Slope anchor rod prestress monitoring system and method

Abstract

The invention discloses a slope anchor rod prestress monitoring system and method, and relates to the technical field of slope abnormity supervision. Comprising a data acquisition module, an acquisition and analysis module, a secondary analysis module and an anomaly analysis module, wherein the data acquisition module is used for acquiring multipoint stress data of a slope anchor rod, slope soil data and slope environment data and preprocessing the acquired data; according to the technical key points, prestressing forces at different positions of the side slope anchor rod can be collected, then the collected prestressing force data at different positions of the side slope anchor rod are evaluated, the attenuation of the prestressing force data is researched, whether the collected data are abnormal or not is further analyzed, and whether the data are abnormal or not can be found in time; therefore, the conditions of low efficiency and high cost caused by inaccurate analysis result, repeated acquisition and analysis due to abnormal acquired data can be effectively avoided, the use effect is good, and the method has a good use prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of slope abnormality supervision, and specifically to a slope anchor prestressing monitoring system and method. Background Art

[0002] The construction industry, as a pillar of the national economy, is playing an increasingly important role and making significant contributions to my country's economic and social development. However, the development and construction of numerous projects, such as urban construction, railway and bridge projects, and mining development, have significantly altered the surrounding geological and ecological environments, including river diversions and mountain leveling. Consequently, numerous soil or rock slopes are inevitably created around these projects.

[0003] During the operation phase of some projects, due to the continuous influence of natural factors such as long-term groundwater erosion, the gravity of the rock and soil, earthquakes, precipitation, and other adverse human factors, the slopes left behind around the project have undergone partial or overall deformation, resulting in a large number of geological disasters such as slope collapse and sliding. The occurrence of slope accidents shows that during the process of urban construction, most artificially modified slopes are directly or indirectly affected by changes in their internal geological conditions and external geomorphological environment. Therefore, an effective support method must be adopted for existing slopes to ensure their stability and safety.

[0004] The construction process of slope support is crucial. Even a perfectly sound design can have serious consequences if the construction process doesn't meet requirements. Only very stable slopes can be excavated without support. For slopes with average stability, excavation is typically carried out one layer at a time, followed by support. Construction is typically carried out in a reverse order, with support applied from the bottom up. For slopes with less stable slopes, pre-excavation support is required.

[0005] Frame prestressed anchor rods are widely used in the support of highway and railway slopes and foundation pits because they can give full play to the bearing capacity and self-stabilization ability of the rock mass and have excellent economy. However, the action mechanism of frame prestressed anchor rods is very complex and is affected by many factors such as climate, environment, and human activities. Therefore, the prestress applied to the frame prestressed anchor rods cannot maintain a stable size.

[0006] Anchor prestress loss is an important issue in slope reinforcement. It is not only an important factor of concern to engineering designers, but also an important issue related to project safety and quality. Therefore, it is necessary to monitor the anchor prestress.

[0007] The existing patent authorization announcement number is CN112095596B, and the patent name is the invention patent of the intelligent monitoring and early warning system and method for prestressed anchor rods on slopes based on cloud platform. The invention patent records that it includes S1, real-time collection of strain values ​​of various points on the anchor rod; S2, construction of anchor rod average shear stress prediction model; S3, calculation of the gap stress function relationship between the anchor rod and the anchor hole; S4, correction of anchor rod average shear stress prediction model to obtain anchor rod shear stress prediction model; S5, measurement of the absolute value D of the difference between the anchor rod shear stress value output by the prediction model and the anchor rod shear stress value measured by the magnetic flux sensor; S6, division of the absolute value D; S7, construction of a neural network model for accident early warning. The present invention uses an intelligent algorithm to construct a neural network model for accident early warning, and predicts the stress bearing condition of the current anchor rod through the real-time collected stress data. When the anchor rod stress bearing condition exceeds the preset value, the fault type is judged.

[0008] The central idea of ​​the scheme recorded in the above patent is to monitor the prestress data of the anchor rods in the slope and the corresponding layout data of the anchor rods, and to process the prestress data of the anchor rods in the slope and the corresponding layout data of the anchor rods using the corresponding anchor rod average shear stress prediction model. This method is suitable for areas with balanced and stable soil, but for slightly complex areas, the prestress of each position of the anchor rod varies greatly, and the soil is affected by the external environment, which will produce different changes. When reflected on the anchor rods, the degree of change is completely different. Using an average method and a fixed model for analysis, it averages the abnormal data, which is difficult to reflect the degree of local abnormality of the slope. By the time it reflects the slope abnormality, the slope abnormality in some areas is often more serious, and the cost of repair is relatively high. Therefore, its applicability is relatively low and does not meet the requirements of complex slopes. For this reason, we have developed a slope anchor prestress monitoring system and method. Summary of the Invention

[0009] (1) Technical problems solved

[0010] In response to the deficiencies in the prior art, the present invention provides a slope anchor prestress monitoring system and method, which can collect the prestress at different positions of the slope anchor, then evaluate the collected prestress data at different positions of the slope anchor, study its attenuation, and further analyze whether there are any abnormalities in the collected data. It can promptly detect whether there are any abnormalities in the data, thereby effectively avoiding inaccurate analysis results caused by abnormal collected data, the need for repeated collection and analysis, and the resulting low efficiency and high cost. The system has good use effect, good use prospects, and effectively solves the problems raised in the background technology.

[0011] (2) Technical solution

[0012] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0013] A slope anchor prestress monitoring system includes a data acquisition module, an acquisition and analysis module, a secondary analysis module, and an abnormality analysis module:

[0014] Data acquisition module: used to collect slope anchor multi-point stress data, slope soil data and slope environment data, and pre-process the collected data;

[0015] Acquisition and Analysis Module: Uses time series analysis technology to dynamically compare real-time collected data with historical data, calculates anchor prestress attenuation, and combines it with the previous monitoring anchor prestress attenuation data at the same monitoring point to analyze the difference ratio between the two. Based on the pre-set standard threshold range, it determines whether data collection is abnormal. If there is a data collection abnormality, the data re-acquisition mechanism is triggered;

[0016] Secondary analysis module: When data collection is normal, current and historical soil parameters and anchor mechanical parameter data are input into the potential analysis model to predict the prestress value at each monitoring point of the anchor, analyze the range adjustment coefficient, and construct the prestress prediction confidence interval. By comparing the real-time monitoring prestress data with the predicted confidence interval, potential abnormalities can be identified;

[0017] Abnormal analysis module: When there is a potential abnormal situation, the real-time and last monitored soil data, anchor parameters, predicted confidence intervals and measured anchor prestress data are input into the slope analysis model to obtain the abnormal risk value. The abnormality degree is evaluated based on the size of the abnormal risk value, and corresponding early warning measures are implemented.

