Method and system for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion

Abstract

The present invention provides a method and system for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion, which relates to the technical field of landslide monitoring. The method comprises: acquiring collected data, and based on the landslide surface deformation time series data set, screening one or more landslide surface deformation time series data that meet preset related conditions with the landslide internal tunnel deformation time series data; using the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data as sample data, and establishing a multivariate time series analysis model based on the sample data. The present invention completes the landslide internal tunnel deformation time series data at the latest time point, and improves the temporal continuity and robustness of the landslide internal tunnel deformation time series data.

Description

Technical Field

[0001] The present invention relates to the technical field of landslide monitoring, and in particular to a method and system for predicting deformation of a tunnel in a landslide based on fusion of multi-source monitoring data. Background Art

[0002] Landslide is the process in which rock and soil move downward along a weak zone under the influence of gravity and other factors. Landslide monitoring refers to the measurement and interpretation of the changing process and triggering factors of the deformation and movement of geological bodies, and then the evaluation of disaster risks and prediction of development trends.

[0003] The monitoring of landslide geology with tunnels can be divided into the following three categories according to the type of monitored objects: one is the internal information of geological bodies, including surface displacement, deep displacement, Newtonian force and tectonic stress; the second is external inducement information, including meteorology, earthquakes, water bank erosion and human activities; the third is other indirect information, including abnormal animal behavior and abnormal vegetation growth. Among them, the deep displacement has better consistency and directness in response to the sliding displacement of the slip surface, and is limitedly affected by external factors in terms of accuracy and stability. Therefore, it is widely used in determining the position of the sliding surface, monitoring the internal deformation of the sliding body, revealing the deformation characteristics of slope creep deformation, evaluating its stability, and guiding landslide early warning and prevention.

[0004] Under the existing technology, there are various means to implement landslide surface deformation monitoring for landslide geology with tunnels. However, due to the limitations of environmental factors, the monitoring methods for tunnels inside landslides are currently mainly based on engineering measurements with a long collection period. Therefore, during the monitoring process, the tunnel deformation monitoring data has problems such as insufficient temporal continuity and weak data robustness, which is not conducive to the joint analysis of tunnel deformation monitoring data and landslide surface deformation. Summary of the invention

[0005] The present invention aims to solve the problem that in the process of monitoring landslide geology with tunnels, the tunnel deformation monitoring data has insufficient time series continuity and weak data robustness.

[0006] To solve the above problems, in a first aspect, the present invention provides a method for predicting tunnel deformation in a landslide based on multi-source monitoring data fusion, comprising:

[0007] Acquire collected data, wherein the collected data includes a landslide surface deformation time series data set and a landslide internal tunnel deformation time series data, wherein the landslide surface deformation time series data set includes a plurality of landslide surface deformation time series data;

[0008] Based on the landslide surface deformation time series data set, screening one or more landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data;

[0009] The screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data are used as sample data, a multivariate time series analysis model for predicting the landslide internal tunnel deformation data is established based on the sample data, and the multivariate time series analysis model is trained and verified to obtain a multivariate time series analysis optimization model;

[0010] The collected data is updated, and the deformation data of the tunnel inside the landslide is predicted based on the updated collected data and the multivariate time series analysis optimization model.

[0011] Optionally, the acquiring of collected data includes:

[0012] Based on a plurality of deformation monitoring sensors in different acquisition modes, a plurality of landslide surface deformation time series data are obtained, wherein the plurality of landslide deformation monitoring sensors include remote sensing radar, laser radar and navigation positioning monitor;

[0013] The engineering survey data collected by the tunnel total station is obtained, and the engineering survey data is used as the time series data of the tunnel deformation inside the landslide.

[0014] Optionally, before screening one or more of the landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data based on the landslide surface deformation time series data set, the method further includes:

[0015] Determine a corresponding preprocessing method according to the type of the landslide surface deformation time series data and the landslide internal tunnel deformation time series data in the collected data, wherein the preprocessing method includes one or more of data anomaly elimination processing, data missing completion processing, data error smoothing processing, and data normalization processing;

[0016] According to the determined preprocessing method, the landslide internal tunnel deformation time series data and the landslide surface deformation time series data are respectively preprocessed accordingly;

[0017] The pre-processed collected data is used as screening data for screening the landslide surface deformation time series data.

[0018] Optionally, the screening of one or more landslide surface deformation time series data sets that meet preset correlation conditions with the landslide internal tunnel deformation time series data comprises:

[0019] sequentially determining the correlation coefficient between each of the plurality of landslide surface deformation time series data and the landslide internal tunnel deformation time series data in the landslide surface deformation time series data set;

[0020] The plurality of landslide surface deformation time series data corresponding to the correlation coefficients greater than a preset coefficient threshold are used as the screened landslide surface deformation time series data.

