the tunnel, uses computer algorithms to fuse multi-channel monitoring data, and perceives and identifies changes in the external environment in real time. It provides early warning and intelligent solutions to drainage problems in the utility tunnel, minimizing the harm caused by leakage.

[0004] Based on the description in the above documents, existing methods for monitoring piping mainly rely on manual inspections and some simple monitoring equipment. They can often only monitor a single parameter and have limited data processing and analysis capabilities, making it difficult to accurately determine the occurrence and development trend of piping. These methods suffer from low monitoring efficiency, poor accuracy, and difficulty in real-time monitoring of piping development. Therefore, this invention provides an artificial intelligence-based levee piping monitoring and management system. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an artificial intelligence-based monitoring and management system for piping in dikes. This system solves the problems of existing piping monitoring methods, which mainly rely on manual inspections and some simple monitoring equipment. Furthermore, the specific situation of piping is easily affected by other factors, and current monitoring methods suffer from low monitoring efficiency, poor accuracy, and difficulty in grasping the development trend of piping in real time.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an artificial intelligence-based monitoring and management system for dike piping, comprising:

[0007] The data acquisition module uses high-definition cameras and various sensors deployed along the dike to collect real-time data related to piping.

[0008] The data transmission module transmits the collected data to the analysis and early warning module via wireless communication technology, and simultaneously transmits it to the database for storage.

[0009] The analysis and early warning module performs preprocessing operations on the collected data, extracts the preprocessed data to establish a piping prediction model, and determines the abnormal characteristics of piping. Based on the abnormal characteristics of piping, it judges the impact of various factors, generates different levels of parameters based on the impact, assigns weights to them to obtain the early warning evaluation value, and generates different levels of early warning instructions based on the early warning evaluation value for transmission.

[0010] The decision management module traces the location information of the piping failure based on the received early warning instructions and matches it with the corresponding maintenance strategy, transmitting it to personnel for processing.

[0011] Preferably, the real-time acquisition of various data related to piping in the data acquisition module is performed as follows:

[0012] The collected data includes image data and parameter data. The image data is collected by a high-definition camera, while the parameter data is collected by a variety of sensors deployed along the dike.

[0013] The distance between sensors deployed along the dike is determined by the image data acquisition range. The image data acquisition range is consistent, and the horizontal and vertical distances of the image data acquisition range are A and B, respectively. The horizontal distance between adjacent sensors C1 satisfies A / 2 < C1 < A, while the vertical distance between adjacent sensors C2 satisfies B / 2 < C2 < B.

[0014] It also integrates image data to form complete dike image data.

[0015] Preferably, the operation of integrating the image data to form complete levee image data is as follows:

[0016] Select any image data as the initial base image and determine the direction of the dike feature in the current base image features;

[0017] If the embankment feature direction is set vertically, then keep the initial base image, remove the (B-C2) distance from the side of the image to be stitched that is close to the initial base image, and then complete the image stitching by aligning the adjacent side corner points;

[0018] If the embankment feature direction is set horizontally, then keep the initial base image, remove the (A-C1) distance from the side of the image to be stitched that is close to the initial base image, and then complete the image stitching by aligning the adjacent side corner points;

[0019] If the dike features are set at the corners of both the longitudinal and transverse directions, then the initial base image is maintained, and the image stitching operation is performed sequentially in both directions.

[0020] Preferably, the wireless communication technology in the data transmission module is:

[0021] The transmission operation is achieved by using LoRa+5G dual-mode communication. LoRa is used to build the network transmission between the data acquisition nodes, and 5G is used to upload data to the database for storage in real time.

[0022] Furthermore, it identifies the frequency of interference signals during transmission and selects an interference-free transmission channel for data transmission.

[0023] Preferably, the analysis and early warning module performs preprocessing operations on the collected data as follows:

[0024] The original data is preprocessed. Data cleaning is used to handle missing values, outliers and duplicate values ​​in the data. Data denoising is used to eliminate random errors or irrelevant signals in the data. Data normalization is used to scale the data to a uniform range.

[0025] Extraction operations are performed on the preprocessed data; the piping prediction model is associated with a search and extraction function.

