Weak surrounding rock tunnel deformation intelligent monitoring method and system

Abstract

The invention discloses a weak surrounding rock tunnel deformation intelligent monitoring method and system. The monitoring method comprises the following steps: acquiring original time sequence data of tunnel deformation monitoring; decomposing the original time sequence data to obtain decomposed time sequence data; obtaining a feature vector set based on the decomposed time sequence data; the feature vector set is input into a tunnel deformation monitoring model, a tunnel deformation monitoring result is obtained, the tunnel deformation monitoring model is constructed through a support vector regression model and is obtained through training of a training set, and the training set is historical monitoring data. Through fusion application of various algorithms, the tunnel deformation monitoring precision and reliability are remarkably improved, and powerful support is provided for safe operation of the tunnel.

Description

Technical Field

[0001] The present invention belongs to the technical field of surrounding rock monitoring, and in particular relates to an intelligent deformation monitoring method and system for soft surrounding rock tunnels. Background Art

[0002] In the process of intelligent deformation monitoring of soft rock tunnels, data preprocessing is a key link. Due to the complex and changeable tunnel environment, the monitoring equipment is easily interfered by the outside world, resulting in a large amount of noise and outliers in the collected raw data. These noises and outliers will have a serious impact on the subsequent feature extraction and model building, reducing the reliability and accuracy of the monitoring system. Therefore, how to effectively clean and denoise data has become a technical problem faced by the intelligent deformation monitoring of soft rock tunnels.

[0003] Traditional data preprocessing methods, such as median filtering and moving average, often fail to achieve ideal results for noise and outliers in complex tunnel environments. Moreover, these methods usually require manual setting of thresholds and parameters, lacking adaptability and flexibility. How to automatically optimize data preprocessing algorithms and parameters according to different tunnel environments and monitoring requirements, and improve the efficiency and accuracy of data cleaning and denoising, is an urgent problem to be solved.

[0004] In addition, tunnel deformation monitoring data has the characteristics of time series, and there are complex correlations and dependencies between data in different time periods. How to fully consider the influence of the time dimension and mine the time series characteristics of data during data preprocessing is also a challenging problem. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes an intelligent deformation monitoring method and system for soft surrounding rock tunnels, which can improve the accuracy and reliability of tunnel deformation monitoring and provide strong support for the safe operation of tunnels.

[0006] The present invention provides an intelligent deformation monitoring method for a soft surrounding rock tunnel, comprising:

[0007] Obtain the original time series data of tunnel deformation monitoring;

[0008] Decomposing the original time series data to obtain decomposed time series data;

[0009] Based on the decomposed time series data, obtaining a set of feature vectors;

[0010] The feature vector set is input into a tunnel deformation monitoring model to obtain a monitoring result of tunnel deformation, wherein the tunnel deformation monitoring model is constructed by a support vector regression model and obtained by training with a training set, and the training set is historical monitoring data.

[0011] Optionally, decompose the original time series data to obtain the decomposed time series data, including:

[0012] Using time series decomposition technology, decompose the original time series data into a combination of a trend component, a periodic component, and a residual component, where

[0013] The trend component is used to judge the overall change trend of tunnel deformation;

[0014] The periodic component is used to identify the periodic characteristics in tunnel deformation.

[0015] Optionally, based on the decomposed time series data, obtain a set of feature vectors, including:

[0016] Judge whether the residual component contains significant noise data, and process the residual component according to the judgment result;

[0017] Recombine the processed residual component with the trend component and the periodic component to obtain the preliminarily cleaned time series data;

[0018] Perform abnormal data elimination processing on the preliminarily cleaned time series data to obtain the cleaned time series data;

[0019] Use the dynamic time warping algorithm to segment the cleaned time series data, obtain the segmented time series data, extract the frequency domain features of the segmented time series data, and generate a set of feature vectors.