[0021] Optionally, the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data are used as sample data, and a multivariate time series analysis model for predicting landslide internal tunnel deformation data is established based on the sample data, including:

[0022] According to the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data, the autoregressive terms of tunnel deformation at multiple first historical moments, the prediction error terms of tunnel deformation at multiple second historical moments, the landslide surface deformation prediction terms at future moments under multiple landslide surface deformation time series data, and the landslide surface deformation prediction error terms are determined, and based on the determination results, the multivariate time series analysis model is constructed.

[0023] Optionally, the multivariate time series analysis model includes:

[0024]

[0025] The four terms on the right side of the above equation are the autoregressive term, the prediction error term of the tunnel deformation, the landslide surface deformation prediction term, and the landslide surface deformation prediction error term. t is the tunnel deformation data at time t, y t-1 is the tunnel deformation data of the tunnel deformation history data set at time t-1, φ i is the coefficient of the ith autoregressive term, p is the order of the autoregressive term, ε t-j is the prediction error of the tunnel deformation time series data of the landslide at time tj, q is the sliding mean order, θ j is the jth sliding average coefficient, X z,t is the Zth type of landslide surface deformation data at time t of the multiple landslide surface deformation time series data, m is the number of types of landslide surface deformation historical data sets, β z is the model coefficient corresponding to the historical data set of landslide surface deformation of the Zth type, ε t is the prediction error of the tunnel deformation time series data inside the landslide at time t.

[0026] Optionally, the training and verifying the multivariate time series analysis model to obtain the multivariate time series analysis optimization model includes:

[0027] Obtaining a value range of the autoregressive term order and the sliding average term order, and determining multiple groups of autoregressive term orders and sliding average term orders according to specific values ​​in the value range;

[0028] According to each group of the autoregressive term order and the moving average term order, a corresponding autoregressive moving average model is established;

[0029] Evaluate each of the autoregressive moving average models using the Bayesian Information Criterion and determine the corresponding BIC value;

[0030] All determined BIC values ​​are compared, and the autoregressive term order and the sliding average term order under the autoregressive moving average model corresponding to the minimum BIC value are selected as the autoregressive term order and the sliding average term order in the multivariate time series analysis optimization model.

[0031] Optionally, the training and verifying the multivariate time series analysis model to obtain the multivariate time series analysis optimization model includes:

[0032] The accuracy is verified by the root mean square error between the predicted value and the measured value of the tunnel deformation data;

[0033] If the accuracy verification is passed, the corresponding multivariate time series analysis model is used as the multivariate time series analysis optimization model;

[0034] If the accuracy verification fails, the multivariate time series analysis model is retrained until the accuracy verification passes, thereby obtaining the multivariate time series analysis optimization model.

[0035] Optionally, updating the collected data and predicting the deformation data of the tunnel inside the landslide based on the updated collected data and the multivariate time series analysis optimization model includes:

[0036] Determine a new batch of sample data according to the latest collected data;

[0037] Each time a new batch of sample data is determined, the earliest batch of sample data is eliminated, and the multivariate time series analysis optimization model is trained and verified, thereby updating the multivariate time series analysis optimization model;

[0038] Substitute the latest landslide surface deformation time series data set in the collected data into the updated multivariate time series analysis optimization model to output the latest landslide internal tunnel deformation data.

[0039] The present invention, based on the landslide surface deformation time series data set, screens one or more of the landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data, that is, the landslide surface deformation time series data that exhibits at least one of different data volumes, different sensitivities, and strong correlations are coupled with the landslide internal tunnel deformation monitoring sensor that exhibits at least one of high precision and high cost, thereby enhancing the connection and fusion between various types of landslide monitoring data; constructs and trains a multivariate time series analysis model based on the landslide internal tunnel deformation time series data and the landslide surface deformation data set that is uniformly distributed, has complete time series, and has rich data sources, thereby improving the quality and diversity of the data set used for the model, thereby increasing the accuracy and generalization ability of the model training; and By using the multivariate time series analysis optimization model and substituting various landslide surface deformation time series data, the deformation of the tunnel inside the landslide can be regressed and calculated. Since the landslide surface data are evenly distributed and the data is sufficient, the predicted internal deformation of the landslide can be evenly distributed at any position in the tunnel, solving the problem of insufficient spatial resolution of the landslide tunnel deformation monitoring data. Secondly, in the period when the time resolution of precision engineering measurements inside the landslide is insufficient or monitoring is missing, the relatively complete time series characteristics of the landslide surface deformation time series data set can be used to supplement the time series data of the tunnel deformation in the landslide at the latest time point, so as to monitor the internal deformation degree of the landslide in real time and improve the robustness of the time series data of the tunnel deformation in the landslide, providing an effective theoretical method and data support for the monitoring and analysis of landslide deformation under landslide geology with tunnels.

