Early warning identification method for slope instability of extremely fractured rock mass in high-cold region

Abstract

The invention, which relates to the technical field of geological disaster monitoring and early warning, discloses an early warning identification method for slope instability of extremely fractured rock in an alpine region, comprising the following steps: collecting and preprocessing original monitoring data to generate a cleaned data set; performing multivariable regression analysis on the cleaned data set by using a neural network model to generate a quantitative coupling model; evaluating the slope instability risk according to the quantitative coupling model, and setting a dynamic early warning threshold in combination with the cleaned data set to generate a risk evaluation report; receiving the cleaned data set in real time according to the risk assessment report and an early warning threshold value, comparing the cleaned data set with the latest monitoring data, and when the latest monitoring data exceeds the early warning threshold value, triggering corresponding early warning and generating early warning information; and performing optimization processing on the latest monitoring data, reducing transmission load, generating processed monitoring data, performing early warning decision, and generating optimized early warning information. According to the invention, the accuracy and response capability of early warning identification are enhanced.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring and early warning technology, and in particular to an early warning and identification method for slope instability of extremely fractured rock masses in high-altitude and cold regions. Background Technology

[0002] With the increasing frequency of geological disasters, especially the slope instability problem of extremely fractured rock masses in high-altitude and cold regions, more and more research is focusing on slope stability monitoring and early warning technologies. Traditional slope monitoring techniques mainly rely on physical indicators such as displacement monitoring and stress-strain monitoring. Physical data of the rock mass is acquired in real time through sensors, and mathematical models are used for analysis and judgment to assess slope stability. Displacement monitoring can be achieved by installing ground displacement sensors to monitor minute changes in slope displacement and provide early warning of potential landslides and collapses.

[0003] Current technologies still have some shortcomings, especially in monitoring slope instability in highly fractured rock masses in high-altitude and cold regions, where they face severe challenges. Due to the special climatic conditions in high-altitude and cold regions (such as low temperatures and frequent freeze-thaw cycles), traditional sensors are easily affected by environmental factors, impacting the accuracy and continuity of monitoring data. Furthermore, a single physical parameter often cannot comprehensively reflect the slope instability risk, especially in highly fractured rock masses where the rock structure is complex and dynamically changes rapidly; relying solely on displacement or stress monitoring is insufficient to promptly identify potential instability risks. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an early warning and identification method for slope instability in extremely fractured rock masses in high-altitude and cold regions, which solves the problem of adjusting the early warning threshold in real time and improving the accuracy of early warning in complex environments.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions, comprising, Collect raw monitoring data and preprocess it to generate a cleaned dataset; Using a neural network model, multivariate regression analysis was performed on the cleaned dataset to generate a quantitative coupling model; The slope instability risk is assessed based on a quantitative coupling model, and a dynamic early warning threshold is set by combining the cleaned dataset to generate a risk assessment report. Based on the risk assessment report and early warning threshold, the cleaned dataset is received in real time and compared with the latest monitoring data. When the latest monitoring data exceeds the early warning threshold, the corresponding early warning is triggered and early warning information is generated. The latest monitoring data is optimized to reduce transmission load, generate processed monitoring data, and make early warning decisions to generate optimized early warning information.

[0007] As a preferred embodiment of the early warning and identification method for slope instability of extremely fractured rock mass in high-altitude cold regions described in this invention, the original monitoring data includes crack width, vibration energy, number of freeze-thaw cycles, slope inclination angle, temperature, humidity and rainfall.

[0008] As a preferred embodiment of the early warning and identification method for slope instability in highly fractured rock masses in high-altitude and cold regions described in this invention, the specific steps for generating the cleaned dataset are as follows: The raw monitoring data is formatted to generate a preliminary cleaned dataset; The initial cleaned dataset is subjected to noise removal and error correction to generate an accurate dataset. The accurate dataset is standardized and redundant features are removed to generate a cleaned dataset.

[0009] As a preferred embodiment of the early warning and identification method for slope instability in extremely fractured rock masses in high-altitude and cold regions described in this invention, the specific steps for generating the quantitative coupling model are as follows: Slope instability-related features were extracted from the cleaned dataset, and data transformation and redundancy removal were performed to generate a standardized feature dataset. The standardized feature dataset is input into the neural network model, trained by multivariate regression analysis, and the neural network weights are optimized using the backpropagation algorithm to generate the trained neural network model. By using the trained neural network model, predictions are made on a new standardized feature dataset and quantitative coupling relationships are established to generate a quantitative coupling model.