[0020] Optionally, judging whether the residual component contains significant noise data and processing the residual component according to the judgment result includes:

[0021] If the fluctuation amplitude of the residual component exceeds the preset noise threshold, it is judged that the residual component contains significant noise data;

[0022] For the residual component containing significant noise, use the wavelet transform denoising algorithm to perform multi-scale decomposition on the residual component, and extract the wavelet coefficients at different frequency scales;

[0023] According to the noise characteristics, perform threshold processing on the wavelet coefficients to remove the noise components in the wavelet coefficients, and obtain the denoised wavelet coefficients;

[0024] Use the denoised wavelet coefficients to reconstruct the residual component through wavelet inverse transform to obtain the smoothed and denoised residual component.

[0025] Optionally, performing abnormal data elimination processing on the preliminarily cleaned time series data to obtain the cleaned time series data includes:

[0026] For the time series data after preliminary cleaning, the local outlier factor algorithm is used to calculate the local density of each data point. By comparing with a preset anomaly threshold, it is determined whether the data point is abnormal data;

[0027] If the local density of the data point is lower than the preset anomaly threshold, the data point is marked as abnormal data and removed from the time series data after preliminary cleaning to obtain the cleaned time series data.

[0028] Optionally, the dynamic time warping algorithm is used to segment the cleaned time series data to obtain the segmented time series data, and the frequency domain features of the segmented time series data are extracted to generate a feature vector set including:

[0029] Set a time window. According to the first data and the time window, the dynamic time warping algorithm is used for calculation to compare the similarity between data of different lengths. By setting a threshold to judge the similarity, if the similarity is greater than the threshold, no segmentation is performed; if the similarity is less than the threshold, the first data is segmented to obtain the second segmented data;

[0030] According to the second segmented data, calculate the mean value of each segment of data. The mean value represents the central tendency of the data in this segment to obtain the mean value data;

[0031] According to the second segmented data and the mean value data, calculate the standard deviation of each segment of data. The standard deviation represents the degree of dispersion of the data in this segment to obtain the standard deviation data;

[0032] According to the second segmented data, the mean value data and the standard deviation data, calculate the kurtosis of each segment of data. The kurtosis represents the degree of steepness of the data distribution form in this segment to obtain the kurtosis data;

[0033] According to the second segmented data, the mean value data and the standard deviation data, calculate the skewness of each segment of data. The skewness represents the direction and degree of skewness of the data distribution in this segment to obtain the skewness data;

[0034] According to the second segmented data, use the fast Fourier transform method to calculate the power spectral density of each segment of data. Through the power spectral density, obtain the frequency domain features of this segment of data. According to the mean value data, the standard deviation data, the kurtosis data, the skewness data and the power spectral density data, generate the feature vector of each segment of data to obtain the feature vector set.

[0035] Optionally, obtaining the training set includes:

[0036] According to the historical tunnel deformation monitoring data, extract the multi-dimensional time series feature vectors reflecting the tunnel deformation characteristics;

[0037] Based on the extracted feature vector set, obtain the training set.

[0038] The present invention also provides an intelligent monitoring system for tunnel deformation in soft surrounding rock, comprising: a data acquisition module, a data decomposition module, a feature extraction module and a monitoring module;

[0039] The data acquisition module is used to acquire the original time series data of tunnel deformation monitoring, and the original time series data includes noise data and abnormal data;

[0040] The data decomposition module is used to decompose the original time series data into a combination of a trend component, a periodic component and a residual component by using time series decomposition technology, so as to obtain the decomposed time series data;

[0041] The feature extraction module is used to obtain a set of feature vectors based on the decomposed time series data

[0042] The monitoring module is used to input the set of feature vectors into a tunnel deformation monitoring model to obtain the monitoring result of tunnel deformation. Among them, the tunnel deformation monitoring model is constructed by a support vector regression model and obtained through training with a training set, and the training set is historical monitoring data.

[0043] Optionally, the feature extraction module includes: a noise processing unit and an anomaly detection unit;

[0044] The noise processing unit is used to judge whether the decomposed residual component contains significant noise data. If the fluctuation amplitude of the residual component exceeds a preset noise threshold, the wavelet transform denoising algorithm is used to smooth the residual component to obtain the denoised residual component;

[0045] The anomaly detection unit is used to recombine the denoised residual component with the trend component and the periodic component to generate the preliminarily cleaned time series data. For the preliminarily cleaned time series data, the local outlier factor algorithm is used to judge whether there is abnormal data. If the local density of the data point is lower than a preset anomaly threshold, it is marked as abnormal data and removed to obtain the cleaned time series data.