[0040] In a second aspect, the present invention also provides a landslide tunnel deformation prediction system based on multi-source monitoring data fusion, comprising:

[0041] An acquisition module, used for acquiring collected data, wherein the collected data includes a landslide surface deformation time series data set and a landslide internal tunnel deformation time series data set, wherein the landslide surface deformation time series data set includes a plurality of landslide surface deformation time series data;

[0042] A screening module, for screening one or more of the landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data based on the landslide surface deformation time series data set;

[0043] A training and verification module, used to use the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data as sample data, establish a multivariate time series analysis model for predicting the landslide internal tunnel deformation data based on the sample data, and train and verify the multivariate time series analysis model to obtain a multivariate time series analysis optimization model;

[0044] An updating and prediction module is used to update the collected data and predict the deformation data of the tunnel inside the landslide based on the updated collected data and the multivariate time series analysis optimization model.

[0045] In a third aspect, the present invention provides an electronic device, including a memory and a processor;

[0046] The memory is used to store computer programs;

[0047] The processor is used to implement the prediction of tunnel deformation in a landslide based on multi-source monitoring data fusion as described in the first aspect when executing the computer program.

[0048] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion as described in the first aspect is implemented.

[0049] The system for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion, the electronic device and the computer-readable storage medium provided by the present invention have the same beneficial effects as the method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion relative to the prior art, which will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic flow chart of a method for predicting deformation of a tunnel in a landslide based on multi-source monitoring data fusion in an embodiment of the present invention is shown;

[0051] Figure 2 A deformation monitoring data distribution diagram of a landslide body and a landslide scene in an embodiment of the present invention is shown;

[0052] Figure 3 A schematic diagram of predicting the internal deformation of a landslide by integrating multi-source data of the landslide surface and the tunnel in a certain landslide body according to an embodiment of the present invention is shown;

[0053] Figure 4 A schematic diagram of a process of fusing landslide surface deformation data and tunnel deformation data based on a multivariate time series analysis model in a landslide scene in an embodiment of the present invention is shown;

[0054] Figure 5 A schematic diagram of the timing of stability and fluctuation in an embodiment of the present invention is shown;

[0055] Figure 6 A schematic diagram showing that the fluctuation of data in an embodiment of the present invention may exceed the actual deformation situation;

[0056] Figure 7 A flowchart of establishing a multivariate time series analysis model for inverting deformation data in a tunnel in a landslide scenario according to an embodiment of the present invention is shown;

[0057] Figure 8 The schematic diagram of the structure of the landslide tunnel deformation prediction system based on multi-source monitoring data fusion in an embodiment of the present invention is shown;

[0058] Fig. 9 A schematic diagram of the structure of an electronic device in an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0059] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0060] It should be noted that, in the present invention, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0061] In the description of this specification, the description with reference to the terms "embodiment", "one embodiment" and "one implementation" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or implementation are included in at least one embodiment or implementation of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or implementation. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or implementations in a suitable manner.

[0062] Reference Figure 1 , Figure 2 and Figure 3 As shown, the embodiment of the present invention proposes a method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion;

[0063] The method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion includes:

[0064] S100: Acquire collected data, wherein the collected data includes a landslide surface deformation time series data set and a landslide internal tunnel deformation time series data, wherein the landslide surface deformation time series data set includes a plurality of landslide surface deformation time series data.

[0065] Specifically, the collected data are all time series data, which characterize the corresponding relationship between deformation and time. The landslide surface deformation time series data set is easy to obtain. For example, it can be obtained by using a scanner through the lidar of an unmanned aerial vehicle during flight, and by installing a SAR instrument on a satellite orbit, a large range of surface deformation information can be obtained without being restricted by ground obstruction. Therefore, a variety of landslide surface deformation time series data measured by landslide surface deformation monitoring sensors with different data volumes, different sensitivities and strong correlations can be obtained, which are also easy to obtain; and for the landslide internal tunnel deformation time series data, it is limited by the tunnel environment, the internal signals and the layout method, etc., and is generally obtained through a total station detector combined with manual monitoring. It is the landslide internal tunnel deformation time series data collected by the high-precision and high-cost landslide internal tunnel deformation monitoring sensor.

[0066] S200: Based on the landslide surface deformation time series data set, one or more landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data are screened.

[0067] Specifically, based on correlation analysis, relevant points are found in the time series data set of the landslide internal tunnel deformation, and the landslide surface deformation data with strong correlation with the landslide internal tunnel deformation data are screened out as feature variables. Feature variables are important input variables used to train models and make predictions, which means that a landslide surface deformation data set with uniform distribution, complete time series and rich data sources can be obtained.

[0068] S300: Using the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data as sample data, establishing a multivariate time series analysis model for predicting landslide internal tunnel deformation data based on the sample data, and training and verifying the multivariate time series analysis model to obtain a multivariate time series analysis optimization model.