[0010] As a preferred embodiment of the early warning and identification method for slope instability in extremely fractured rock masses in high-altitude and cold regions described in this invention, the steps of inputting a standardized feature dataset into a neural network model, training it through multivariate regression analysis, optimizing the neural network weights using a backpropagation algorithm, and generating a trained neural network model are as follows. The standardized feature dataset is input into the neural network model, and the activation values ​​are calculated through forward propagation to generate preliminary prediction results; Based on the preliminary prediction results, the backpropagation algorithm is used to adjust the neural network weights to generate optimized neural network weights. Through iterative training and cross-validation, the neural network parameters and weights are continuously adjusted to generate a trained neural network model.

[0011] As a preferred embodiment of the early warning and identification method for slope instability in extremely fractured rock masses in high-altitude and cold regions described in this invention, the specific steps for generating the risk assessment report are as follows: Based on the quantitative coupling model, the cleaned dataset is analyzed to calculate the assessment value of slope instability risk and generate preliminary risk assessment results. Based on the crack width, temperature change, and vibration frequency characteristics in the cleaned dataset, an adaptive algorithm is used to calculate and set early warning thresholds related to the current slope condition. The preliminary risk assessment results are compared with the warning threshold to generate a risk assessment report.

[0012] As a preferred embodiment of the early warning and identification method for slope instability in extremely fractured rock masses in high-altitude and cold regions described in this invention, the specific steps for generating early warning information are as follows: Based on the risk assessment report and early warning thresholds, receive the latest monitoring data of the cleaned dataset in real time; The latest monitoring data is compared with the warning threshold to determine whether the warning threshold has been exceeded. If the latest monitoring data does not exceed the warning threshold, normal monitoring status shall be maintained; When the latest monitoring data exceeds the warning threshold, the corresponding warning will be automatically triggered and a warning message will be generated.

[0013] As a preferred embodiment of the early warning and identification method for slope instability in extremely fractured rock masses in high-altitude and cold regions described in this invention, the specific steps for generating and processing the monitoring data are as follows: The latest monitoring data is denoised, compressed, and quantized to generate an optimized dataset; The optimized dataset is converted into a standardized format to generate a standardized format dataset. Standardized format data is processed for transmission optimization to reduce transmission latency and improve transmission efficiency, generating processed monitoring data.

[0014] As a preferred embodiment of the early warning and identification method for slope instability in extremely fractured rock masses in high-altitude and cold regions described in this invention, the specific steps for denoising, compressing, and quantizing the latest monitoring data to generate an optimized dataset are as follows. Kalman filtering is used to filter noise from the original monitoring data, removing sensor errors and environmental interference noise, and generating a denoised dataset. A lossless compression method is used to compress the denoised dataset to generate a compressed dataset. By using a dynamic quantization algorithm, the compressed dataset is transformed into a format that is efficient for processing and storage, generating an optimized dataset.

[0015] As a preferred embodiment of the early warning and identification method for slope instability in extremely fractured rock masses in high-altitude and cold regions described in this invention, the specific steps for generating optimized early warning information are as follows: Analyze the processed monitoring data to identify potential risk patterns and generate a risk feature dataset; Based on the risk feature dataset, a quantitative coupling model is used to assess the slope instability risk and generate preliminary risk assessment results. By comparing and analyzing the preliminary risk assessment results and the early warning threshold, early warning judgment criteria are generated, and then compared with the processed monitoring data to generate preliminary early warning information. The initial warning information is verified, a warning verification report is generated, and the current warning level is assessed by combining the processed monitoring data to generate optimized warning information.

[0016] The beneficial effects of this invention are as follows: by using a neural network model to perform multivariate regression analysis on the cleaned dataset, a quantitative coupling model is generated, which realizes accurate modeling of the complex nonlinear relationship of slope instability, thereby improving the adaptability and accuracy of the quantitative coupling model, providing a reliable basis for setting dynamic early warning thresholds and risk assessment, and thus enhancing the accuracy and responsiveness of early warning identification. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart for an early warning and identification method for slope instability in extremely fractured rock masses in high-altitude and cold regions.

[0019] Figure 2 This is a flowchart of the generated dataset after cleaning.

[0020] Figure 3 A flowchart generated for a quantitative coupling model.

[0021] Figure 4 A flowchart for generating early warning information. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides an early warning and identification method for slope instability of extremely fractured rock mass in high-altitude and cold regions, including the following steps: S1. Collect raw monitoring data and preprocess it to generate a cleaned dataset.

[0026] S1.1 The original monitoring data includes crack width, vibration energy, number of freeze-thaw cycles, slope inclination angle, temperature, humidity and rainfall.