[0018] Furthermore, multi-point stress data of slope anchors are collected through fiber optic Bragg grating stress sensors and fiber optic demodulators installed at N positions along the length of each slope anchor. Slope soil data include slope soil density and moisture content, slope soil particle gradation and porosity, slope soil cohesion and internal friction angle, slope soil compression modulus and deformation modulus, foundation bearing capacity and slope soil permeability coefficient. Slope environmental data include slope gradient and height, slope precipitation data, and slope temperature and humidity data.

[0019] Furthermore, the steps for preprocessing the collected data are as follows:

[0020] Statistical summary: Use data statistical techniques to comprehensively sort out the collected data, use decision tree classification algorithms to classify the data, use data aggregation functions to perform summary operations, and generate structured data summary tables;

[0021] Data testing: Use data quality assessment indicators to audit data and check whether there are missing values ​​and duplicate records in the data;

[0022] Missing data processing: If missing data is detected, the data collection process will be initiated according to the missing data mechanism to collect the missing data;

[0023] Duplicate processing: If data duplication is detected, a data hash deduplication algorithm is used to locate and delete the duplicate records originally collected. At the same time, the data recollection process is initiated according to the data duplication mechanism to collect duplicate data.

[0024] Furthermore, dynamic comparison is to compare the slope anchor prestress collected in real time with the slope anchor prestress collected at the corresponding position last time, and calculate the anchor prestress attenuation Psj by the formula Psj = Pls-Pss, where Pls is the slope anchor prestress collected at a certain position last time, and Pss is the slope anchor prestress collected at a certain position in real time, and obtain N groups of anchor prestress attenuation;

[0025] Determine the positive and negative values ​​of the prestress attenuation of N groups of anchor rods. If the prestress attenuation of N groups of anchor rods is negative, it is determined that there is an abnormality in data collection. If the prestress attenuation of N groups of anchor rods is positive, it is marked as normal.

[0026] Get the normal anchor prestress attenuation amount corresponding to the last anchor prestress attenuation amount Psjs at the monitoring position;

[0027] Calculate the difference between the normal anchor prestress attenuation and the previous anchor prestress attenuation at the corresponding monitoring position.

[0028] Compare and analyze the difference ratio P with the standard threshold interval. If P is within the standard threshold interval, it is determined that the data collection is normal and no processing is performed;

[0029] If P is outside the standard threshold range, the data collection is judged to be abnormal, and the data re-collection mechanism is triggered to collect K consecutive data of the anchor prestress at this location, and obtain the maximum, minimum and average values ​​of the data. If the difference between the maximum and minimum values ​​is less than 3% of the average value, the average value is used as the anchor prestress at this location, and the data collection abnormality at this location is eliminated;

[0030] If the difference between the maximum and minimum values ​​is greater than or equal to 3% of the average value, it is marked as an abnormal supervision state and an abnormal supervision warning signal is issued.

[0031] Furthermore, the potential analysis model is a hybrid model based on a convolutional neural network and a long short-term memory network. The steps of predicting the prestress value of each monitoring point of the anchor rod using the potential analysis model are as follows:

[0032] The spatial feature vector is extracted based on the convolutional neural network branch, and the temporal feature vector is extracted based on the long short-term memory network branch;

[0033] The spatial feature vector and the temporal feature vector are subjected to feature splicing, fusion and dimensionality reduction processing to form a spatiotemporal feature vector;

[0034] Feature preprocessing, Dropout regularization and multi-layer perceptron mapping are performed based on the spatiotemporal feature vector to obtain the prestress prediction value γ of N points.

[0035] Furthermore, the steps to analyze the range adjustment coefficient are as follows:

[0036] Obtain the most recent P group of historical slope anchor prestress prediction data and actual monitoring data at the monitoring point;

[0037] Calculate the difference ratio between the historical slope anchor prestress prediction data and the actual monitoring data of the monitoring point group P to obtain the difference ratio data set;

[0038] The exponential smoothing method in time series analysis is used to process the data in the difference ratio dataset, and the historical difference ratio data are used to predict the difference ratio at the current moment, and the predicted difference ratio is used as the range adjustment coefficient ε.

[0039] Furthermore, based on the range adjustment coefficient ε of each monitoring point, the prestress prediction confidence interval [γ-γ×ε, γ+γ×ε] of each monitoring point is calculated. If the real-time monitoring prestress data Pss of the monitoring point is within [γ-γ×ε, γ+γ×ε], the monitoring point is judged to be normal. If the real-time monitoring prestress data Pss of the monitoring point is outside [γ-γ×ε, γ+γ×ε], the monitoring point is judged to be potentially abnormal.

[0040] Furthermore, the slope analysis model uses a trained convolutional neural network, and the training steps of the convolutional neural network are as follows:

[0041] Data collection and annotation: retrieve historical data and mark several groups of data time points with potential anomalies in the historical data;

[0042] Obtain soil data, prestress prediction interval, measured anchor prestress data, and abnormality degree data at the data time point of potential abnormal conditions. Use data mining technology to further extract the soil data monitored last time at the data time point of the abnormal conditions. Then, normalize and integrate the extracted data into an abnormality data package. The abnormality degree data includes the characteristic data reflecting the abnormality degree and its corresponding abnormality risk value.

[0043] Aggregate several groups of abnormal data packets to form an abnormal data set, and divide the abnormal data set into a training set and a validation set;

[0044] Construct a convolutional neural network to perform regression tasks. Use the data in the training set except the abnormal risk value as the input of the convolutional neural network, and use the label of the characteristic data reflecting the degree of abnormality as the output. Train the convolutional neural network to obtain the initial convolutional neural network.

[0045] The initial convolutional neural network is model verified using the validation set, and the mean square error evaluation index is used to output the initial convolutional neural network with a test accuracy greater than or equal to the preset threshold as the trained convolutional neural network.