[0069] Specifically, a multivariate time series analysis model for predicting the deformation data of the tunnel inside the landslide is established based on the sample data, such as a data regression model. The multivariate time series analysis model is constructed and trained based on the time series data of the tunnel deformation inside the landslide and the landslide surface deformation data set with uniform distribution, complete time series and rich data sources, and the data mapping relationship between the characteristic variables of the landslide surface deformation and the deformation inside the tunnel is learned; due to the precision engineering measurement distributed inside the tunnel, although the data accuracy is very high, there are objective problems such as insufficient time resolution and discrete spatial distribution, and it is impossible to be evenly distributed in the complete branch tunnel of the tunnel, while the landslide deformation data is evenly distributed and the historical data is sufficient, which is convenient for inputting into the time series analysis model for training, which can improve the quality and diversity of the data set used for the model, thereby increasing the accuracy and generalization ability of model training;

[0070] S400: updating the collected data, and predicting the deformation data of the tunnel inside the landslide based on the updated collected data and the multivariate time series analysis optimization model.

[0071] Combined with the above analysis, the landslide surface data at the most recent time node is sufficient, but the landslide deformation time series in the tunnel is missing. This step can firstly optimize the model again by updating the collected data to obtain the multivariate time series analysis optimization model. By using the multivariate time series analysis optimization model and substituting various landslide surface deformation time series data, the deformation of the tunnel inside the landslide can be regressed and calculated. Since the landslide surface data is evenly distributed and the data is sufficient, the predicted internal deformation of the landslide can be evenly distributed at any position in the tunnel, solving the problem of insufficient spatial resolution of the landslide tunnel deformation monitoring data. Secondly, in the period when the time resolution of the precision engineering measurement inside the landslide is insufficient or the monitoring is missing, the time series of the landslide surface deformation time series data set can be used to complete the tunnel deformation time series data at the latest time point, so as to monitor the internal deformation degree of the landslide in real time.

[0072] It should be noted that in some actual application scenarios, external inducement information, including relatively stable conditions such as meteorology, earthquakes, water bank erosion, and human activities, as well as other indirect information, including relatively stable conditions such as abnormal animal behavior and abnormal vegetation growth, can be combined with the present embodiment to complete the time series data of tunnel deformation in the landslide at the latest time point.

[0073] When the embodiment of the present invention is applied in practice, based on the landslide surface deformation time series data set, one or more of the landslide surface deformation time series data that meet the preset correlation conditions with the landslide internal tunnel deformation time series data are screened, that is, the landslide surface deformation time series data that exhibits at least one of different data volumes, different sensitivities, and strong correlations are coupled with the landslide internal tunnel deformation monitoring sensor that exhibits at least one of high precision and high cost, thereby enhancing the connection and fusion between various types of landslide monitoring data; based on the landslide internal tunnel deformation time series data and the landslide surface deformation data set with uniform distribution, complete time series, and rich data sources, a multivariate time series analysis model is constructed and trained, which can improve the quality and diversity of the data set used for the model, thereby increasing the accuracy and Generalization ability: by using the multivariate time series analysis optimization model and substituting various landslide surface deformation time series data, the deformation of the tunnel inside the landslide can be regressed and calculated. Since the landslide surface data are evenly distributed and the data is sufficient, the predicted internal deformation of the landslide can be evenly distributed at any position in the tunnel, solving the problem of insufficient spatial resolution of the landslide tunnel deformation monitoring data. Secondly, in the period when the time resolution of precision engineering measurements inside the landslide is insufficient or monitoring is missing, the time series of the landslide surface deformation time series data set can be used to complete the time series data of the tunnel deformation in the landslide at the latest time point, so as to monitor the internal deformation degree of the landslide in real time and improve the robustness of the time series data of the tunnel deformation in the landslide, providing an effective theoretical method and data support for the monitoring and analysis of landslide deformation under landslide geology with tunnels.

[0074] In summary, the present invention decomposes landslide deformation monitoring into two relatively independent monitoring methods based on sensor location distribution, and solves the problems of insufficient utilization, insufficient fusion and insufficient joint analysis of tunnel monitoring data in the current landslide monitoring process through data association and joint modeling; through the meta-time series analysis model, the landslide surface deformation data set with different time series coverage and different monitoring implementation types is fused with the landslide tunnel deformation data set, providing a new and easy-to-implement solution for the landslide monitoring process, effectively improving the integrity and effectiveness of the landslide monitoring time series.

[0075] like Figure 3 As shown, as an optional embodiment of the present invention, the acquisition of collected data includes:

[0076] Based on a plurality of deformation monitoring sensors in different acquisition modes, a plurality of landslide surface deformation time series data are obtained, wherein the plurality of landslide deformation monitoring sensors include remote sensing radar, laser radar and navigation positioning monitor;

[0077] Specifically, according to the positional relationship of landslide deformation monitoring sensors, they can be divided into space-based InSAR (remote sensing radar) for monitoring landslide surface deformation, UAV LiDAR (UAV laser radar), and Beidou / GNSS positioning (navigation and positioning monitor).

[0078] The engineering survey data collected by the tunnel total station is obtained, and the engineering survey data is used as the time series data of the tunnel deformation inside the landslide.

[0079] Specifically, the precise engineering measurement of the deformation inside the tunnel of the landslide is monitored, and the time series data of the deformation inside the tunnel of the landslide is collected by the tunnel total station.