[0027] Specifically, real-time observation of the monitoring area records changes in crack width to reflect the propagation of rock mass cracks; vibration energy is analyzed to assess the dynamic response and stability of the slope; the number of freeze-thaw cycles is determined by recording temperature changes to determine the freeze-thaw frequency of the rock mass; the slope inclination angle is monitored by comparing time series data; temperature and humidity data are recorded based on the regularity of environmental changes to reflect the hydro-climatic characteristics of the slope; rainfall is recorded based on meteorological data to assess the impact of rainfall on slope stability. After standardization and processing, the data are compiled into a raw monitoring dataset. S1.2 Format the raw monitoring data to generate a preliminary cleaned dataset.

[0028] Specifically, each data point in the original monitoring data (such as crack width, vibration energy, freeze-thaw cycle count, slope inclination angle, temperature, humidity, and rainfall) is sorted by timestamp to ensure time series alignment. The values ​​in the original monitoring data are converted to standard units, for example, crack width is converted to standard millimeters, and vibration energy is converted to standard acceleration units. The original monitoring data is grouped according to different feature categories, ensuring that each group has a consistent structure; for example, temperature, humidity, and rainfall should belong to one category, while crack width, vibration energy, and freeze-thaw cycle count should belong to another. Missing values ​​are checked and filled using interpolation methods to ensure the data integrity for each feature. The processed data is then organized into a unified format to generate a preliminary cleaned dataset.

[0029] S1.3. Remove noise and correct errors from the initially cleaned dataset to generate an accurate dataset.

[0030] Specifically, outlier detection is performed on each feature in the initially cleaned dataset using the standard deviation method. Outliers exceeding the reasonable range for each feature are identified. For the detected outliers, regression analysis is used to correct them, replacing the outlier data with reasonable values. Kalman filtering is used to remove noise from the outlier data, eliminating the effects of sensor errors and environmental interference. The denoised data is then smoothed to ensure its continuity and consistency. Finally, the smoothed data is checked again to ensure there are no errors or omissions, generating an accurate dataset.

[0031] It should also be noted that the basis for the reasonable range is usually based on the distribution characteristics of the data and statistical principles. In the standard deviation method, the reasonable range is generally set based on the mean and standard deviation of the dataset. For example, assuming that the data conforms to a normal distribution, it is generally considered that the values ​​of the data points within the range of mean ± 2 times the standard deviation are normal, accounting for about 95% of the total data; while the values ​​within the range of mean ± 3 times the standard deviation are considered to be extremely common, accounting for about 99.7%.

[0032] S1.4 Standardize the accurate dataset and remove redundant features to generate a cleaned dataset.

[0033] Specifically, each feature in the precise dataset is standardized, typically using the Z-score standardization method to ensure that each feature has the same scale; correlation analysis is used to identify and remove redundant features, and variables that are highly correlated with other features are removed to reduce the dimensionality of the data and improve computational efficiency; the dataset after standardization and redundant feature removal is the cleaned dataset.

[0034] S2. Using a neural network model, perform multivariate regression analysis on the cleaned dataset to generate a quantitative coupling model.

[0035] S2.1 Extract slope instability-related features from the cleaned dataset, and perform data transformation and redundancy removal to generate a standardized feature dataset.

[0036] Specifically, based on previous analysis, features such as crack width, vibration energy, freeze-thaw cycle count, slope inclination angle, temperature, humidity, and rainfall were selected and data transformations were performed to ensure data consistency. For example, discrete data was converted into continuous data, or categorical data was encoded into numerical data. Correlation analysis was used to remove highly correlated or redundant features, retaining features with high information content. Z-score standardization was used to standardize each feature so that the mean of each feature is 0 and the standard deviation is 1, ensuring that all features are analyzed on the same scale and generating a standardized feature dataset.

[0037] S2.2 Input the standardized feature dataset into the neural network model, train it through multivariate regression analysis, optimize the neural network weights using the backpropagation algorithm, and generate the trained neural network model.

[0038] S2.2.1 Input the standardized feature dataset into the neural network model, calculate the activation values ​​through forward propagation, and generate preliminary prediction results.

[0039] Specifically, after inputting the standardized feature dataset into the neural network model, each standardized feature value is used as a node value in the input layer. Through the forward propagation process, the node values ​​of the input layer are passed to the hidden layers of the neural network in sequence. Each layer performs a nonlinear transformation on the input through activation functions (such as ReLU and Sigmoid). After inter-layer weighting and bias adjustment, the values ​​are passed to the output layer to generate preliminary prediction results. The output of each node is obtained by processing the mapping result of the output of the previous layer node in the parameter space of the current layer through activation functions.

[0040] S2.2.2 Based on the preliminary prediction results, the backpropagation algorithm is used to adjust the neural network weights to generate optimized neural network weights.