[0046] Compared with the prior art, the present invention has the following advantages and technical effects:

[0047] In view of the noise and anomaly problems existing in the original time series data, this invention uses time series decomposition, wavelet transform denoising, and local outlier factor algorithm for data cleaning. The cleaned data is segmented through the dynamic time warping algorithm, statistical and frequency domain features are extracted, and a set of feature vectors is generated. The feature vectors are input into the support vector regression model for training to obtain a tunnel deformation monitoring model. This method can effectively remove the noise and outliers in the original data, extract key features, and achieve accurate monitoring and prediction of tunnel deformation. Through the integrated application of multiple algorithms, this invention significantly improves the accuracy and reliability of tunnel deformation monitoring, providing strong support for the safe operation of tunnels. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments and descriptions thereof of this application are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0049] Figure 1 is a flowchart of an intelligent monitoring method for the deformation of a soft surrounding rock tunnel in an embodiment of this invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. This application will be described in detail below with reference to the drawings and in combination with the embodiments.

[0051] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0052] The following elaborates in detail on the professional terms involved in this embodiment:

[0053] The main reasons for the deformation of a soft surrounding rock tunnel include geological conditions, meteorological conditions, groundwater level, and human factors. Specifically:

[0054] Geological conditions: For example, the strata, faults, landslides, earthquakes, etc. of folded mountains will cause soil layer deformation.

[0055] Meteorological conditions: Extreme climates such as heavy rain, strong wind, ice and snow will cause the soft soil to collapse due to water infiltration and erosion.

[0056] Groundwater level: Groundwater can increase the pore pressure of the soil, making the soil in a "supersaturated" state, thus destroying the stability of the soil.

[0057] Human factors: For example, the construction of projects such as airplane airports, railway tunnels, and cable tunnels will also cause underground rock and soil deformation.

[0058] The main characteristics of the deformation of tunnels in soft surrounding rock include:

[0059] Large deformation: The deformation of soft rock tunnels can reach more than 10 cm, and that of squeezing soft rock tunnels can reach more than 30 cm, or even more than 100 cm.

[0060] High deformation rate: Especially in the early stage, the deformation rate is large and fast.

[0061] Long deformation duration: Due to the high rheology and low strength of soft surrounding rock, the duration of stress redistribution after excavation is long.

[0062] Control measures for the deformation of tunnels in soft surrounding rock

[0063] In order to effectively control the deformation of tunnels in soft surrounding rock, the following measures can be taken:

[0064] Use advanced mathematical models for simulation analysis, and obtain the possible degree of surrounding rock deformation through simulation analysis, so as to formulate appropriate measures before implementation of control.

[0065] Strengthen the support of the tunnel head, use bolts or cable bolts with higher strength for anchoring, and at the same time strengthen the anti-seepage treatment of groundwater at the tunnel head, effectively reduce the influence of poor surrounding rock on the tunnel head, and ensure the stability of the tunnel entrance.

[0066] Adopt appropriate support methods, such as precast steel grid bridges, steel and wooden structure systems, and installation of brackets on the inner wall of the tunnel, etc., to increase the adaptability to peak periods of tunnel closure, fire disasters, earthquake disasters, etc.

[0067] Strengthen construction supervision, and strictly construct in accordance with design documents and construction specifications to prevent the out-of-control of soft surrounding rock of the tunnel due to quality problems.

[0068] This embodiment proposes an intelligent monitoring method for the deformation of tunnels in soft surrounding rock, as Figure 1 shown, which specifically includes the following steps:

[0069] Obtain the original time series data of tunnel deformation monitoring;

[0070] Decompose the original time series data to obtain the decomposed time series data;

[0071] Based on the decomposed time series data, obtain the set of feature vectors;

[0072] Input the set of feature vectors into the tunnel deformation monitoring model to obtain the monitoring results of tunnel deformation, where the tunnel deformation monitoring model is constructed by a support vector regression model and obtained through training with a training set, and the training set is historical monitoring data.