[0046] Furthermore, the abnormality degree is evaluated based on the abnormal risk value, including:

[0047] Set the abnormality threshold, which includes YC1 and YC2, and YC1>YC2>0;

[0048] Compare the abnormal risk value with the abnormality degree threshold;

[0049] If the abnormal risk value is greater than or equal to the abnormality threshold YC1, the slope abnormality is determined to be severe, the monitoring point is marked, the corresponding abnormality level light flashes as a prompt, and an abnormality report and abnormal area blockade report are sent through the emergency management platform;

[0050] If the abnormal risk value is less than the abnormality threshold YC1 and greater than or equal to the abnormality threshold YC2, the slope abnormality is determined to be moderate, and the monitoring point is marked. A light corresponding to the abnormality level flashes as a prompt. A manual inspection work order is sent to the operation and maintenance platform for regular manual verification. A spatiotemporal heat map is generated, and the abnormality diffusion trend is marked.

[0051] If the abnormal risk value is less than the abnormality threshold YC2, the slope abnormality is judged to be mild. The data is verified. After verification, the monitoring period is shortened, the monitoring point is marked, and the light corresponding to the abnormality level is flashed. At the same time, the data is included in the next round of convolutional neural network as a training sample.

[0052] Furthermore, a slope anchor prestressing monitoring method includes the following steps:

[0053] Collect slope anchor multi-point stress data, slope soil data, and slope environment data, and pre-process the collected data;

[0054] Use time series analysis technology to dynamically compare real-time collected data with historical data to calculate the anchor prestress attenuation. Combined with the previous monitoring anchor prestress attenuation data at the same monitoring point, analyze the difference ratio between the two, and refer to the pre-set standard threshold range to determine whether the data collection is abnormal. If there is a data collection abnormality, the data re-collection mechanism is triggered;

[0055] When data collection is normal, current and historical soil parameters and anchor mechanical parameter data are input into the potential analysis model to predict the prestress value at each monitoring point of the anchor, analyze the range adjustment coefficient, and construct the prestress prediction confidence interval. By comparing the real-time monitoring prestress data with the prediction confidence interval, potential abnormalities can be identified.

[0056] When there is a potential abnormal situation, the real-time and last monitored soil data, anchor parameters, predicted confidence intervals and measured anchor prestress data are input into the slope analysis model to obtain the abnormal risk value. The degree of abnormality is evaluated based on the size of the abnormal risk value, and corresponding early warning measures are implemented.

[0057] (3) Beneficial effects

[0058] The present invention provides a slope anchor prestressing monitoring system and method, which has the following beneficial effects:

[0059] 1. The present invention records a slope anchor prestress monitoring system and method, which can collect the prestress at different positions of the slope anchor, then evaluate the collected prestress data at different positions of the slope anchor, study its attenuation, and further analyze whether there are any abnormalities in the collected data. It can promptly detect whether there are any abnormalities in the data, thereby effectively avoiding inaccurate analysis results caused by abnormal collected data, the need for repeated collection and analysis, and the resulting low efficiency and high cost. The system has good use effect and has a good prospect for use.

[0060] 2. The present invention records a slope anchor prestress monitoring system and method, which, when in use, obtains external soil change data, performs analysis based on external soil changes and historical anchor prestress data, studies the anchor prestress change law and the influence of the external environment on the anchor prestress change, thereby effectively realizing the effective prediction of anchor prestress, providing a basis for anchor prestress anomaly analysis, making anchor prestress analysis more accurate and having better analysis effect, and having good application prospects.

[0061] 3. The present invention records a slope anchor prestress monitoring system and method. When an abnormality occurs in the anchor prestress, the system combines the real-time and last monitored soil data, anchor parameters, predicted confidence intervals and measured anchor prestress data for analysis, and predicts the slope abnormality risk value caused by the change in anchor prestress and soil data. It can serve as basic data for slope safety early warning, making it convenient for relevant technical personnel to understand the slope risk level and for corresponding management personnel to adjust the corresponding protection strategy in time. It has good use effect and good prospects for use. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of a slope anchor prestress monitoring system according to the present invention;

[0063] Figure 2 This is a partial layout diagram of the slope anchor monitoring points of the present invention. DETAILED DESCRIPTION

[0064] 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.

[0065] Since the central idea of ​​the solution recorded in the existing patent is to monitor the prestress data of the anchor rods in the slope and the corresponding layout data of the anchor rods, and process the prestress data of the anchor rods in the slope and the corresponding layout data of the anchor rods using the corresponding anchor rod average shear stress prediction model, the stress data collected in real time predicts the current stress bearing condition of the anchor rod, and further realizes that the stress bearing condition of the anchor rod exceeds the preset value.

[0066] However, in actual use, this solution has the following limitations:

[0067] Single applicability: This method is suitable for areas with balanced and stable soil. However, for slightly more complex areas, the prestressing force at each anchor position varies greatly. In addition, the soil is affected by the external environment, which will produce different changes. The degree of change reflected in the anchor is completely different. The average method and the fixed model used for analysis will average the abnormal data, which is difficult to reflect the degree of local slope anomaly. By the time it reflects the slope anomaly, the slope anomaly in some areas is often more serious, and the cost of repair is high.

[0068] Low accuracy: This method analyzes stress changes during use, and stress changes are difficult to reflect the situation between the anchor rod and the slope, so the accuracy of the analysis is not high.

[0069] This invention constructs a new technical framework for slope prestressing monitoring and slope anomaly assessment through multiple verifications of collected data, dynamic comparisons in time and space dimensions, and multiple innovations in comprehensive analysis of slope anchors. Its core lies not in the improvement of a single technology, but in the systematic solution of the inherent defects of traditional methods through the organic collaboration of cross-domain technologies, providing reusable technical specifications for the intelligent analysis of slope anomaly assessment, and laying the technical foundation for providing anchor and slope anomaly analysis.

[0070] Example 1

[0071] See also Figure 1 This embodiment provides a rehabilitation training evaluation system for schizophrenia patients. The specific scheme is as follows:

[0072] 1. System Architecture

[0073] 1.1 Data Collection

[0074] Data acquisition is based on the data acquisition module: it is used to collect slope anchor multi-point stress data, slope soil data and slope environment data, and pre-process the collected data.

[0075] Multi-point stress data of slope anchors are collected using fiber Bragg grating stress sensors and fiber demodulators installed at N positions along the length of each slope anchor. Slope soil data include slope soil density and moisture content, slope soil particle gradation and porosity, slope soil cohesion and internal friction angle, slope soil compression modulus and deformation modulus, foundation bearing capacity, and slope soil permeability coefficient. Slope environmental data include slope gradient and height, slope precipitation data, and slope temperature and humidity data.