[0026] The search and extraction function uses the data types required by the piping prediction model to select sensors in different locations as the first search label and the title of the required data type as the second search label. The content of the second search label is input based on the first search label after the search is completed. The data after the two searches are completed is extracted to the corresponding piping prediction model for use.

[0027] Preferably, the operation of establishing the piping prediction model in the analysis and early warning module is as follows:

[0028] A regional model simulating real-time conditions is formed by combining GIS geographic information with parameter information collected in the current area.

[0029] By introducing historical data into the regional model, dynamic pre-training is achieved, ultimately forming a piping prediction model. The piping prediction model is updated synchronously after real-time data updates.

[0030] Preferably, the operation in the analysis and early warning module that determines the characteristics of piping anomalies and then judges the impact of various factors based on these characteristics is as follows:

[0031] The complete levee image data is converted to grayscale, and a boundary of the levee features is selected as the basic boundary. Multiple detection boundaries parallel to the basic boundary are set at equal intervals on the outer slope of the levee.

[0032] Based on the equidistant setting of detection points on the detection boundary, different features are determined according to the gray value changes of the detection points. Then, the different features are matched with the known feature database to determine the soil moisture characteristics, and the soil moisture characteristics are analyzed to determine the piping anomaly characteristics.

[0033] Select a piping anomaly feature and extract the parameters of the current piping anomaly feature location. The influencing parameters are obtained by combining the water level height, hydraulic gradient and piping changes.

[0034] Preferably, the operation of analyzing the soil moisture characteristics to determine the piping anomaly characteristics is as follows:

[0035] By extracting image data of the same location at different time periods, the image data of the two time periods are compared and overlapped, and the soil moisture features of the upper image data are made transparent.

[0036] Furthermore, if the range of soil moisture characteristics in later image data is greater than that in the initial image data, it is identified as a piping anomaly and a seepage phenomenon. If the soil moisture characteristics show dynamic water outflow, then the piping anomaly is identified as a piping phenomenon.

[0037] Preferably, the operation of deriving the influencing parameters by combining water level height, hydraulic gradient, and piping changes is as follows:

[0038] The location of the internal outlet and external piping outlet at the current piping location is determined, and the head difference is denoted as ΔH, where ΔH = |height of internal outlet - height of external piping outlet|. Based on this, the hydraulic gradient is calculated: I i =ΔH / L, I i Let U be the hydraulic gradient at position i, and L be the flow length during the water infiltration process obtained from the levee design drawings. The calculated hydraulic gradient is compared with the hydraulic gradient threshold to obtain the hydraulic gradient influence parameter, labeled U. x ;

[0039] The water pressure difference within the current dike is calculated as ΔP = ρgh, where ΔP is the pressure difference between the inner outlet at the piping location and the water surface, ρ is the liquid density, g is the acceleration due to gravity, and h is the vertical distance from the liquid surface to the inner outlet at the piping location. The calculated water pressure difference is then compared with a water pressure difference threshold to derive the water pressure difference influence parameter, denoted as V. y ;

[0040] The channel area of ​​the external piping outlet at the current piping location is collected, and the seepage flow rate through the dike per unit time is calculated: Q i =k×I i ×S, Q i Let Q be the seepage flow rate at the i-th location, S be the channel area of ​​the external piping outlet, k be the permeability coefficient, and the calculated seepage flow rate Q be... i The influencing parameter of seepage flow, labeled W, is obtained by comparing it with the seepage flow threshold. z .

[0041] Preferably, the operation of assigning weights to parameters of different levels to obtain the early warning evaluation value in the analysis and early warning module is as follows:

[0042] Achieve influence on hydraulic gradient parameter U x Water pressure difference affects parameter V y And the influence parameter W of seepage flow z Perform weight assignment;

[0043] The expression for the early warning assessment value is: F = α × U x +β×V y +γ×W z ;

[0044] F is the early warning assessment value, α is the weight value of the hydraulic gradient influence parameter, β is the weight value of the water pressure difference influence parameter, and γ is the weight value of the seepage flow influence parameter.