[0073] Specifically, the original time series data of tunnel deformation monitoring is obtained. The original time series data contains noise data and abnormal data. The time series decomposition technology is used to decompose the original time series data into a combination of trend component, periodic component and residual component to obtain the decomposed time series data;

[0074] For the decomposed residual component, determine whether it contains significant noise data. If the fluctuation amplitude of the residual component exceeds the preset noise threshold, the wavelet transform denoising algorithm is used to smooth the residual component to obtain the denoised residual component.

[0075] The denoised residual component is recombined with the trend component and the periodic component to generate the time series data after preliminary cleaning. For the time series data after preliminary cleaning, the local anomaly factor algorithm is used to determine whether there is abnormal data. If the local density of the data point is lower than the preset abnormal threshold, it is marked as abnormal data and removed to obtain the cleaned time series data;

[0076] According to the cleaned time series data, the dynamic time warping algorithm is used to segment the data to obtain the segmented time series data. For the segmented time series data, the statistical characteristics such as the mean, standard deviation, kurtosis, skewness and frequency domain characteristics such as power spectrum density of each segment are extracted to generate a set of feature vectors.

[0077] The feature vector set is input into the preset support vector regression model for model training and optimization to obtain the tunnel deformation monitoring model. The tunnel deformation monitoring model is used to predict and analyze the time series data collected in real time, and the monitoring results of tunnel deformation are output.

[0078] Furthermore, the original time series data is decomposed to obtain the decomposed time series data including:

[0079] Using time series decomposition technology, the original time series data is decomposed into a combination of trend component, period component and residual component, where:

[0080] Trend component, used to determine the overall trend of tunnel deformation;

[0081] The periodic component is used to identify periodic features in tunnel deformation.

[0082] Specifically, obtain the original time series data of tunnel deformation monitoring. For the noise data and abnormal data contained in the original data, use time series decomposition technology for processing. Through time series decomposition, the original time series data is decomposed into a trend component, a periodic component, and a residual component. According to the obtained trend component, judge the overall change trend of tunnel deformation. If the trend component shows an obvious upward or downward trend, it indicates that there is a risk of continuous deformation of the tunnel, and it is necessary to further analyze the cause of deformation and take corresponding measures. For the obtained periodic component, use methods such as Fourier transform for frequency domain analysis to identify the periodic characteristics in tunnel deformation. If there is obvious periodic deformation, it is necessary to combine the environmental conditions of the tunnel to judge the cause of the periodic deformation, such as the influence of factors such as temperature change and traffic flow load. For the obtained residual component, use statistical methods for outlier detection and processing. By setting a residual threshold, the data points exceeding the threshold range are marked as outliers. For the identified outliers, methods such as interpolation and smoothing can be used for correction to eliminate the influence of abnormal data on subsequent analysis. After completing the outlier processing, reconstruct the corrected residual component with the trend component and the periodic component to obtain the tunnel deformation time series data after removing noise and outliers. Use the reconstructed time series data to establish a tunnel deformation prediction model, such as machine learning models like support vector machines and neural networks, to realize the prediction of the tunnel deformation trend in the future for a period of time, and provide support for tunnel safety monitoring and maintenance decision-making. According to the tunnel deformation prediction results, comprehensively consider factors such as the importance of the tunnel, the deformation rate, and the deformation amount to evaluate the health status of the tunnel, determine the safety level and maintenance level of the tunnel, and formulate corresponding monitoring frequencies and maintenance measure plans to ensure the safe operation of the tunnel.

[0083] Furthermore, based on the decomposed time series data, obtain a set of feature vectors including:

[0084] Judge whether the residual component contains significant noise data, and process the residual component according to the judgment result;

[0085] Recombine the processed residual component with the trend component and the periodic component to obtain the preliminarily cleaned time series data;

[0086] Perform outlier removal processing on the preliminarily cleaned time series data to obtain the cleaned time series data;

[0087] Use the dynamic time warping algorithm to segment the cleaned time series data to obtain the segmented time series data, extract the frequency domain features of the segmented time series data, and generate a set of feature vectors.