[0076] The density of slope soil reflects the overall quality of soil particles and pores, and affects the distribution of self-weight stress. The greater the density, the greater the self-stabilizing pressure of the slope, which may increase the anchor load demand.

[0077] The increase in slope soil moisture content will reduce soil cohesion and internal friction angle (especially for clay soil), causing soil softening and increased deadweight, which can easily cause sliding and may cause anchor prestress loss or failure.

[0078] Slope soil particle size: Coarse-grained soil (such as gravel and sand) has strong permeability and large pores, and requires higher prestressing force to resist sliding when anchoring the anchor rod; fine-grained soil (such as clay and silt) has low permeability and is easily softened by water accumulation, which may lead to a decrease in the adhesion of the soil around the anchor rod.

[0079] The larger the porosity, the higher the soil compressibility. The anchor prestress may decay due to soil compression, affecting long-term stability.

[0080] The cohesion of slope soil reflects the bonding ability between soil particles. The lower the cohesion, the easier it is for the soil to slide along the shear surface, and higher anchor prestressing is required to compensate for the insufficient anti-slip force.

[0081] The internal friction angle determines the frictional resistance of the soil. The smaller the internal friction angle, the weaker the soil's anti-slip ability, and the anchor rod needs to provide greater tension to maintain balance.

[0082] When the compressibility of the slope soil is high, after the anchor rod is prestressed, the soil compression may cause the anchor rod to move or the prestress to relax, causing the slope to become unstable.

[0083] The deformation modulus of slope soil reflects the ability of soil to resist deformation. Soil with low modulus is prone to large deformation under the action of anchor load, which affects the anchoring effect.

[0084] If the bearing capacity of the foundation is insufficient, the anchor section of the anchor rod may fail due to shear damage of the surrounding soil, resulting in the inability to effectively transfer the prestress and causing slope accidents.

[0085] Soil with a high permeability coefficient on the slope is easily affected by the penetration of groundwater, resulting in phenomena such as piping and sand flow, which weaken the soil strength and may lead to hollowing of the soil around the anchor or anchor failure.

[0086] In actual use, the expansion rate of slope soil, soil self-weight stress and lateral pressure, as well as soil structure and crack development degree can also be collected.

[0087] In actual use, fiber Bragg grating stress sensors and fiber demodulators can also be used to install vibrating wire stress gauges, strain gauges and other monitoring instruments at key points of slope anchors, and connect to data acquisition terminals through wired (such as RS485 bus) or wireless (such as LoRa, NB-IoT) communication methods. LoRa is a long-distance radio technology, and NB-IoT is a narrowband Internet of Things technology based on cellular networks.

[0088] The specific stress monitoring components are as follows:

[0089] Vibrating wire strain gauges have strong anti-interference capabilities and are suitable for long-term monitoring. They output stress values ​​through frequency signals. During installation, the strain gauge is welded to the anchor steel bar or connected through a steel sleeve to ensure coaxial force. The sensor wire is led out along the anchor rod, and the connection is wrapped with waterproof tape to avoid damage during grouting.

[0090] Resistance strain gauges have high accuracy and fast response, but require shielded cables and temperature compensation and are suitable for short-term or dynamic monitoring.

[0091] Distributed monitoring of fiber optic Bragg grating stress sensors can obtain multi-point data simultaneously, resist electromagnetic interference, and is suitable for complex environments. During installation, the sensor is adhered to the surface of the anchor rod with special glue, or fixed with a stainless steel clamp. Redundant length must be reserved for the optical fiber line to avoid breakage during tensioning or deformation, and anti-wear protection must be provided (such as insertion into a PVC pipe). When in use, stress is calculated based on wavelength drift, taking into account the cross-sensitivity characteristics of temperature and strain.

[0092] The data acquisition instruments corresponding to the stress monitoring elements are as follows:

[0093] The vibrating string collector is equipped with a vibrating string sensor, which supports manual or automatic frequency data collection;

[0094] The static strain collector is connected to the resistance strain gauge to record the strain value in real time and convert it into stress;

[0095] Fiber optic interrogators are used in fiber grating sensors and can synchronously collect multi-channel wavelength signals.

[0096] The data acquisition terminal receives the raw data transmitted by the sensor (such as stress value, temperature value, timestamp, etc.) in real time, and transmits the data to the cloud server or local server through the edge computing node or gateway.

[0097] The temperature and humidity data in the slope environmental data are collected synchronously by temperature and humidity sensors.

[0098] The slope soil data is collected by using the soil moisture sensor and the pore water pressure gauge.

[0099] When deploying sensors, place them 0.5 to 1 m above the potential slip surface to ensure they are located in the effective force-bearing area.

[0100] Each anchor rod is equipped with 2-4 monitoring points along its length (such as the hole mouth, the middle, and the end of the anchoring section), and the spacing is adjusted according to the length of the anchor rod.

[0101] The sensor can be installed in advance by welding or tying the sensor to the steel bar during anchor construction and then embedding it into the drilled hole along with the anchor.

[0102] When wiring, wired sensors use shielded twisted pair cables and set up repeaters every 100m; wireless sensors deploy gateways (such as LoRaWAN gateways) with a coverage radius of ≥1km.

[0103] 1.2 Data Transmission

[0104] Edge Layer

[0105] Data acquisition terminal: uses industrial-grade RTU (remote terminal unit) and supports multi-protocol conversion (Modbus, CAN bus).

[0106] Edge computing node: equipped with lightweight algorithms to filter invalid data in real time (such as mutation points where stress values ​​exceed the range by ±20%).

[0107] Transport layer

[0108] Wireless transmission: 4G / 5G network (applicable to urban areas) or satellite communication (applicable to remote mountainous areas), data is encrypted and uploaded to the cloud (HTTPS protocol).

[0109] Wired transmission: Optical fiber direct connection (bandwidth ≥ 10Mbps), used for close connection between data center and monitoring site.

[0110] 1.3 Data Preprocessing

[0111] The steps for preprocessing the collected data are as follows:

[0112] Statistical summary: Use statistical techniques to comprehensively sort out the collected data, use a decision tree classification algorithm (based on data type, monitoring point location, and data characteristics) to classify the data, and use data aggregation functions (SUM, COUNT, AVG, etc.) to perform summary operations and generate a structured data summary table, thus achieving preliminary regularization of the original data and laying the foundation for subsequent analysis;

[0113] The decision tree classification algorithm can also use the K-nearest neighbor classification algorithm.