[0041] Specifically, based on the preliminary prediction results, the mean squared error is used to calculate the prediction error, and the error is backpropagated into the neural network. The weights of each layer are updated by calculating the error and the gradient of the output layer. Using the backpropagation algorithm, the error signal is propagated sequentially from the output layer to the input layer. The error gradient of each layer is calculated using the chain rule, and the weights of each node in the neural network are adjusted using the gradient descent method. Through multiple iterations, such as 1000 times, the weights of the neural network are gradually adjusted to an optimized state. In each iteration, the neural network updates the weights of each layer according to the calculated gradient, reducing the prediction error. After sufficient iterations, the weights of the neural network will converge near the local minimum or global minimum, generating the optimized neural network weights.

[0042] It should also be noted that the chain rule is used to calculate the derivative of a composite function, especially in the backpropagation algorithm. By calculating the gradient of each layer layer by layer, the chain rule helps to pass the error from the output layer to the input layer. The chain rule can effectively pass error information to the entire neural network, ensuring that the weights of each layer are correctly adjusted, thereby optimizing the learning process of the neural network and reducing prediction errors.

[0043] S2.2.3. Through iterative training and cross-validation, the neural network parameters and weights are continuously adjusted to generate a trained neural network model.

[0044] Specifically, through iterative training, the training dataset is divided into multiple subsets, and cross-validation is used, with a different subset used as the validation set each time, and the rest used as the training set. During each training process, the neural network parameters are updated based on the error value, and the network weights are adjusted using the gradient descent algorithm. During training, the error between each prediction result and the true value is calculated, and the weights of the neural network are adjusted to reduce the error. Through multiple iterations, the weight parameters of the neural network are gradually optimized, and cross-validation is used to evaluate the effect of each round of training to prevent overfitting and ensure the generalization ability of the neural network parameters. Through continuous training and cross-validation, a trained neural network model is generated.

[0045] S2.3. Using the trained neural network model, predict the new standardized feature dataset and establish quantitative coupling relationships to generate a quantitative coupling model.

[0046] Specifically, using the trained neural network model, a new standardized feature dataset is input into the neural network for forward propagation calculation. Through the calculation of each layer in the neural network, the prediction result for each sample is generated. The prediction result reflects the risk assessment value of slope instability. Based on the risk assessment value, a quantitative coupling relationship between the prediction result and each feature is established. Based on the prediction result and the quantitative coupling relationship, a quantitative coupling model is generated.

[0047] It should be noted that by performing multivariate regression analysis on the cleaned dataset using a neural network model, large-scale and multi-feature monitoring data can be effectively processed, and potential slope instability risk factors can be automatically identified. This not only improves the adaptability and accuracy in complex environments, but also automatically adjusts the parameters of the quantitative coupling model based on historical data, thereby maintaining high prediction accuracy in different geological environments and avoiding the reliance on human assumptions and simplified models in traditional methods.

[0048] S3. Assess the slope instability risk based on the quantitative coupling model, and combine the cleaned dataset to set a dynamic early warning threshold and generate a risk assessment report.

[0049] S3.1. Based on the quantitative coupling model, analyze the cleaned dataset, calculate the assessment value of slope instability risk, and generate preliminary risk assessment results.

[0050] Specifically, features related to slope instability, such as crack width, temperature change, and vibration frequency, are extracted from the cleaned dataset. These features are then input into a quantitative coupling model, which processes them to obtain a slope instability risk assessment value for each sample. Finally, the slope instability risk assessment values ​​are integrated to generate a preliminary risk assessment result.

[0051] It should also be noted that crack width, temperature variation, and vibration frequency characteristics are important parameters for analyzing slope instability risk. Crack width reflects the development of fissures in the rock mass or soil; a larger crack width usually indicates poor stability of the rock mass or soil, making it prone to instability. Temperature variation is closely related to the freeze-thaw cycle of the slope, and is particularly important in cold regions. Drastic temperature fluctuations may cause the rock mass to expand and contract, increasing the risk of instability. Vibration frequency characteristics reflect the vibration state of the rock mass or soil under external forces; changes in frequency may indicate crack propagation or stress concentration within the rock mass, and these changes may foreshadow the risk of slope instability. By analyzing crack width, temperature variation, and vibration frequency characteristics, slope stability can be effectively assessed and instability risk can be predicted.

[0052] S3.2 Based on the crack width, temperature change, and vibration frequency characteristics in the cleaned dataset, an adaptive algorithm is used to calculate and set an early warning threshold related to the current slope condition.