[0080] like Figure 4 As shown, as an optional embodiment of the present invention, before screening one or more of the landslide surface deformation time series data that meet the preset correlation conditions with the landslide internal tunnel deformation time series data based on the landslide surface deformation time series data set, it also includes:

[0081] Determine a corresponding preprocessing method according to the type of the landslide surface deformation time series data and the landslide internal tunnel deformation time series data in the collected data, wherein the preprocessing method includes one or more of data anomaly elimination processing, data missing completion processing, data error smoothing processing, and data normalization processing;

[0082] Specifically, the data outliers are eliminated based on the following formula to eliminate the outliers in the historical monitoring data of landslide deformation:

[0083]

[0084] Among them, x is the sample data, x i is the i-th deformation data in the continuous time series data in the sample data set; n is the number of historical monitoring data; x is the average value of the continuous time series data, S x is the standard deviation. If the absolute value of the difference between a data point and the mean is greater than three times the standard deviation, the data point will be identified as an outlier and removed;

[0085] The missing values ​​in the historical monitoring data of landslide deformation are filled by the following formula:

[0086]

[0087] Where N is the time period or sliding window size for selecting the moving average method; X t-1 +X t-2 +…+X t-Nis the sum of the landslide deformation monitoring data in the past N time points; then SMA t is the moving average at time t in the historical monitoring dataset.

[0088] Data error smoothing is performed by using five-point cubic smoothing on the historical monitoring data of landslide deformation to reduce the measurement error contained in each data point. The formula used is as follows:

[0089]

[0090] in, Y i The value after data smoothing, Y i is the i-th deformation data in the continuous time series data in the sample data set; the value range of i is from 1 to n, and n is the number of historical monitoring data.

[0091] Data normalization processing, the maximum and minimum normalization method is used to standardize the historical monitoring data of landslide deformation, as shown in the following formula:

[0092]

[0093] in, is the result of normalization of the sample data set; zi is the i-th deformation data in the continuous time series data in the sample data set; z min is the minimum value of the continuous time series data in the sample data set; z max It is the maximum value of the continuous time series data in the sample data set.

[0094] According to the determined preprocessing method, the landslide internal tunnel deformation time series data and the landslide surface deformation time series data are respectively preprocessed accordingly;

[0095] The preprocessed collected data is used as the screening data for screening the landslide surface deformation time series data. The data preprocessing process can obtain a complete landslide deformation regular grid data set, and the time series of each point can be supplemented (that is, both the time resolution and the spatial resolution can be satisfied).

[0096] Specifically, there are two preprocessing methods. One is to directly select a preprocessing method, and the other is to directly process in the order of data anomaly elimination, data missing completion, data error smoothing, and data normalization.

[0097] The processing of different types of data should be selected by category. For example, space-based InSAR data has a wide coverage and less missing data in daily units. More attention should be paid to data outlier processing and error smoothing, especially in areas with large cloud coverage, where data fluctuations may exceed actual deformation.

[0098] In the precision engineering measurement of landslides, the time resolution of data is often low (generally, monitoring data is returned only once in a long time), which requires the completion of key data. However, due to the high accuracy of precision engineering measurement, the processing of data error smoothing is relatively ignored. The project monitoring accuracy in a specific example is shown as follows: Figure 6 shown.

[0099] However, ignoring the work efficiency, the four types of data processing can be carried out in sequence to obtain relatively effective results. The accuracy and resolution of data collection methods such as ground-based InSAR, GNSS static observation, and precision engineering measurement are statistically shown in Table 1 in a specific example;

[0100] Table 1:

[0101]

[0102] Among them, 1-3 is the collection method of landslide surface deformation time series data, and 8 is the collection method of precision engineering measurement, that is, the collection method of landslide internal tunnel deformation time series data.

[0103] In summary, during the data preprocessing process, one data processing process or a combination of multiple data processing processes can be used each time. The choice of specific data processing needs to be combined with the characteristics of the data itself. The purpose is to enhance the effectiveness and stability of the data set, improve the efficiency of subsequent data screening, and improve modeling efficiency.

[0104] As an optional embodiment of the present invention, the screening of one or more landslide surface deformation time series data sets that meet preset correlation conditions with the landslide internal tunnel deformation time series data includes:

[0105] sequentially determining the correlation coefficient between each of the plurality of landslide surface deformation time series data and the landslide internal tunnel deformation time series data in the landslide surface deformation time series data set;

[0106] The plurality of landslide surface deformation time series data corresponding to the correlation coefficients greater than a preset coefficient threshold are used as the screened landslide surface deformation time series data.

[0107] Specifically, the correlation coefficient is expressed as follows:

[0108]

[0109] In the above formula, r is the correlation coefficient formula; X i represents the i-th sample in the X variable in the landslide surface deformation dataset, is the mean of the X variable; Y i represents the i-th sample in the landslide tunnel deformation dataset, is the mean of the Y variable.