[0045] The warning threshold range is set to [d, e], and different levels of warning instructions are generated based on the warning assessment value as follows:

[0046] If F < [d, e], a Level 1 warning instruction is formed; if F ∈ [d, e], a Level 2 warning instruction is formed; if F > [d, e], a Level 3 warning instruction is formed, and the severity of the Level 1 to Level 3 warning instructions increases sequentially.

[0047] This invention provides an artificial intelligence-based monitoring and management system for levee piping. Compared with existing technologies, it has the following advantages:

[0048] 1. This AI-based levee piping monitoring and management system preprocesses the collected data, extracts the preprocessed data to establish a piping prediction model, identifies piping anomaly characteristics, and judges the impact of various factors based on these characteristics. Different levels of parameters are generated based on these impacts, weighted, and weighted to produce an early warning assessment value. Based on the early warning assessment value, different levels of early warning commands are generated and transmitted. The piping prediction model analyzes and processes the collected data, accurately determining whether there is piping risk in the levee and the severity of the piping. It can analyze historical data to uncover the patterns and influencing factors of piping occurrence, predict the future state of the levee, and issue corresponding early warning signals according to different warning levels, improving the accuracy and timeliness of early warnings and achieving adaptive piping monitoring and management operations.

[0049] 2. This AI-based levee piping monitoring and management system determines the distance of sensors deployed along the levee by collecting image data. It selects any image data as the initial base image, determines the levee feature direction in the current base image features, and completes the formation of complete image data by sequentially stitching images together. This enables better subsequent analysis of levee piping monitoring, avoids repetitive data processing, and ensures the efficiency of piping monitoring and management.

[0050] 3. This AI-based levee piping monitoring and management system performs grayscale processing on complete levee image data and selects a boundary of the levee features as the basic boundary. Multiple detection boundaries, parallel to the basic boundary, are set at equal intervals on the outer slope of the levee. Detection points are set at equal intervals on these boundaries, and different features are determined based on the grayscale value changes of the detection points. These distinguished features are then matched with a known feature database to determine soil moisture characteristics. Analysis of these soil moisture characteristics identifies piping anomaly characteristics. A piping anomaly is selected, and parameters at its location are extracted. Influencing parameters are derived by combining water level height, hydraulic gradient, and piping changes. This adaptively enables early warning assessment, improves the completeness of levee piping monitoring and management, and adaptively matches strategies according to the level to ensure levee safety. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the piping monitoring and management system of the present invention.

[0052] Figure 2 This is a flowchart illustrating the data preprocessing operation of the present invention.

[0053] Figure 3 The flowchart illustrates the operation of determining the characteristics of piping anomalies in this invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Please see Figures 1-3 This invention provides two technical solutions:

[0056] Example 1: An artificial intelligence-based monitoring and management system for dike piping, comprising:

[0057] The data acquisition module uses high-definition cameras and various sensors deployed along the dike to collect real-time data related to piping.

[0058] The data transmission module transmits the collected data to the analysis and early warning module via wireless communication technology, and simultaneously transmits it to the database for storage.

[0059] The analysis and early warning module performs preprocessing operations on the collected data, extracts the preprocessed data to establish a piping prediction model, and determines the abnormal characteristics of piping. Based on the abnormal characteristics of piping, it judges the impact of various factors, generates different levels of parameters based on the impact, assigns weights to them to obtain the early warning evaluation value, and generates different levels of early warning instructions based on the early warning evaluation value for transmission.

[0060] The decision management module traces the location information of the piping failure based on the received early warning instructions and matches it with the corresponding maintenance strategy, transmitting it to personnel for processing.

[0061] By preprocessing the collected data and extracting the preprocessed data to establish a piping prediction model, and determining the characteristics of piping anomalies, the model assesses the impact of various factors based on these characteristics. Different levels of parameters are generated based on these impacts, weighted, and weighted to produce an early warning assessment value. Based on this assessment value, different levels of early warning commands are generated and transmitted. This piping prediction model analyzes and processes the collected data, accurately determining whether piping risks exist in the dikes and the severity of the piping. It can also analyze historical data to uncover the patterns and influencing factors of piping occurrence, predict the future state of the dikes, and issue corresponding early warning signals according to different warning levels, improving the accuracy and timeliness of early warnings and enabling adaptive piping monitoring and management.