[0053] Specifically, based on the crack width, temperature change, and vibration frequency characteristics in the cleaned dataset, these characteristics are standardized. An adaptive algorithm is then used to calculate the warning threshold based on the current slope condition and the real-time changes in crack width, temperature change, and vibration frequency characteristics. The expression is as follows: ; in, Indicates the warning threshold. Indicates the adjustment factor. The weights assigned to crack width, temperature variation, and vibration frequency characteristics are represented. Standardized values ​​representing crack width, temperature variation, and vibration frequency characteristics. Indicates the bias term. An index representing crack width, temperature variation, and vibration frequency characteristics; The adaptive algorithm dynamically adjusts based on changes in each feature. and By calculating new warning thresholds, the real-time stability of the slope can be reflected, ensuring that the warning values ​​can be flexibly adjusted under different environmental conditions.

[0054] S3.3 Compare the preliminary risk assessment results with the early warning threshold to generate a risk assessment report.

[0055] Specifically, the preliminary risk assessment results are compared with the warning threshold. When the preliminary risk assessment results are less than the warning threshold, normal monitoring is maintained, and the latest monitoring data in the cleaned dataset is received and analyzed in real time. When the preliminary risk assessment results are greater than the warning threshold, a risk assessment report is generated, which includes information such as the slope instability risk level and assessment time, to provide data support for subsequent decision-making. The warning threshold reflects the real-time changes in crack width, temperature changes, and vibration frequency characteristics.

[0056] It should be noted that by using a quantitative coupling model and dynamic early warning threshold setting, the early warning threshold can be automatically adjusted based on the latest monitoring data under different times and geological conditions, accurately assessing the risk of slope instability. Compared with traditional methods, this improves the adaptability and real-time performance of the early warning method, avoids the rigidity of traditional fixed threshold settings, and enables the early warning to be automatically optimized according to environmental changes. This not only improves the accuracy of risk assessment but also reduces the risk of false alarms and missed alarms, enhancing the effectiveness of early warning.

[0057] S4. Based on the risk assessment report and warning threshold, receive the cleaned dataset in real time and compare it with the latest monitoring data. When the latest monitoring data exceeds the warning threshold, trigger the corresponding warning and generate warning information.

[0058] S4.1 Receive the latest monitoring data of the cleaned dataset in real time based on the risk assessment report and early warning threshold.

[0059] Specifically, based on the risk assessment report and early warning thresholds, the frequency and conditions for receiving real-time monitoring data are determined to ensure that the data can be acquired on time. The latest monitoring data in the cleaned dataset, including characteristic data such as crack width, temperature changes, and vibration frequency, is received in real time and enters the processing flow through the transmission path, ensuring that all data formats are consistent with the preset standards. After the latest monitoring data is received, a preliminary check and verification are performed to ensure that the data is complete and error-free, and it is prepared for comparison with the risk assessment report and early warning thresholds.

[0060] S4.2 Compare the latest monitoring data with the warning threshold to determine whether the warning threshold has been exceeded.

[0061] Specifically, when comparing the latest monitoring data with the warning threshold, the crack width, temperature change, and vibration frequency in the latest monitoring data are extracted one by one, and the data format is ensured to meet the predetermined standard. Each feature value is compared with the warning threshold one by one to determine whether it exceeds the set warning threshold.

[0062] S4.3 When the latest monitoring data does not exceed the warning threshold, normal monitoring status shall be maintained.

[0063] Specifically, when the latest monitoring data does not exceed the warning threshold, it is confirmed that all characteristic data (such as crack width, temperature change, and vibration frequency) do not exceed the warning threshold. The latest raw monitoring data continues to be received and processed, and the normal monitoring status is maintained without triggering any warning signals. The monitoring data will continue to be collected, stored, and analyzed to ensure that potential changes or anomalies are detected in a timely manner, and to wait for further monitoring data or changes in the warning threshold. If the new monitoring data still remains within the warning threshold, the monitoring status will continue.

[0064] S4.4 When the latest monitoring data exceeds the warning threshold, the corresponding warning will be automatically triggered and warning information will be generated.

[0065] Specifically, when the latest monitoring data exceeds the warning threshold, the monitoring data exceeding the warning threshold is marked, the degree of exceeding the range is confirmed, and according to the set warning rules, combined with the severity of exceeding the warning threshold, the corresponding warning information is generated. The warning information includes the specific characteristics of exceeding the limit, the reason for triggering the warning, and possible risks.

[0066] It should also be noted that the early warning rules are mainly based on historical monitoring data, slope stability analysis results, geological characteristics, meteorological conditions, and the dynamic change patterns of the slope. The early warning rules set corresponding early warning thresholds based on different risk levels of the slope, such as the changing trends of characteristic values ​​like crack width, vibration energy, and temperature changes. Based on the fluctuation range of historical data and actual environmental conditions, reasonable early warning trigger points are determined by analyzing past early warning cases and disaster occurrence patterns to ensure that early warning signals can be issued in a timely manner when potential risks occur on the slope.