[0110] For example, when the calculated correlation coefficient is greater than 0.7, the corresponding landslide internal tunnel deformation time series data and landslide surface deformation time series data set (which contains several landslide surface deformation data at this time) can be identified as feature point data and enter the model training. However, if the amount of data involved in the training of the correlation coefficient is larger, the more time it takes, and the theoretically wider the applicability of the model, both should be considered comprehensively.

[0111] When this embodiment is applied in practice, it is possible to distinguish several types of landslide surface deformation data whose correlation coefficients with the landslide internal tunnel deformation time series data are higher than a preset coefficient threshold, thereby obtaining highly correlated characteristic variables, which are used as considerations when constructing a multivariate time series analysis and prediction model and as input variables after the model is constructed.

[0112] As an optional embodiment of the present invention, the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data are used as sample data, and a multivariate time series analysis model for predicting landslide internal tunnel deformation data is established based on the sample data, including:

[0113] According to the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data, the autoregressive terms of tunnel deformation at multiple first historical moments, the prediction error terms of tunnel deformation at multiple second historical moments, the landslide surface deformation prediction terms at future moments under multiple landslide surface deformation time series data, and the landslide surface deformation prediction error terms are determined, and based on the determination results, the multivariate time series analysis model is constructed.

[0114] Exemplarily, the multivariate time series analysis model includes:

[0115]

[0116] The four terms on the right side of the above equation are the autoregressive term, the prediction error term of the tunnel deformation, the landslide surface deformation prediction term, and the landslide surface deformation prediction error term. t is the tunnel deformation data at time t, y t-1 is the tunnel deformation data of the tunnel deformation history data set at time t-1, φ iis the coefficient of the ith autoregressive term, p is the order of the autoregressive term, ε t-j is the prediction error of the tunnel deformation time series data of the landslide at time tj, q is the sliding mean order, θ j is the jth sliding average coefficient, X z,t is the Zth type of landslide surface deformation data at time t of the multiple landslide surface deformation time series data, m is the number of types of landslide surface deformation historical data sets, β z is the model coefficient corresponding to the historical data set of landslide surface deformation of the Zth type, ε t is the prediction error of the tunnel deformation time series data inside the landslide at time t.

[0117] When this embodiment is applied in practice, the autoregressive term is used to capture the time series characteristics of the tunnel deformation, and the future trend is predicted through historical data. The landslide surface deformation is considered: the landslide surface deformation data is introduced, and the connection between the landslide surface deformation and the tunnel deformation is established, which improves the prediction accuracy. The prediction error is considered: the sliding average term is used to capture the sequence characteristics of the prediction error, and it is incorporated into the model to reduce the prediction error. Combining autoregression, sliding average and landslide surface deformation data, a multi-factor comprehensive prediction model is constructed, which improves the interpretability and prediction accuracy of the model.

[0118] See also Figure 7 As an optional embodiment of the present invention, the training and verification of the multivariate time series analysis model to obtain the multivariate time series analysis optimization model includes:

[0119] Obtaining a value range of the autoregressive term order and the sliding average term order, and determining multiple groups of autoregressive term orders and sliding average term orders according to specific values ​​in the value range;

[0120] Specifically, the value range of the initialization parameters p and q, for example, the value range of p and q can be set to 1 to 10; choosing appropriate p and q values ​​can enable the multivariate time series analysis model to better fit the data and make more accurate predictions.

[0121] According to each group of the autoregressive term order and the moving average term order, a corresponding autoregressive moving average model is established;

[0122] Specifically, for each group (p, q), a corresponding autoregressive moving average model (ARMA model) is established.

[0123] Evaluate each of the autoregressive moving average models using the Bayesian Information Criterion and determine the corresponding BIC value;

[0124] Each model was evaluated using the Bayesian Information Criterion (BIC) and the BIC value was recorded.

[0125] All determined BIC values ​​are compared, and the autoregressive term order and the sliding average term order under the autoregressive moving average model corresponding to the minimum BIC value are selected as the autoregressive term order and the sliding average term order in the multivariate time series analysis optimization model.

[0126] The calculation formula of BIC value is as follows:

[0127] BIC=kln(n 1 )-2ln(L);

[0128] Where: k is the number of parameters, n 1 is the sample size, and L is the likelihood function.

[0129] The BIC values ​​of all models are compared, and the autoregressive term order and the sliding average term order under the autoregressive moving average model corresponding to the minimum BIC value are selected as the autoregressive term order and the sliding average term order in the multivariate time series analysis optimization model.

[0130] The (p, q) value corresponding to the minimum BIC value is used as the final selected autoregressive term order and sliding average term order. The final determined (p, q) value is the optimal autoregressive term order and sliding average term order in the multivariate time series analysis optimization model. The other parameters in the model are constants.

[0131] The smaller the BIC value, the better the model parameters, and thus the better the performance of the multivariate time series analysis optimization model, which serves as the basis for subsequent model verification.