[0062] In this embodiment of the invention, the real-time acquisition of various data related to piping in the data acquisition module is as follows:

[0063] The collected data includes image data and parameter data. The image data is collected by a high-definition camera, while the parameter data is collected by a variety of sensors deployed along the dike.

[0064] The distance between sensors deployed along the dike is determined by the image data acquisition range. The image data acquisition range is consistent, and the horizontal and vertical distances of the image data acquisition range are A and B, respectively. The horizontal distance between adjacent sensors C1 satisfies A / 2 < C1 < A, while the vertical distance between adjacent sensors C2 satisfies B / 2 < C2 < B.

[0065] It also integrates image data to form complete dike image data.

[0066] In this embodiment of the invention, the operation of integrating image data to form complete dike image data is as follows:

[0067] Select any image data as the initial base image and determine the direction of the dike feature in the current base image features;

[0068] If the embankment feature direction is set vertically, then keep the initial base image, remove the (B-C2) distance from the side of the image to be stitched that is close to the initial base image, and then complete the image stitching by aligning the adjacent side corner points;

[0069] If the embankment feature direction is set horizontally, then keep the initial base image, remove the (A-C1) distance from the side of the image to be stitched that is close to the initial base image, and then complete the image stitching by aligning the adjacent side corner points;

[0070] If the dike features are set at the corners of both the longitudinal and transverse directions, then the initial base image is maintained, and the image stitching operation is performed sequentially in both directions.

[0071] By determining the distance of sensors deployed along the dike by collecting image data, and selecting any image data as the initial base image, the direction of the dike features in the current base image is determined. The image stitching operation is then performed to complete the formation of complete image data. This enables better subsequent analysis of dike piping monitoring, avoids repetitive data processing, and ensures the efficiency of piping monitoring and management.

[0072] In this embodiment of the invention, the wireless communication technology in the data transmission module is:

[0073] The transmission operation is achieved by using LoRa+5G dual-mode communication. LoRa is used to build the network transmission between the data acquisition nodes, and 5G is used to upload data to the database for storage in real time.

[0074] Furthermore, it identifies the frequency of interference signals during transmission and selects an interference-free transmission channel for data transmission.

[0075] In this embodiment of the invention, the preprocessing operation of the collected data in the analysis and early warning module is as follows:

[0076] The original data is preprocessed. Data cleaning is used to handle missing values, outliers and duplicate values ​​in the data. Data denoising is used to eliminate random errors or irrelevant signals in the data. Data normalization is used to scale the data to a uniform range.

[0077] Extraction operations are performed on the preprocessed data; the piping prediction model is associated with a search and extraction function.

[0078] The search and extraction function uses the data types required by the piping prediction model to select sensors in different locations as the first search label and the title of the required data type as the second search label. The content of the second search label is input based on the first search label after the search is completed. The data after the two searches are completed is extracted to the corresponding piping prediction model for use.

[0079] In this embodiment of the invention, the operation of establishing a piping prediction model in the analysis and early warning module is as follows:

[0080] A regional model simulating real-time conditions is formed by combining GIS geographic information with parameter information collected in the current area.

[0081] By introducing historical data into the regional model, dynamic pre-training is achieved, ultimately forming a piping prediction model. The piping prediction model is updated synchronously after real-time data updates.

[0082] In this embodiment of the invention, the operation of determining the impact of various factors based on the piping anomaly characteristics after identifying them in the analysis and early warning module is as follows:

[0083] The complete levee image data is converted to grayscale, and a boundary of the levee features is selected as the basic boundary. Multiple detection boundaries parallel to the basic boundary are set at equal intervals on the outer slope of the levee.

[0084] Based on the equidistant setting of detection points on the detection boundary, different features are determined according to the gray value changes of the detection points. Then, the different features are matched with the known feature database to determine the soil moisture characteristics, and the soil moisture characteristics are analyzed to determine the piping anomaly characteristics.

[0085] Select a piping anomaly feature and extract the parameters of the current piping anomaly feature location. The influencing parameters are obtained by combining the water level height, hydraulic gradient and piping changes.