[0114] Data testing: Audit data using data quality assessment indicators (completeness, consistency, accuracy, etc.). Apply descriptive statistical methods (such as mean, median, and standard deviation calculations) and data visualization techniques (such as box plots and histograms) to check for missing values ​​and duplicate records. Missing data usually manifests as empty field values ​​or specific identifiers. Duplicate data refers to multiple records that are identical in key attributes.

[0115] Missing data processing: If missing data is detected, the data collection process will be initiated according to the missing data mechanism (such as completely random missing, random missing, non-random missing) to collect the missing data and ensure the integrity of the missing data;

[0116] Duplicate processing: If data duplication is detected, a data hash deduplication algorithm is used to locate and delete the duplicate records originally collected. At the same time, the data recollection process is initiated according to the data duplication mechanism to collect duplicate data to ensure the uniqueness and accuracy of the data.

[0117] The hash deduplication algorithm can also be replaced by a deduplication algorithm based on similarity calculation. Once data duplication is detected, the source of the data duplication can be traced back, which may be caused by reasons such as collection equipment failure and system interface abnormality.

[0118] The data preprocessing tools use SQL and Python. SQL can achieve batch deduplication and unique identifier management of large-scale time series data, while Python can achieve complex interpolation, model prediction, and batch processing.

[0119] 2. Collect data and analyze it

[0120] The data collection analysis is mainly based on the collection and analysis module: the real-time collection data is dynamically compared with the historical data using time series analysis technology to calculate the anchor prestress attenuation. The difference ratio between the two is analyzed by combining the last monitoring anchor prestress attenuation data at the same monitoring point. The data collection is judged to be abnormal by referring to the pre-set standard threshold range. If there is any data collection abnormality, the data re-collection mechanism is triggered.

[0121] Anchor prestress monitoring points such as Figure 2 shown.

[0122] The collected data includes the collection time (timestamp accurate to seconds), anchor detection position (code that uniquely identifies the anchor), and prestress value (raw data measured by the sensor).

[0123] Before dynamic data comparison, real-time data and historical data are aligned by timestamp to ensure that the time dimension of the dynamic comparison is consistent.

[0124] Dynamic comparison is to compare the slope anchor prestress collected in real time with the slope anchor prestress collected at the corresponding position last time, and calculate the anchor prestress attenuation Psj by the formula Psj = Pls-Pss, where Pls is the slope anchor prestress collected at a certain position last time, and Pss is the slope anchor prestress collected at a certain position in real time, and the N groups of anchor prestress attenuation are obtained;

[0125] For example, the test is performed once a month, that is, the data collected this time is compared with the data collected last month. Except for the time stamp interval of one month, the data at the same anchor rod test position are compared, and the prestress value is compared during the comparison.

[0126] Determine the positive and negative values ​​of the prestress attenuation of N groups of anchor rods. If the prestress attenuation of N groups of anchor rods is negative, it is determined that there is an abnormality in data collection (under normal circumstances, the prestress of anchor rods decays. Therefore, when the prestress attenuation of anchor rods is negative, it means that the prestress increases, which is obviously abnormal). If the prestress attenuation of N groups of anchor rods is positive, it is marked as normal.

[0127] Get the normal anchor prestress attenuation amount corresponding to the last anchor prestress attenuation amount Psjs at the monitoring position;

[0128] Under normal circumstances, the fluctuation of anchor prestress attenuation at the same position has a certain range and will not change suddenly. However, the attenuation at different anchor positions is different, so it cannot be processed evenly. It is necessary to analyze and process each monitoring point separately. Therefore, compare the last anchor prestress attenuation Psjs.

[0129] Calculate the difference between the normal anchor prestress attenuation and the previous anchor prestress attenuation at the corresponding monitoring position.

[0130] Since the difference in anchor prestress attenuation at each position is different, the difference ratio is more effective than the difference.

[0131] Compare and analyze the difference ratio P with the standard threshold interval. If P is within the standard threshold interval, it is determined that the data collection is normal and no processing is performed;

[0132] The standard threshold interval is [0, 5%], that is, when the difference ratio P is less than or equal to 5%, the data collection is judged to be normal and no processing is performed;

[0133] When the difference ratio P is greater than 5%, the data collection is judged to be abnormal.

[0134] If P is outside the standard threshold range, the data collection is judged to be abnormal, and the data re-collection mechanism is triggered to collect K consecutive data of the anchor prestress at this location, and obtain the maximum, minimum and average values ​​of the data. If the difference between the maximum and minimum values ​​is less than 3% of the average value, the average value is used as the anchor prestress at this location, and the data collection abnormality at this location is eliminated;

[0135] If the difference between the maximum and minimum values ​​is greater than or equal to 3% of the average value, it is marked as an abnormal supervision state and an abnormal supervision warning signal is issued.

[0136] This situation indicates that the data collected at this monitoring point is highly volatile and inaccurate, most of which is caused by sensor failure.

[0137] The present invention records a slope anchor prestress monitoring system and method, which can collect the prestress at different positions of the slope anchor, then evaluate the collected prestress data at different positions of the slope anchor, study its attenuation, and further analyze whether there are any abnormalities in the collected data. It can promptly discover whether there are any abnormalities in the data, thereby effectively avoiding inaccurate analysis results caused by abnormal collected data, the need for repeated collection and analysis, and the resulting low efficiency and high cost. The system has good use effect and has a good prospect for use.

[0138] 3. Potential anomaly analysis

[0139] Potential anomaly analysis is based on the secondary analysis module: When data collection is normal, current and historical soil parameters and anchor mechanical parameter data are input into the potential analysis model to predict the prestress value at each anchor monitoring point, analyze the range adjustment coefficient, and construct the prestress prediction confidence interval. By comparing the real-time monitoring prestress data with the prediction confidence interval, potential anomalies can be identified.

[0140] The potential analysis model is a hybrid model based on convolutional neural networks and long short-term memory networks. The combination of LSTM and CNN in the hybrid model based on convolutional neural networks and long short-term memory networks adopts a parallel branch-feature fusion architecture.

[0141] The steps for predicting the prestress value of each monitoring point of the anchor bolt using the potential analysis model are as follows:

[0142] Extract spatial feature vectors based on convolutional neural network branches;

[0143] The specific steps are temporal and spatial feature separation: for each time, extract the spatial matrix;

[0144] Convolution operation: Apply a 3×3 convolution kernel to extract local spatial features, such as the stress gradient of adjacent anchors;

[0145] Spatiotemporal feature reconstruction: Through global average pooling, the spatial dimension is compressed into a feature vector, and then the feature vector is output;

[0146] Extracting temporal feature vectors based on long short-term memory network branches;

[0147] The specific steps are time series feature extraction, long-term and short-term dependency modeling, and obtaining time feature vectors.