[0088] Specifically, obtain the decomposed residual components and perform noise judgment processing on the residual components. Calculate the fluctuation amplitude of the residual components and compare it with a preset noise threshold. If the fluctuation amplitude of the residual components exceeds the preset noise threshold, it is determined that the residual components contain significant noise data. For the residual components containing significant noise, use the wavelet transform denoising algorithm to denoise the residual components. In the wavelet transform denoising algorithm, perform multi-scale decomposition on the residual components through wavelet basis functions to extract wavelet coefficients at different frequency scales. According to the noise characteristics, perform threshold processing on the wavelet coefficients to remove the noise components in the wavelet coefficients and obtain the denoised wavelet coefficients. Use the denoised wavelet coefficients to reconstruct the residual components through wavelet inverse transform to obtain the smoothed denoised residual components.

[0089] Further, determine whether the residual components contain significant noise data, and the processing of the residual components according to the judgment result includes:

[0090] If the fluctuation amplitude of the residual components exceeds the preset noise threshold, it is determined that the residual components contain significant noise data;

[0091] For the residual components containing significant noise, use the wavelet transform denoising algorithm to perform multi-scale decomposition on the residual components and extract wavelet coefficients at different frequency scales;

[0092] According to the noise characteristics, perform threshold processing on the wavelet coefficients to remove the noise components in the wavelet coefficients and obtain the denoised wavelet coefficients;

[0093] Use the denoised wavelet coefficients to reconstruct the residual components through wavelet inverse transform to obtain the smoothed denoised residual components.

[0094] Further, perform abnormal data elimination processing on the preliminarily cleaned time series data to obtain the cleaned time series data, including:

[0095] For the preliminarily cleaned time series data, use the local outlier factor algorithm to calculate the local density of each data point, and judge whether the data point is abnormal data by comparing it with a preset outlier threshold;

[0096] If the local density of the data point is lower than the preset outlier threshold, mark the data point as abnormal data and remove it from the preliminarily cleaned time series data to obtain the cleaned time series data.

[0097] Specifically, obtain the original time series data, perform denoising processing on the original time series data to obtain the denoised residual component, trend component, and periodic component. Recombine the denoised residual component, trend component, and periodic component, and obtain the initially cleaned time series data through a combination algorithm. For the initially cleaned time series data, use the Local Outlier Factor algorithm to calculate the local density of each data point, and determine whether the data point is an abnormal data point by comparing it with a preset anomaly threshold. If the local density of the data point is lower than the preset anomaly threshold, mark the data point as abnormal data and remove it from the initially cleaned time series data to obtain the further cleaned time series data. For the further cleaned time series data, use the Z-score normalization method to standardize the data, eliminate the dimensional differences between different data points, and obtain the standardized time series data. According to the standardized time series data, decompose it into a trend term, a periodic term, and a random term through a time series decomposition algorithm, and establish corresponding prediction models respectively to obtain the predicted values of the time series data for a future period. Combine the predicted trend term, periodic term, and random term, and restore them to the scale of the original data through inverse normalization processing, and finally obtain the predicted results of the time series data for a future period as a reference basis for business decisions.

[0098] More specifically, for the further cleaned time series data, use the Z-score normalization method to standardize the data, eliminate the dimensional differences between different data points, and obtain the standardized time series data.

[0099] Sort the data points according to the timestamps of the time series data in the time dimension to ensure the continuity and integrity of the time series. Use data smoothing methods such as moving average and median filtering to denoise the time series data, eliminate the influence of outliers and random fluctuations, and obtain the smoothed time series data. Calculate the mean and standard deviation of the time series data, and perform normalization processing on each data point according to the Z-score normalization formula to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1. Through the Z-score normalization process, the dimensional differences between different data points are eliminated, and the time series data is converted into dimensionless normalized data, which is convenient for subsequent data analysis and modeling. Extract features from the normalized time series data to obtain statistical features of the time series, such as mean, variance, skewness, kurtosis, etc., as descriptive indicators of the time series. Use the time series decomposition method to decompose the normalized time series data into a trend term, a seasonal term, and a random term, and model and predict different components separately. According to the normalized time series data and the extracted features, select suitable machine learning algorithms, such as ARIMA, LSTM, etc., to build a time series prediction model for predicting and analyzing future time series data.