[0086] In this embodiment of the invention, the operation of analyzing soil moisture characteristics to determine the characteristics of piping anomalies is as follows:

[0087] By extracting image data of the same location at different time periods, the image data of the two time periods are compared and overlapped, and the soil moisture features of the upper image data are made transparent.

[0088] Furthermore, if the range of soil moisture characteristics in later image data is greater than that in the initial image data, it is identified as a piping anomaly and a seepage phenomenon. If the soil moisture characteristics show dynamic water outflow, then the piping anomaly is identified as a piping phenomenon.

[0089] By performing grayscale processing on complete levee image data and selecting one boundary of the levee features as the basic boundary, multiple detection boundaries parallel to the basic boundary are set at equal intervals on the outer slope of the levee. Detection points are set at equal intervals on the detection boundaries, and different features are determined based on the grayscale value changes of the detection points. Then, the distinguished features are matched with a known feature database to determine soil moisture characteristics, and the soil moisture characteristics are analyzed to determine piping anomaly characteristics. A piping anomaly is selected, and the parameters of the current piping anomaly location are extracted. By combining water level height, hydraulic gradient, and piping changes, the influencing parameters are obtained, thereby adaptively realizing early warning assessment, improving the completeness of levee piping monitoring and management, and adaptively matching strategies according to the level to ensure the safety of the levee.

[0090] In this embodiment of the invention, the operation of deriving the influencing parameters by combining water level height, hydraulic gradient, and piping changes is as follows:

[0091] The location of the internal outlet and external piping outlet at the current piping location is determined, and the head difference is denoted as ΔH, where ΔH = |height of internal outlet - height of external piping outlet|. Based on this, the hydraulic gradient is calculated: I i =ΔH / L, I i Let U be the hydraulic gradient at position i, and L be the flow length during the water infiltration process obtained from the levee design drawings. The calculated hydraulic gradient is compared with the hydraulic gradient threshold to obtain the hydraulic gradient influence parameter, labeled U. x ;

[0092] The water pressure difference within the current dike is calculated as ΔP = ρgh, where ΔP is the pressure difference between the inner outlet at the piping location and the water surface, ρ is the liquid density, g is the acceleration due to gravity, and h is the vertical distance from the liquid surface to the inner outlet at the piping location. The calculated water pressure difference is then compared with a water pressure difference threshold to derive the water pressure difference influence parameter, denoted as V. y ;

[0093] The channel area of ​​the external piping outlet at the current piping location is collected, and the seepage flow rate through the dike per unit time is calculated: Q i =k×I i ×S, Q i Let Q be the seepage flow rate at location i, S be the channel area of ​​the external piping outlet, and k be the permeability coefficient, where the permeability coefficient k is a parameter predetermined based on soil survey data at the corresponding location of the dike. The calculated seepage flow rate Q will then be used to calculate the seepage flow rate. i The influencing parameter of seepage flow, labeled W, is obtained by comparing it with the seepage flow threshold. z .

[0094] Among them, the hydraulic gradient influence parameter U x The method for determining it is as follows:

[0095] For example, the hydraulic gradient threshold is set to [I r I q ], and I i <I r When, then the current U x =1, while I i ∈[I r I q When ], then the current U x =2, if I i＞ I q When, then the current U x =3;

[0096] The influence parameter V of water pressure difference can be deduced by analogy. y And the influence parameter W of seepage flow z A comparison.

[0097] In this embodiment of the invention, the operation of assigning weights to parameters of different levels in the analysis and early warning module to obtain the early warning evaluation value is as follows:

[0098] Achieve influence on hydraulic gradient parameter U x Water pressure difference affects parameter V y And the influence parameter W of seepage flow z Perform weight assignment;

[0099] The expression for the early warning assessment value is: F = α × U x +β×V y +γ×W z ;

[0100] F is the early warning assessment value, α is the weight value of the hydraulic gradient influence parameter, β is the weight value of the water pressure difference influence parameter, and γ is the weight value of the seepage flow influence parameter.