[0148] The spatial feature vector and the temporal feature vector are subjected to feature splicing, fusion and dimensionality reduction processing to form a spatiotemporal feature vector;

[0149] The specific steps are to concatenate the spatial feature vector and the temporal feature vector, and then perform nonlinear mapping (feature fusion and dimensionality reduction through the fully connected layer).

[0150] Feature preprocessing, Dropout regularization and multi-layer perceptron mapping are performed based on the spatiotemporal feature vector to obtain the prestress prediction value γ of N points.

[0151] The steps to analyze the range adjustment factor are as follows:

[0152] Obtain the most recent P group of historical slope anchor prestress prediction data and actual monitoring data at the monitoring point;

[0153] For example, if data is collected once a month, the slope anchor prestress prediction data for the previous P months will be obtained.

[0154] Calculate the difference ratio between the historical slope anchor prestress prediction data and the actual monitoring data of the monitoring point group P to obtain the difference ratio data set;

[0155] The calculation is performed each time the data is collected, so the difference ratio can be directly retrieved from the database without the need for real-time calculation.

[0156] The exponential smoothing method in time series analysis is used to process the data in the difference ratio dataset, and the historical difference ratio data are used to predict the difference ratio at the current moment, and the predicted difference ratio is used as the range adjustment coefficient ε.

[0157] The exponential smoothing method uses a simple exponential smoothing model when processing data in the difference ratio data set:

[0158] The prediction formula is as follows:

[0159] ε=d×ε p +d×(1-d)×ε p-1 +d×(1-d) 2 ×ε p-2+…+d×(1-d) p-1 ×ε p-(p-1)

[0160] Where d is the smoothing coefficient, which is determined by minimizing the MSE. The data in the difference ratio dataset is: (ε p , ε p-1 , ε p-2 …ε p-(p-1) );

[0161] Based on the range adjustment coefficient ε of each monitoring point, the prestress prediction confidence interval [γ-γ×ε, γ+γ×ε] of each monitoring point is calculated. If the real-time monitoring prestress data Pss of the monitoring point is within [γ-γ×ε, γ+γ×ε], the monitoring point is judged to be normal. This situation indicates that the attenuation rate of the anchor rod prestress is within the prediction range, which belongs to normal prestress attenuation; if the real-time monitoring prestress data Pss of the monitoring point is outside [γ-γ×ε, γ+γ×ε], the monitoring point is judged to be potentially abnormal. This situation indicates that the attenuation rate of the anchor rod prestress is too fast, indicating that its attenuation is abnormal.

[0162] The present invention records a slope anchor prestress monitoring system and method, which, when in use, obtains external soil change data, performs analysis based on external soil changes and historical anchor prestress data, studies the anchor prestress change law and the influence of the external environment on the anchor prestress change, thereby effectively realizing the effective prediction of anchor prestress, providing a basis for anchor prestress anomaly analysis, making anchor prestress analysis more accurate and having better analysis effect, and having good application prospects.

[0163] 4. Abnormal analysis

[0164] When an anomaly exists, it is necessary to further analyze the anomaly. The analysis of the anomaly mainly depends on the anomaly analysis module. The anomaly analysis module: when there is a potential abnormal situation, the real-time and last monitored soil data, anchor parameters, predicted confidence intervals and measured anchor prestress data are input into the slope analysis model to obtain the abnormal risk value, and the degree of abnormality is evaluated according to the size of the abnormal risk value, and corresponding early warning measures are implemented.

[0165] The slope analysis model uses a trained convolutional neural network. The training steps of the convolutional neural network are as follows:

[0166] Data collection and annotation: retrieve historical data and mark several groups of data time points with potential anomalies in the historical data;

[0167] Obtain soil data, prestress prediction interval, measured anchor prestress data, and abnormality degree data at the data time point of potential abnormal conditions. Use data mining technology to further extract the soil data monitored last time at the data time point of the abnormal conditions. Then, normalize and integrate the extracted data into an abnormality data package. The abnormality degree data includes the characteristic data reflecting the abnormality degree and its corresponding abnormality risk value.

[0168] Aggregate several groups of abnormal data packets to form an abnormal data set, and divide the abnormal data set into a training set and a validation set with a division ratio of 7:3;

[0169] Construct a convolutional neural network to perform regression tasks. Use the data in the training set except the abnormal risk value as the input of the convolutional neural network, and use the label of the characteristic data reflecting the degree of abnormality as the output. Train the convolutional neural network to obtain the initial convolutional neural network.

[0170] The output calculation formula of neurons in convolutional neural networks is h l =f(W l ×h l-1 +b l ), where h l is the output of the l-th layer of neurons, W l is the weight matrix of the lth layer, h l-1 is the output of the l-1 layer neurons, b l is the bias vector of the lth layer, and f is the activation function.

[0171] The mean square error loss function is used during training.

[0172] The gradient of the loss function with respect to the network parameters (weight W and bias b) is calculated through the back-propagation algorithm, and the network parameters are updated using an optimizer (such as stochastic gradient descent SGD, Adam, etc.). The training is iterated continuously until the loss function converges to obtain the initial convolutional neural network.

[0173] The initial convolutional neural network is model verified using the validation set, and the mean square error evaluation index is used to output the initial convolutional neural network with a test accuracy greater than or equal to the preset threshold as the trained convolutional neural network.

[0174] The formula of the mean square error loss function is as follows:

[0175]

[0176] Where MSE is mean square error, N is the number of samples, and yb i The i-th true label value, y i Predict the label value for the i-th model.

[0177] The evaluation of abnormality degree based on the abnormal risk value includes:

[0178] Set the abnormality threshold, which includes YC1 and YC2, and YC1>YC2>0;

[0179] Compare the abnormal risk value with the abnormality degree threshold;

[0180] If the abnormal risk value is greater than or equal to the abnormality threshold YC1, the slope abnormality is determined to be severe, the monitoring point is marked, the corresponding abnormality level light prompt (for example, red) flashes, and an abnormality report and abnormal area blockade report are sent through the emergency management platform;

[0181] If the abnormal risk value is less than the abnormality threshold YC1 and greater than or equal to the abnormality threshold YC2, the slope abnormality is determined to be moderate, and the monitoring point is marked. The corresponding abnormality level light prompt (for example, yellow) flashes, and a manual inspection work order is sent to the operation and maintenance platform for regular manual verification. A spatiotemporal heat map is generated, and the abnormality diffusion trend is marked.