[0100] Furthermore, use the dynamic time warping algorithm to segment the cleaned data to obtain the segmented time series data, and extract the frequency domain features of the segmented time series data to generate a set of feature vectors including:

[0101] Set a time window, and use the dynamic time warping algorithm to calculate according to the first data and the time window, compare the similarity between data of different lengths, and judge the similarity through the set threshold. If the similarity is greater than the threshold, no segmentation is performed. If the similarity is less than the threshold, the first data is segmented to obtain the second segmented data;

[0102] According to the second segmented data, calculate the mean of each segment of data. The mean represents the central tendency of the data in this segment to obtain the mean data;

[0103] According to the second segmented data and the mean data, calculate the standard deviation of each segment of data. The standard deviation represents the degree of dispersion of the data in this segment to obtain the standard deviation data;

[0104] According to the second segmented data, the mean data, and the standard deviation data, calculate the kurtosis of each segment of data. The kurtosis represents the steepness of the distribution form of the data in this segment to obtain the kurtosis data;

[0105] According to the second segmented data, the mean data, and the standard deviation data, calculate the skewness of each segment of data. The skewness represents the direction and degree of skewness of the data distribution in this segment to obtain the skewness data;

[0106] According to the second data after segmentation, the power spectral density of each segment of data is calculated using the fast Fourier transform method, and the frequency-domain characteristics of this segment of data are obtained through the power spectral density. According to the mean data, standard deviation data, kurtosis data, skewness data, and power spectral density data, a feature vector for each segment of data is generated to obtain a set of feature vectors.

[0107] Furthermore, obtaining the training set includes:

[0108] According to the historical tunnel deformation monitoring data, extract the multi-dimensional time series feature vectors reflecting the tunnel deformation characteristics;

[0109] Based on the extracted set of feature vectors, obtain the training set.

[0110] Specifically, according to the historical tunnel deformation monitoring data, extract the multi-dimensional time series feature vectors reflecting the tunnel deformation characteristics. Take the extracted set of feature vectors as training samples and input them into a preset support vector regression model for training. During the training process of the support vector regression model, optimize the model hyperparameters through grid search and cross-validation methods to obtain the tunnel deformation monitoring model with the optimal performance. Use the wireless sensor network to collect the surrounding environment parameters of the tunnel and deformation monitoring data such as strain and displacement in real time to form multi-dimensional time series data. Preprocess the collected time series data, including data cleaning, data synchronization, data normalization, etc., to obtain the standardized time series input data. Input the standardized time series input data into the trained and optimized tunnel deformation monitoring model for deformation trend prediction and abnormal deformation detection and analysis. According to the prediction and analysis results of the tunnel deformation monitoring model, output the tunnel deformation monitoring report, including the deformation trend prediction curve, abnormal deformation warning information, etc., to provide a decision-making basis for the safe operation and management of the tunnel.

[0111] More specifically, input the standardized time series input data into the trained and optimized tunnel deformation monitoring model for deformation trend prediction and abnormal deformation detection and analysis.

[0112] Obtain the time series data of tunnel deformation monitoring, preprocess the data, including denoising, smoothing, etc., to obtain the standardized time series input data. According to the standardized time series input data, adopt a long short-term memory neural network model, and construct a tunnel deformation trend prediction model by setting the input layer, hidden layer, and output layer. Utilize the historical data of tunnel deformation monitoring to train and optimize the constructed prediction model. Through the backpropagation algorithm, adjust the model parameters to minimize the prediction error and obtain the optimized tunnel deformation prediction model. Input the standardized time series tunnel deformation monitoring data into the optimized prediction model, and through forward propagation calculation, obtain the tunnel deformation trend prediction results for a period of time in the future. According to the tunnel deformation monitoring data and prediction results, calculate indicators such as deformation rate and acceleration, and judge whether there is abnormal deformation by comparing with the preset threshold. If abnormal deformation is detected, trigger the early warning mechanism, and according to the location, degree, etc. of the abnormal deformation, take corresponding safety measures, such as reinforcement, closure, etc., to ensure the safe operation of the tunnel. Visualize the tunnel deformation prediction results and abnormal detection results, generate deformation trend curves, abnormal deformation distribution maps, etc., to facilitate engineering personnel to intuitively analyze the tunnel deformation situation and timely discover and handle potential safety hazards.