[0067] S5. Optimize the latest monitoring data, reduce transmission load, generate processed monitoring data, make early warning decisions, and generate optimized early warning information.

[0068] S5.1 Denoise, compress, and quantize the latest monitoring data to generate an optimized dataset.

[0069] S5.1.1 Use Kalman filtering to filter noise from the original monitoring data, remove sensor errors and environmental interference noise, and generate a denoised dataset.

[0070] Specifically, based on the original monitoring data, the existing sensor errors and environmental interference noise are identified. The state estimate and error covariance matrix of the Kalman filter are initialized, and the process noise covariance and observation noise covariance are set. According to the prediction steps of the Kalman filter, the state estimate of the previous moment and the control input are used to predict the state of the current moment, and new observation data are acquired. The predicted value is calculated using the state estimate of the previous moment and the control input, reflecting the expected state of the current moment. The observed value is the data acquired by the sensor in real time, reflecting the actual physical state, such as crack width and temperature changes. The residual calculation method is used to calculate the difference between the observed data and the predicted state. The Kalman filter updates the state estimate using the predicted value and the observed value, corrects the noise in the original monitoring data, and reduces sensor errors and environmental interference. Through repeated iterations, the Kalman filter calculates the residual based on each updated predicted value and the new observed value, and uses the residual to adjust the state estimate, continuously optimizing the matching degree between the predicted value and the actual observed value. Each iteration adjusts the state estimate based on the updated residual, reduces the influence of noise, and generates a denoised dataset.

[0071] S5.1.2. Use lossless compression to compress the denoised dataset to generate a compressed dataset.

[0072] Specifically, the denoised dataset is analyzed to identify redundant information in crack width, temperature change, and vibration frequency features. The LZ77 compression algorithm is used to compress this redundant information. During compression, based on the frequency of occurrence of each element of crack width, temperature change, and vibration frequency features in the denoised dataset, features with higher frequencies are represented by shorter codes, while features with lower frequencies are represented by longer codes. This optimizes storage space while maintaining the integrity of crack width, temperature change, and vibration frequency features. The compressed dataset is then generated after processing with the LZ77 compression algorithm. The compressed dataset can be decompressed to restore the denoised dataset, ensuring data accuracy and integrity.

[0073] It should also be noted that the LZ77 compression algorithm reduces data redundancy and achieves compression by finding duplicate strings in the input data and replacing these duplicate parts with pointers to the positions and lengths of previously occurring identical strings.

[0074] S5.1.3 Utilize dynamic quantization algorithms to transform the compressed dataset into a format that is efficient for processing and storage, generating an optimized dataset.

[0075] Specifically, a dynamic quantization algorithm is used to process the compressed dataset. The compressed dataset is divided into blocks, and each block of data values ​​is divided according to a preset quantization level to obtain the discrete value of each data point. By dynamically adjusting the quantization step size, the error in the quantization process is minimized. Based on the discrete value of each data point, the quantization level is dynamically selected so that the data representation can be compressed efficiently during storage and processing, while keeping important information intact. This generates an optimized dataset that can be stored and processed efficiently in a more compact format.

[0076] It should also be noted that the preset quantization levels are set based on the feature distribution and actual needs of the dataset. The quantization range and accuracy are determined based on statistical indicators such as the maximum, minimum, and standard deviation of crack width, temperature change, and vibration frequency characteristics. According to the quantization range, the dataset is divided into multiple levels to ensure that storage space can be effectively compressed without losing information. The quantization levels are generally divided into several discrete levels, each corresponding to a data range. The specific level classification is adaptively adjusted according to the distribution of feature data. For example, for crack width, the range of 0 to 5 mm may be divided into 5 levels, while for temperature change, it may be divided into 10 levels based on the fluctuation characteristics.

[0077] S5.2 Convert the optimized dataset into a standardized format to generate a standardized format dataset.

[0078] Specifically, each feature data in the optimized dataset is normalized to convert the range of each feature data to a uniform scale, usually 0 to 1, to avoid the impact of differences in the dimensions of different features on subsequent analysis; mean removal and standard deviation scaling are performed on each feature data so that the mean of each feature data is 0 and the standard deviation is 1, in order to eliminate the influence of different data dimensions; each data item after processing conforms to a standardized format and can be organized and saved as a standardized format dataset through formatting.