[0182] If the abnormal risk value is less than the abnormality threshold YC2, the slope abnormality is judged to be mild. The data is verified. After verification, the monitoring period is shortened, the monitoring point is marked, and the light prompt corresponding to the abnormality level (for example, green) is flashed. At the same time, the data is included in the next round of convolutional neural network as a training sample.

[0183] The present invention records a slope anchor prestress monitoring system and method. When an abnormality occurs in the anchor prestress, the system combines real-time and last monitored soil data, anchor parameters, predicted confidence intervals and measured anchor prestress data for analysis, and predicts the slope abnormality risk value caused by changes in anchor prestress and soil data. The system can serve as basic data for slope safety early warning, making it convenient for relevant technical personnel to understand the slope risk level and for corresponding management personnel to adjust the corresponding protection strategy in a timely manner. The system has good use effect and has a good prospect for use.

[0184] The weight coefficient involved in the above formula is determined by the coefficient of variation method. The coefficient of variation method is a method of weighting each indicator according to the degree of variation between the current value of each evaluation indicator and the target value. If the numerical difference of a certain indicator is large, it can clearly distinguish the evaluated objects, indicating that the indicator has rich discrimination information, and thus the indicator should be given a larger weight. On the contrary, if the numerical difference of each evaluated object on a certain indicator is small, then the ability of this indicator to distinguish the evaluation objects is weak, and thus the indicator should be given a smaller weight. This method directly uses the information contained in each indicator to obtain the weight of the indicator through calculation, and therefore is objective.

[0185] Example 2

[0186] A slope anchor prestressing monitoring method comprises the following steps:

[0187] Collect slope anchor multi-point stress data, slope soil data, and slope environment data, and pre-process the collected data;

[0188] Use time series analysis technology to dynamically compare real-time collected data with historical data to calculate the anchor prestress attenuation. Combined with the previous monitoring anchor prestress attenuation data at the same monitoring point, analyze the difference ratio between the two, and refer to the pre-set standard threshold range to determine whether the data collection is abnormal. If there is a data collection abnormality, the data re-collection mechanism is triggered;

[0189] When data collection is normal, current and historical soil parameters and anchor mechanical parameter data are input into the potential analysis model to predict the prestress value at each monitoring point of the anchor, analyze the range adjustment coefficient, and construct the prestress prediction confidence interval. By comparing the real-time monitoring prestress data with the prediction confidence interval, potential abnormalities can be identified.

[0190] When there is a potential abnormal situation, the real-time and last monitored soil data, anchor parameters, predicted confidence intervals and measured anchor prestress data are input into the slope analysis model to obtain the abnormal risk value. The degree of abnormality is evaluated based on the size of the abnormal risk value, and corresponding early warning measures are implemented.

[0191] In this application, the several formulas involved are all calculated by taking their numerical values ​​after removing the dimensions, and the formulas are established by collecting a large amount of data and performing software simulation to obtain a formula for the most recent real situation. Some coefficients or weights in the formulas are set by technical personnel in this field according to actual conditions, so they will not be elaborated here.

[0192] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of the various examples 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.

[0193] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0194] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A slope anchor prestress monitoring system, characterized by: include: Data acquisition module: used to collect slope anchor multi-point stress data, slope soil data and slope environment data, and pre-process the collected data; Acquisition and Analysis Module: Uses time series analysis technology to dynamically compare real-time collected data with historical data, calculates anchor prestress attenuation, and combines it with the previous monitoring anchor prestress attenuation data at the same monitoring point to analyze the difference ratio between the two. Based on the pre-set standard threshold range, it determines whether data collection is abnormal. If there is a data collection abnormality, the data re-acquisition mechanism is triggered; Secondary analysis module: When data collection is normal, current and historical soil parameters and anchor mechanical parameter data are input into the potential analysis model to predict the prestress value at each monitoring point of the anchor, analyze the range adjustment coefficient, and construct the prestress prediction confidence interval. By comparing the real-time monitoring prestress data with the predicted confidence interval, potential abnormalities can be identified; Abnormal analysis module: When there is a potential abnormal situation, the real-time and last monitored soil data, anchor parameters, predicted confidence intervals and measured anchor prestress data are input into the slope analysis model to obtain the abnormal risk value. The abnormality degree is evaluated based on the size of the abnormal risk value, and corresponding early warning measures are implemented.

2. The slope anchor prestressing monitoring system according to claim 1, characterized in that: Multi-point stress data of slope anchors are collected using fiber Bragg grating stress sensors and fiber demodulators installed at N positions along the length of each slope anchor. Slope soil data include slope soil density and moisture content, slope soil particle gradation and porosity, slope soil cohesion and internal friction angle, slope soil compression modulus and deformation modulus, foundation bearing capacity, and slope soil permeability coefficient. Slope environmental data include slope gradient and height, slope precipitation data, and slope temperature and humidity data.

3. The slope anchor prestress monitoring system according to claim 2, characterized in that: The steps for preprocessing the collected data are as follows: Statistical summary: Use data statistical techniques to comprehensively sort out the collected data, use decision tree classification algorithms to classify the data, use data aggregation functions to perform summary operations, and generate structured data summary tables; Data testing: Use data quality assessment indicators to audit data and check whether there are missing values ​​and duplicate records in the data; Missing data processing: If missing data is detected, the data collection process will be initiated according to the missing data mechanism to collect the missing data; Duplicate processing: If data duplication is detected, a data hash deduplication algorithm is used to locate and delete the duplicate records originally collected. At the same time, the data recollection process is initiated according to the data duplication mechanism to collect duplicate data.