[0132] It should be noted that before determining p and q, it is sometimes necessary to perform an ADF test to determine the order of the multivariate time series analysis prediction model for smoothing processing, so as to facilitate the determination of the optimal multivariate time series analysis optimization model, that is, determining the order by looking at the graph, or determining the order by recognizing the graph. The specific process is explained below.

[0133] See also Figure 5 The "graph" here refers to the time series graph. The left graph of a stable sequence fluctuates around a constant; while the right graph of an unstable sequence has an obvious trend of growth or decrease.

[0134] A specific standard may be selected as a standard for judging a convergence result. In the embodiment, we selected the unit root test (ADF), and the concept is as follows:

[0135] When using time series models, such as ARMA and ARIMA, the time series is required to be stationary. Therefore, when studying a time series, the first step is to perform a stationarity test. In addition to the objective method of detecting the trend of statistical charts with the naked eye, another commonly used rigorous statistical test method is the ADF test, also called the unit root test.

[0136] The full name of ADF test is Augmented Dickey-Fuller test. As the name implies, ADF is an augmented form of Dickey-Fuller test. DF test can only be applied to first-order cases. When there is a high-order lag correlation in the sequence, ADF test can be used. Therefore, ADF is an extension of DF test.

[0137] When an autoregressive process: y t =by t-1 +a+∈ t If the lag term coefficient b is 1, it is called a unit root. When a unit root exists, the relationship between the independent variable and the dependent variable is deceptive because any error in the residual sequence will not decay as the sample size (i.e., the number of periods) increases, which means that the impact of the residual in the model is permanent. This kind of regression is also called pseudo regression. If a unit root exists, this process is a random walk.

[0138] The principle of the ADF test is to determine whether the sequence has a unit root: if the sequence is stable, there is no unit root; otherwise, there is a unit root. Therefore, the H of the ADF test is 0 The assumption is that there is a unit root. If the obtained significance test statistic is less than three confidence levels (10%, 5%, 1%), then there is a corresponding confidence of (90%, 95%, 99%) to reject the null hypothesis. Here, in the specific embodiment, we choose the unit root test P value (ADF value) less than 0.05 to consider it stable.

[0139] Using the above method, the most suitable stable minimum difference order d is selected using the order determination method. By observing the time series inversion results of the tunnel deformation data, the first-order or second-order difference is selected; then p and q are determined.

[0140] As an optional embodiment of the present invention, the training and verification of the multivariate time series analysis model to obtain the multivariate time series analysis optimization model includes:

[0141] The accuracy is verified by the root mean square error between the predicted value and the measured value of the tunnel deformation data;

[0142] Specifically, the root mean square error (MSE) calculation formula is as follows:

[0143]

[0144] Where m is the number of validation samples of the landslide tunnel deformation dataset, X i represents the i-th sample in the X variable in the landslide surface deformation data set (corresponding to one of the various landslide surface deformation time series data), F(X i ) is the predicted value of the landslide tunnel deformation data calculated by the X variable based on the multivariate time series analysis model; Y i represents the measured value of the i-th sample in the landslide tunnel deformation dataset.

[0145] If the accuracy verification is passed, the corresponding multivariate time series analysis model is used as the multivariate time series analysis optimization model; the corresponding multivariate time series analysis model is the multivariate time series analysis optimization model corresponding to the accuracy verification.

[0146] If the accuracy verification fails, the multivariate time series analysis model is retrained until the accuracy verification passes, thereby obtaining the multivariate time series analysis optimization model.

[0147] Specifically, the RMS value needs to be less than the set threshold value for the accuracy verification to pass, which means that the lower the data dispersion, the more concentrated the data is around the average value, indicating that the predicted value is close enough to the measured value, and also indicating that the predicted value is accurate enough.

[0148] When the RMS value is greater than or equal to the set threshold, the accuracy verification fails. Then the q and q values ​​are reselected for optimization training until the accuracy verification passes. The accuracy of the model is optimized by multivariate time series analysis.

[0149] When this embodiment is applied in practice, through the accuracy verification process, it can be ensured that a multivariate time series analysis optimization model that meets the accuracy is obtained. The multivariate time series analysis optimization model can adapt to the accurate prediction of the landslide internal tunnel deformation data among various landslide surface deformation time series data.

[0150] As an optional embodiment of the present invention, the updating of the collected data and predicting the deformation data of the tunnel inside the landslide based on the updated collected data and the multivariate time series analysis optimization model includes:

[0151] Determine a new batch of sample data according to the latest collected data;

[0152] The data were collected according to the collection frequency as shown in Table 2;

[0153] Table 2;

[0154]

[0155] Specifically, after the data is collected, it is still pre-processed and analyzed for relevance in the aforementioned manner to obtain sample data.