[0079] S5.3 Optimize the transmission of standardized format data to reduce transmission delay and improve transmission efficiency, and generate processed monitoring data.

[0080] Specifically, lossless compression of standardized format data is performed using data compression methods to reduce storage space requirements and transmission load; optimized encoding of the compressed data is performed using data encoding, and Huffman coding is used to convert the data into a format more suitable for efficient transmission; network latency is reduced and transmission efficiency is improved by selecting appropriate data transmission protocols and optimizing transmission paths, ensuring stability and reliability during data transmission; the optimized data is saved as processed monitoring data.

[0081] S5.4 Analyze the processed monitoring data, identify potential risk patterns, and generate a risk feature dataset.

[0082] Specifically, feature extraction is performed on the processed monitoring data to identify potential risk factors related to slope instability, such as crack width, vibration energy, and temperature changes; principal component analysis is used to analyze crack width, vibration energy, and temperature changes to identify potential risk patterns; the risk patterns obtained from the analysis are quantified to generate a risk feature dataset.

[0083] S5.5 Based on the risk feature dataset, a quantitative coupling model is used to assess the slope instability risk and generate preliminary risk assessment results.

[0084] Specifically, the various features in the risk feature dataset are associated with a quantitative coupling model. Each risk feature is processed through the quantitative coupling model, and the mathematical relationships within the model are used to calculate the degree of influence of each feature on the slope instability risk. Based on the output of the quantitative coupling model, the correlation between each feature and the slope instability risk is analyzed, and the contribution of each feature to the overall risk is calculated. Based on the degree of influence of each feature, a preliminary risk assessment result for slope instability is obtained. Based on the preliminary assessment result, the assessment result is divided according to different risk levels to generate risk level data. Combined with the actual environment and geological conditions, the adaptability and real-time nature of the assessment result are ensured, and a preliminary risk assessment result is generated.

[0085] It should also be noted that actual environment and geological conditions refer to the natural environmental factors and geological structure characteristics in a specific area. The actual environment includes climatic factors such as temperature, humidity and precipitation, which may affect the expansion, contraction or freeze-thaw cycle of the rock mass, thereby changing the stability of the rock mass. Geological conditions, including the degree of rock fragmentation, crack distribution, soil type, and slope inclination angle, determine the rock mass's strength, deformation capacity, and resistance to external forces. When conducting risk assessments, it is important to consider factors that can provide more accurate slope instability risk assessment results, so that the assessment results can reflect the actual risk situation under the current environmental and geological conditions.

[0086] S5.6 Compare and analyze the preliminary risk assessment results and early warning thresholds to generate early warning judgment criteria, and compare them with the processed monitoring data to generate preliminary early warning information.

[0087] Specifically, the preliminary risk assessment results are compared and analyzed with the warning thresholds. This requires sorting the various risk values ​​in the preliminary risk assessment results and comparing them with the warning thresholds to determine whether each risk value exceeds the warning threshold. Based on the comparison results, warning judgment criteria are set. If a risk value exceeds the warning threshold, it is determined to be a warning triggered state, and preliminary warning information is generated. The processed monitoring data is then compared with the warning judgment criteria to verify whether the processed monitoring data meets the warning judgment criteria. The warning criteria are further optimized as needed to ensure the accuracy and timeliness of the warning information.

[0088] S5.7 Verify the preliminary warning information, generate a warning verification report, and combine the processed monitoring data to assess the current warning level and generate optimized warning information.

[0089] Specifically, based on the indicators in the preliminary warning information, the actual measured values ​​in the processed monitoring data are compared to analyze whether the warning triggering conditions meet the actual situation. A warning verification report is generated, including the comparison results between the preliminary warning information and the actual monitoring data, and pointing out potential errors or deviations. Based on the comparison results, combined with key parameters in the processed monitoring data (such as crack width and temperature changes), the current warning level is assessed to determine whether it needs to be adjusted. Based on the assessment results, optimized warning information is generated to ensure that the warning level can accurately reflect the actual situation of slope instability risk.

[0090] In summary, this invention utilizes a neural network model to perform multivariate regression analysis on the cleaned dataset, generating a quantitative coupling model. This enables precise modeling of the complex nonlinear relationships in slope instability, thereby improving the adaptability and accuracy of the quantitative coupling model. It provides a reliable basis for setting dynamic early warning thresholds and risk assessment, and ultimately enhances the accuracy and responsiveness of early warning identification.