[0113] This embodiment also provides an intelligent monitoring system for tunnel deformation in soft surrounding rock, including: a data acquisition module, a data decomposition module, a feature extraction module, and a monitoring module;

[0114] The data acquisition module is used to obtain the original time series data of tunnel deformation monitoring, and the original time series data includes noise data and abnormal data;

[0115] The data decomposition module is used to adopt time series decomposition technology to decompose the original time series data into a combination of trend components, periodic components, and residual components to obtain the decomposed time series data;

[0116] The feature extraction module is used to obtain a set of feature vectors based on the decomposed time series data

[0117] The monitoring module is used to input the set of feature vectors into the tunnel deformation monitoring model to obtain the monitoring results of tunnel deformation. Among them, the tunnel deformation monitoring model is constructed by a support vector regression model and obtained through training with a training set, and the training set is historical monitoring data.

[0118] Furthermore, the feature extraction module includes: a noise processing unit and an abnormal detection unit;

[0119] The noise processing unit is used to judge whether the decomposed residual component contains significant noise data. If the fluctuation amplitude of the residual component exceeds the preset noise threshold, the wavelet transform denoising algorithm is used to smooth the residual component to obtain the denoised residual component;

[0120] An anomaly detection unit is used to recombine the denoised residual component with the trend component and the periodic component to generate the preliminarily cleaned time series data. For the preliminarily cleaned time series data, the local outlier factor algorithm is adopted to determine whether there is abnormal data. If the local density of a data point is lower than the preset anomaly threshold, it is marked as abnormal data and removed to obtain the cleaned time series data.

[0121] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An intelligent deformation monitoring method for soft surrounding rock tunnels, characterized in that: include: Obtain the original time series data of tunnel deformation monitoring; Decomposing the original time series data to obtain decomposed time series data; Based on the decomposed time series data, obtaining a set of feature vectors; The feature vector set is input into a tunnel deformation monitoring model to obtain a monitoring result of tunnel deformation, wherein the tunnel deformation monitoring model is constructed by a support vector regression model and obtained by training with a training set, and the training set is historical monitoring data.

2. The method for intelligently monitoring deformation of soft surrounding rock tunnels according to claim 1 is characterized in that: Decomposing the original time series data to obtain the decomposed time series data includes: The original time series data is decomposed into a combination of trend component, period component and residual component by using time series decomposition technology, wherein: The trend component is used to determine the overall change trend of tunnel deformation; The periodic component is used to identify periodic features in tunnel deformation.

3. The method for intelligently monitoring deformation of soft surrounding rock tunnels according to claim 2 is characterized in that: Based on the decomposed time series data, obtaining a feature vector set includes: Determine whether the residual component contains significant noise data, and process the residual component according to the determination result; Recombining the processed residual component with the trend component and the period component to obtain time series data after preliminary cleaning; Performing abnormal data elimination processing on the time series data after the preliminary cleaning to obtain the cleaned time series data; A dynamic time warping algorithm is used to segment the cleaned time series data to obtain segmented time series data, and frequency domain features of the segmented time series data are extracted to generate a feature vector set.

4. The method for intelligently monitoring deformation of soft surrounding rock tunnels according to claim 3 is characterized in that: Determining whether the residual component contains significant noise data, and processing the residual component according to the determination result includes: If the fluctuation amplitude of the residual component exceeds the preset noise threshold, it is judged that the residual component contains significant noise data; For the residual components containing significant noise, the wavelet transform denoising algorithm is used to perform multi-scale decomposition on the residual components and extract the wavelet coefficients at different frequency scales; According to the noise characteristics, the wavelet coefficients are threshold processed to remove the noise components in the wavelet coefficients and obtain the denoised wavelet coefficients; The denoised wavelet coefficients are used to reconstruct the residual component through inverse wavelet transform to obtain the denoised residual component after smoothing.