4. The slope anchor prestressing monitoring system according to claim 3, characterized in that: Dynamic comparison is to compare the slope anchor prestress collected in real time with the slope anchor prestress collected at the corresponding position last time, and calculate the anchor prestress attenuation Psj by the formula Psj = Pls-Pss, where Pls is the slope anchor prestress collected at a certain position last time, and Pss is the slope anchor prestress collected at a certain position in real time, and the N groups of anchor prestress attenuation are obtained; Determine the positive and negative values ​​of the prestress attenuation of N groups of anchor rods. If the prestress attenuation of N groups of anchor rods is negative, it is determined that there is an abnormality in data collection. If the prestress attenuation of N groups of anchor rods is positive, it is marked as normal. Get the normal anchor prestress attenuation amount corresponding to the last anchor prestress attenuation amount Psjs at the monitoring position; Calculate the difference ratio between the normal anchor prestress attenuation and the previous anchor prestress attenuation at the corresponding monitoring position Compare and analyze the difference ratio P with the standard threshold interval. If P is within the standard threshold interval, it is determined that the data collection is normal and no processing is performed; If P is outside the standard threshold range, the data collection is judged to be abnormal, and the data re-collection mechanism is triggered to collect K consecutive data of the anchor prestress at this location, and obtain the maximum, minimum and average values ​​of the data. If the difference between the maximum and minimum values ​​is less than 3% of the average value, the average value is used as the anchor prestress at this location, and the data collection abnormality at this location is eliminated; If the difference between the maximum and minimum values ​​is greater than or equal to 3% of the average value, it is marked as an abnormal supervision state and an abnormal supervision warning signal is issued.

5. The slope anchor prestress monitoring system according to claim 4, characterized in that: The potential analysis model is a hybrid model based on convolutional neural networks and long short-term memory networks. The steps for predicting the prestress value of each monitoring point of the anchor rod using the potential analysis model are as follows: The spatial feature vector is extracted based on the convolutional neural network branch, and the temporal feature vector is extracted based on the long short-term memory network branch; The spatial feature vector and the temporal feature vector are subjected to feature splicing, fusion and dimensionality reduction processing to form a spatiotemporal feature vector; Feature preprocessing, Dropout regularization and multi-layer perceptron mapping are performed based on the spatiotemporal feature vector to obtain the prestress prediction value γ of N points.

6. The slope anchor prestress monitoring system according to claim 5, characterized in that: The steps to analyze the range adjustment factor are as follows: Obtain the most recent P group of historical slope anchor prestress prediction data and actual monitoring data at the monitoring point; Calculate the difference ratio between the historical slope anchor prestress prediction data and the actual monitoring data of the monitoring point group P to obtain the difference ratio data set; The exponential smoothing method in time series analysis is used to process the data in the difference ratio dataset, and the historical difference ratio data are used to predict the difference ratio at the current moment, and the predicted difference ratio is used as the range adjustment coefficient ε.

7. The slope anchor prestress monitoring system according to claim 6, characterized in that: Based on the range adjustment coefficient ε of each monitoring point, the prestress prediction confidence interval [γ-γ×ε, γ+γ×ε] of each monitoring point is calculated. If the real-time monitoring prestress data Pss of the monitoring point is within [γ-γ×ε, γ+γ×ε], the monitoring point is judged to be normal. If the real-time monitoring prestress data Pss of the monitoring point is outside [γ-γ×ε, γ+γ×ε], the monitoring point is judged to be potentially abnormal.

8. The slope anchor prestress monitoring system according to claim 7, characterized in that: The slope analysis model uses a trained convolutional neural network. The training steps of the convolutional neural network are as follows: Data collection and annotation: retrieve historical data and mark several groups of data time points with potential anomalies in the historical data; Obtain soil data, prestress prediction interval, measured anchor prestress data, and abnormality degree data at the data time point of potential abnormal conditions. Use data mining technology to further extract the soil data monitored last time at the data time point of the abnormal conditions. Then, normalize and integrate the extracted data into an abnormality data package. The abnormality degree data includes the characteristic data reflecting the abnormality degree and its corresponding abnormality risk value. Aggregate several groups of abnormal data packets to form an abnormal data set, and divide the abnormal data set into a training set and a validation set; Construct a convolutional neural network to perform regression tasks. Use the data in the training set except the abnormal risk value as the input of the convolutional neural network, and use the label of the characteristic data reflecting the degree of abnormality as the output. Train the convolutional neural network to obtain the initial convolutional neural network. The initial convolutional neural network is model verified using the validation set, and the mean square error evaluation index is used to output the initial convolutional neural network with a test accuracy greater than or equal to the preset threshold as the trained convolutional neural network.

9. The slope anchor prestressing monitoring system according to claim 8, characterized in that: The evaluation of abnormality degree based on the abnormal risk value includes: Set the abnormality threshold, which includes YC1 and YC2, and YC1>YC2>0; Compare the abnormal risk value with the abnormality degree threshold; If the abnormal risk value is greater than or equal to the abnormality threshold YC1, the slope abnormality is determined to be severe, the monitoring point is marked, the corresponding abnormality level light flashes as a prompt, and an abnormality report and abnormal area blockade report are sent through the emergency management platform; If the abnormal risk value is less than the abnormality threshold YC1 and greater than or equal to the abnormality threshold YC2, the slope abnormality is determined to be moderate, and the monitoring point is marked. A light corresponding to the abnormality level flashes as a prompt. A manual inspection work order is sent to the operation and maintenance platform for regular manual verification. A spatiotemporal heat map is generated, and the abnormality diffusion trend is marked. If the abnormal risk value is less than the abnormality threshold YC2, the slope abnormality is judged to be mild. The data is verified. After verification, the monitoring period is shortened, the monitoring point is marked, and the light corresponding to the abnormality level is flashed. At the same time, the data is included in the next round of convolutional neural network as a training sample.

10. A method for monitoring prestress of slope anchor rods, using the system according to any one of claims 1 to 9, characterized in that: The steps include: Collect slope anchor multi-point stress data, slope soil data, and slope environment data, and pre-process the collected data; Use time series analysis technology to dynamically compare real-time collected data with historical data to calculate the anchor prestress attenuation. Combined with the previous monitoring anchor prestress attenuation data at the same monitoring point, analyze the difference ratio between the two, and refer to the pre-set standard threshold range to determine whether the data collection is abnormal. If there is a data collection abnormality, the data re-collection mechanism is triggered; When data collection is normal, current and historical soil parameters and anchor mechanical parameter data are input into the potential analysis model to predict the prestress value at each monitoring point of the anchor, analyze the range adjustment coefficient, and construct the prestress prediction confidence interval. By comparing the real-time monitoring prestress data with the prediction confidence interval, potential abnormalities can be identified. When there is a potential abnormal situation, the real-time and last monitored soil data, anchor parameters, predicted confidence intervals and measured anchor prestress data are input into the slope analysis model to obtain the abnormal risk value. The degree of abnormality is evaluated based on the size of the abnormal risk value, and corresponding early warning measures are implemented.

Images