[0091] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions, characterized in that: include, Collect raw monitoring data and preprocess it to generate a cleaned dataset; Using a neural network model, multivariate regression analysis was performed on the cleaned dataset to generate a quantitative coupling model; The slope instability risk is assessed based on a quantitative coupling model, and a dynamic early warning threshold is set by combining the cleaned dataset to generate a risk assessment report. Based on the risk assessment report and early warning threshold, the cleaned dataset is received in real time and compared with the latest monitoring data. When the latest monitoring data exceeds the early warning threshold, the corresponding early warning is triggered and early warning information is generated. The latest monitoring data is optimized to reduce transmission load, generate processed monitoring data, and make early warning decisions to generate optimized early warning information.

2. The method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions as described in claim 1, characterized in that: The original monitoring data includes crack width, vibration energy, number of freeze-thaw cycles, slope inclination angle, temperature, humidity, and rainfall.

3. The method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions as described in claim 1, characterized in that: The specific steps for generating the cleaned dataset are as follows. The raw monitoring data is formatted to generate a preliminary cleaned dataset; The initial cleaned dataset is subjected to noise removal and error correction to generate an accurate dataset. The accurate dataset is standardized and redundant features are removed to generate a cleaned dataset.

4. The method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions as described in claim 1, characterized in that: The specific steps for generating the quantitative coupling model are as follows: Slope instability-related features were extracted from the cleaned dataset, and data transformation and redundancy removal were performed to generate a standardized feature dataset. The standardized feature dataset is input into the neural network model, trained by multivariate regression analysis, and the neural network weights are optimized using the backpropagation algorithm to generate the trained neural network model. By using the trained neural network model, predictions are made on a new standardized feature dataset and quantitative coupling relationships are established to generate a quantitative coupling model.

5. The method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions as described in claim 4, characterized in that: The process involves inputting a standardized feature dataset into a neural network model, training it using multivariate regression analysis, optimizing the neural network weights using the backpropagation algorithm, and generating the trained neural network model. The specific steps are as follows: The standardized feature dataset is input into the neural network model, and the activation values ​​are calculated through forward propagation to generate preliminary prediction results; Based on the preliminary prediction results, the backpropagation algorithm is used to adjust the neural network weights to generate optimized neural network weights. Through iterative training and cross-validation, the neural network parameters and weights are continuously adjusted to generate a trained neural network model.

6. The method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions as described in claim 1, characterized in that: The specific steps for generating the risk assessment report are as follows: Based on the quantitative coupling model, the cleaned dataset is analyzed to calculate the assessment value of slope instability risk and generate preliminary risk assessment results. Based on the crack width, temperature change, and vibration frequency characteristics in the cleaned dataset, an adaptive algorithm is used to calculate and set early warning thresholds related to the current slope condition. The preliminary risk assessment results are compared with the warning threshold to generate a risk assessment report.

7. The method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions as described in claim 1, characterized in that: The specific steps for generating the early warning information are as follows: Based on the risk assessment report and early warning thresholds, receive the latest monitoring data of the cleaned dataset in real time; The latest monitoring data is compared with the warning threshold to determine whether the warning threshold has been exceeded. If the latest monitoring data does not exceed the warning threshold, normal monitoring status shall be maintained; When the latest monitoring data exceeds the warning threshold, the corresponding warning will be automatically triggered and a warning message will be generated.

8. The method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions as described in claim 1, characterized in that: The specific steps for generating the processed monitoring data are as follows. The latest monitoring data is denoised, compressed, and quantized to generate an optimized dataset; The optimized dataset is converted into a standardized format to generate a standardized format dataset. Standardized format data is processed for transmission optimization to reduce transmission latency and improve transmission efficiency, generating processed monitoring data.

9. The method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions as described in claim 8, characterized in that: The specific steps for denoising, compressing, and quantizing the latest monitoring data to generate an optimized dataset are as follows. Kalman filtering is used to filter noise from the original monitoring data, removing sensor errors and environmental interference noise, and generating a denoised dataset. A lossless compression method is used to compress the denoised dataset to generate a compressed dataset. By using a dynamic quantization algorithm, the compressed dataset is transformed into a format that is efficient for processing and storage, generating an optimized dataset.

10. The method for early warning and identification of slope instability in extremely fractured rock masses in high-altitude and cold regions as described in claim 1, characterized in that: The specific steps for generating the optimized early warning information are as follows: Analyze the processed monitoring data to identify potential risk patterns and generate a risk feature dataset; Based on the risk feature dataset, a quantitative coupling model is used to assess the slope instability risk and generate preliminary risk assessment results. By comparing and analyzing the preliminary risk assessment results and the early warning threshold, early warning judgment criteria are generated, and then compared with the processed monitoring data to generate preliminary early warning information. The initial warning information is verified, a warning verification report is generated, and the current warning level is assessed by combining the processed monitoring data to generate optimized warning information.

Images