5. The method for intelligently monitoring deformation of soft surrounding rock tunnels according to claim 3 is characterized in that: The time series data after the preliminary cleaning is subjected to abnormal data elimination processing, and the time series data after the cleaning is obtained includes: For the time series data after preliminary cleaning, the local anomaly factor algorithm is used to calculate the local density of each data point, and by comparing it with the preset anomaly threshold, it is determined whether the data point is an abnormal data; If the local density of a data point is lower than a preset abnormal threshold, the data point is marked as abnormal data and removed from the time series data after preliminary cleaning to obtain the cleaned time series data.

6. The method for intelligently monitoring deformation of soft surrounding rock tunnels according to claim 3 is characterized in that: The cleaned time series data is segmented using a dynamic time warping algorithm to obtain segmented time series data, and the frequency domain features of the segmented time series data are extracted to generate a feature vector set including: A time window is set, and a dynamic time warping algorithm is used to perform calculations based on the first data and the time window, and the similarities between data of different lengths are compared. The similarity is determined by a set threshold. If the similarity is greater than the threshold, no segmentation is performed; if the similarity is less than the threshold, the first data is segmented to obtain segmented second data. According to the segmented second data, the mean of each segment of data is calculated, and the mean represents the central trend of the segment of data, and the mean data is obtained; According to the segmented second data and mean data, the standard deviation of each segment of data is calculated, and the standard deviation represents the degree of dispersion of the segment of data, and the standard deviation data is obtained; According to the segmented second data, mean data and standard deviation data, the kurtosis of each segment of data is calculated. The kurtosis represents the steepness of the distribution of the data in this segment, and the kurtosis data is obtained. According to the segmented second data, mean data and standard deviation data, the skewness of each segment of data is calculated. The skewness represents the direction and degree of the skewness of the data distribution in this segment, and the skewness data is obtained. According to the segmented second data, the fast Fourier transform method is used to calculate the power spectral density of each segment of data, and the frequency domain characteristics of the segment of data are obtained through the power spectral density. According to the mean data, standard deviation data, kurtosis data, skewness data and power spectral density data, the feature vector of each segment of data is generated to obtain a feature vector set.

7. The method for intelligently monitoring deformation of soft surrounding rock tunnels according to claim 1 is characterized in that: Acquiring the training set includes: According to the historical monitoring data of tunnel deformation, the multi-dimensional time series feature vector reflecting the tunnel deformation characteristics is extracted; Based on the extracted feature vector set, the training set is obtained.

8. An intelligent deformation monitoring system for soft surrounding rock tunnels, characterized in that: include: Data acquisition module, data decomposition module, feature extraction module and monitoring module; The data acquisition module is used to acquire original time series data of tunnel deformation monitoring, wherein the original time series data includes noise data and abnormal data; The data decomposition module is used to decompose the original time series data into a combination of trend component, period component and residual component by using time series decomposition technology to obtain decomposed time series data; The feature extraction module is used to obtain a feature vector set based on the decomposed time series data. The monitoring module is used to input the feature vector set into the tunnel deformation monitoring model to obtain the monitoring result of the tunnel deformation, wherein the tunnel deformation monitoring model is constructed by a support vector regression model and obtained by training with a training set, and the training set is historical monitoring data.

9. The intelligent monitoring system for deformation of soft surrounding rock tunnels according to claim 8 is characterized in that: The feature extraction module includes: a noise processing unit and an anomaly detection unit; The noise processing unit is used to determine whether the decomposed residual component contains significant noise data, and if the fluctuation amplitude of the residual component exceeds a preset noise threshold, the residual component is smoothed by using a wavelet transform denoising algorithm to obtain a denoised residual component; The anomaly detection unit is used to recombine the denoised residual component with the trend component and the periodic component to generate time series data after preliminary cleaning. For the time series data after preliminary cleaning, a local anomaly factor algorithm is used to determine whether there is abnormal data. If the local density of the data point is lower than a preset anomaly threshold, it is marked as abnormal data and removed to obtain the cleaned time series data.

Images