[0101] The warning threshold range is set to [d, e], and different levels of warning instructions are generated based on the warning assessment value as follows:

[0102] If F < [d, e], a Level 1 warning instruction is formed; if F ∈ [d, e], a Level 2 warning instruction is formed; if F > [d, e], a Level 3 warning instruction is formed, and the severity of the Level 1 to Level 3 warning instructions increases sequentially.

[0103] Example 2 differs from Example 1 in that: Monitoring operations are performed on the same area using both the existing monitoring and management system and the dike piping monitoring and management system of this invention. The piping anomaly characteristics and specific conditions of the current area are known. The detection of piping anomalies and the false positive rate during the detection process are recorded, and the recorded results are matched with the known conditions. Specific results are shown in Table 1.

[0104] Table 1 Monitoring Record Sheet

[0105]

[0106] Experimental results show that the system of the present invention is significantly better than the comparison system in terms of timeliness (real-time detection) and accuracy (100%) in detecting abnormal features, and the false judgment rate (2%) is much lower than that of the comparison system (21%), which proves its superiority and reliability in practical applications.

[0107] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0108] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0109] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based monitoring and management system for dike piping, characterized in that: include: The data acquisition module uses high-definition cameras and various sensors deployed along the dike to collect real-time data related to piping. The data transmission module transmits the collected data to the analysis and early warning module via wireless communication technology, and simultaneously transmits it to the database for storage. The analysis and early warning module performs preprocessing operations on the collected data, extracts the preprocessed data to establish a piping prediction model, and determines the abnormal characteristics of piping. Based on the abnormal characteristics of piping, it judges the impact of various factors, generates different levels of parameters based on the impact, assigns weights to them to obtain the early warning evaluation value, and generates different levels of early warning instructions based on the early warning evaluation value for transmission. The decision management module, based on the received early warning instructions, traces the location information of the piping and matches it with the corresponding maintenance strategy, transmitting it to personnel for processing. The data acquisition module performs the following operations to collect various data related to piping in real time: The collected data includes image data and parameter data. The image data is collected by a high-definition camera, while the parameter data is collected by a variety of sensors deployed along the dike. The distance between sensors deployed along the dike is determined by the image data acquisition range. The image data acquisition range is consistent, and the horizontal and vertical distances of the image data acquisition range are A and B, respectively. The horizontal distance between adjacent sensors C1 satisfies A / 2 < C1 < A, while the vertical distance between adjacent sensors C2 satisfies B / 2 < C2 < B. And achieve the integration of image data to form complete dike image data; The operation in the analysis and early warning module that determines the characteristics of piping anomalies and then judges the impact of various factors based on these characteristics is as follows: The complete levee image data is converted to grayscale, and a boundary of the levee features is selected as the basic boundary. Multiple detection boundaries parallel to the basic boundary are set at equal intervals on the outer slope of the levee. Based on the equidistant setting of detection points on the detection boundary, different features are determined according to the gray value changes of the detection points. Then, the different features are matched with the known feature database to determine the soil moisture characteristics, and the soil moisture characteristics are analyzed to determine the piping anomaly characteristics. Select a piping anomaly feature and extract the parameters of the current piping anomaly feature location. The influencing parameters are obtained by combining the water level height, hydraulic gradient and piping changes. The procedure for analyzing soil moisture characteristics to determine the characteristics of piping anomalies is as follows: By extracting image data of the same location at different time periods, the image data of the two time periods are compared and overlapped, and the soil moisture features of the upper image data are made transparent. If the range of soil moisture characteristics in later image data is greater than that in the initial image data, then it is determined to be a piping anomaly and a seepage phenomenon. If the soil moisture characteristics show dynamic water outflow, then the piping anomaly is a piping phenomenon. The operation of influencing parameters, derived by combining water level height, hydraulic gradient, and piping changes, is as follows: The location of the internal outlet and external piping outlet at the current piping location is determined, and the head difference is denoted as ΔH, where ΔH = |height of internal outlet - height of external piping outlet|. Based on this, the hydraulic gradient is calculated: I i =ΔH / L, I i Let U be the hydraulic gradient at position i, and L be the flow length during the water infiltration process obtained from the levee design drawings. The calculated hydraulic gradient is compared with the hydraulic gradient threshold to obtain the hydraulic gradient influence parameter, labeled U. x ; The water pressure difference within the current dike is calculated as ΔP = ρgh, where ΔP is the pressure difference between the inner outlet at the piping location and the water surface, ρ is the liquid density, g is the acceleration due to gravity, and h is the vertical distance from the liquid surface to the inner outlet at the piping location. The calculated water pressure difference is then compared with a water pressure difference threshold to derive the water pressure difference influence parameter, denoted as V. y ; The channel area of ​​the external piping outlet at the current piping location is collected, and the seepage flow rate through the dike per unit time is calculated: Q i =k×I i ×S, Q i Let Q be the seepage flow rate at the i-th location, S be the channel area of ​​the external piping outlet, k be the permeability coefficient, and the calculated seepage flow rate Q be... i The influencing parameter of seepage flow, labeled W, is obtained by comparing it with the seepage flow threshold. z ; The operation in the analysis and early warning module to assign weights to parameters of different levels to obtain the early warning evaluation value is as follows: Achieve influence on hydraulic gradient parameter U x Water pressure difference affects parameter V y And the influence parameter W of seepage flow z Perform weight assignment; The expression for the early warning assessment value is: F = α × U x +β×V y +γ×W z ; F is the early warning assessment value, α is the weight value of the hydraulic gradient influence parameter, β is the weight value of the water pressure difference influence parameter, and γ is the weight value of the seepage flow influence parameter. The warning threshold range is set to [d, e], and different levels of warning instructions are generated based on the warning assessment value as follows: If F < [d, e], a Level 1 warning instruction is formed; if F ∈ [d, e], a Level 2 warning instruction is formed; if F > [d, e], a Level 3 warning instruction is formed, and the severity of the Level 1 to Level 3 warning instructions increases sequentially. The operation of integrating the image data to form complete levee image data is as follows: Select any image data as the initial base image and determine the direction of the dike feature in the current base image features; If the embankment feature direction is set vertically, then keep the initial base image, remove the (B-C2) distance from the side of the image to be stitched that is close to the initial base image, and then complete the image stitching by aligning the adjacent side corner points; If the embankment feature direction is set horizontally, then keep the initial base image, remove the (A-C1) distance from the side of the image to be stitched that is close to the initial base image, and then complete the image stitching by aligning the adjacent side corner points; If the dike features are set at the corners of both the longitudinal and transverse directions, then the initial base image is maintained, and the image stitching operation is performed sequentially in both directions.

2. The levee piping monitoring and management system based on artificial intelligence according to claim 1, characterized in that: The wireless communication technology in the data transmission module is: The transmission operation is achieved by using LoRa+5G dual-mode communication. LoRa is used to build the network transmission between the data acquisition nodes, and 5G is used to upload data to the database for storage in real time. Furthermore, it identifies the frequency of interference signals during transmission and selects an interference-free transmission channel for data transmission.

3. The levee piping monitoring and management system based on artificial intelligence according to claim 1, characterized in that: The analysis and early warning module performs the following preprocessing operations on the collected data: The original data is preprocessed. Data cleaning is used to handle missing values, outliers and duplicate values ​​in the data. Data denoising is used to eliminate random errors or irrelevant signals in the data. Data normalization is used to scale the data to a uniform range. Extraction operations are performed on the preprocessed data; the piping prediction model is associated with a search and extraction function. The search and extraction function uses the data types required by the piping prediction model to select sensors in different locations as the first search label and the title of the required data type as the second search label. The content of the second search label is input based on the first search label after the search is completed. The data after the two searches are completed is extracted to the corresponding piping prediction model for use.

4. The levee piping monitoring and management system based on artificial intelligence according to claim 1, characterized in that: The operation for establishing a piping prediction model in the analysis and early warning module is as follows: A regional model simulating real-time conditions is formed by combining GIS geographic information with parameter information collected in the current area. By introducing historical data into the regional model, dynamic pre-training is achieved, ultimately forming a piping prediction model. The piping prediction model is updated synchronously after real-time data updates.

Images