[0156] Each time a new batch of sample data is determined, the earliest batch of sample data is eliminated, and the multivariate time series analysis optimization model is trained and verified, thereby updating the multivariate time series analysis optimization model;

[0157] Specifically, whenever a batch of new sample data is obtained, they are added to the training data of the model. Since the capacity of sample data is generally fixed, the earliest batch of sample data (also the earliest collected) is eliminated, and then the multivariate time series analysis model is trained and verified in the same way as mentioned above to obtain the multivariate time series analysis optimization model, and the update of the multivariate time series analysis optimization model can ensure that reliable and latest landslide internal tunnel deformation data are finally obtained.

[0158] Substitute the latest landslide surface deformation time series data set in the collected data into the updated multivariate time series analysis optimization model to output the latest landslide internal tunnel deformation data.

[0159] Specifically, since the analysis and optimization model is the latest, it conforms to the changes in the deformation of the tunnel inside the landslide corresponding to the latest landslide surface deformation time series data set of the collected data, continuously updates the collected data, and utilizes the landslide surface deformation data set with uniform distribution, complete time series, and rich data sources. The deformation prediction value inside the landslide tunnel is calculated through the multivariate time series analysis model. In general, by completing the landslide deformation data set in the study area, based on the large sample size, long time series and diverse surface deformation monitoring data, the long time series and high-precision deformation monitoring data of the landslide internal tunnel are completed, so that the deformation degree of the internal area of ​​the landslide can be monitored.

[0160] When this embodiment is applied in practice, the landslide surface data at the most recent time node is sufficient, but the landslide deformation time series in the tunnel is missing. First, the landslide surface data can be used to regress and deduce the deformation data of the tunnel inside the landslide. Since the landslide surface data is evenly distributed, the deduced and predicted internal deformation of the landslide can be evenly distributed at any position in the tunnel, solving the problem of insufficient spatial resolution of the landslide tunnel deformation monitoring data. Secondly, in the period when the time resolution of the precision engineering measurement inside the landslide is insufficient or the monitoring is missing, the time series of the landslide surface deformation time series (after preprocessing) can be used to complete the time series data of the tunnel deformation in the landslide at the latest time point, so as to monitor the internal deformation degree of the landslide in real time.

[0161] like Figure 8As shown, the present invention also provides a landslide tunnel deformation prediction system 200 based on multi-source monitoring data fusion, comprising:

[0162] The acquisition module 210 is used to acquire collected data, wherein the collected data includes a landslide surface deformation time series data set and a landslide internal tunnel deformation time series data set, wherein the landslide surface deformation time series data set includes a plurality of landslide surface deformation time series data;

[0163] A screening module 220 is used to screen one or more of the landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data based on the landslide surface deformation time series data set;

[0164] The training and verification module 230 is used to use the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data as sample data, establish a multivariate time series analysis model for predicting the landslide internal tunnel deformation data based on the sample data, and train and verify the multivariate time series analysis model to obtain a multivariate time series analysis optimization model;

[0165] An update and prediction module 240 is used to update the collected data and predict the deformation data of the tunnel inside the landslide based on the updated collected data and the multivariate time series analysis optimization model.

[0166] The specific implementation of this embodiment can refer to the corresponding implementation method mentioned above and will not be described again here.

[0167] like Fig. 9 As shown, an electronic device 300 provided by an embodiment of the present invention includes a memory 310 and a processor 320; the memory 310 is used to store a computer program; the processor 320 is used to implement the above-mentioned method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion when executing the computer program.

[0168] In other words, an electronic device 300 includes a memory 310 and a processor 320 coupled to the memory 310; the memory 310 is configured to store a computer program; and the processor 320 is configured to perform the following operations when executing the computer program:

[0169] Acquire collected data, wherein the collected data includes a landslide surface deformation time series data set and a landslide internal tunnel deformation time series data, wherein the landslide surface deformation time series data set includes a plurality of landslide surface deformation time series data;

[0170] Based on the landslide surface deformation time series data set, screening one or more landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data;

[0171] The screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data are used as sample data, a multivariate time series analysis model for predicting the landslide internal tunnel deformation data is established based on the sample data, and the multivariate time series analysis model is trained and verified to obtain a multivariate time series analysis optimization model;

[0172] The collected data is updated, and the deformation data of the tunnel inside the landslide is predicted based on the updated collected data and the multivariate time series analysis optimization model.

[0173] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for predicting deformation of a tunnel in a landslide based on multi-source monitoring data fusion as described above is implemented.

[0174] In other words, a non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor performs the following operations:

[0175] Acquire collected data, wherein the collected data includes a landslide surface deformation time series data set and a landslide internal tunnel deformation time series data, wherein the landslide surface deformation time series data set includes a plurality of landslide surface deformation time series data;

[0176] Based on the landslide surface deformation time series data set, screening one or more landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data;

[0177] The screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data are used as sample data, a multivariate time series analysis model for predicting the landslide internal tunnel deformation data is established based on the sample data, and the multivariate time series analysis model is trained and verified to obtain a multivariate time series analysis optimization model;

[0178] The collected data is updated, and the deformation data of the tunnel inside the landslide is predicted based on the updated collected data and the multivariate time series analysis optimization model.

[0179] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0180] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

[0181] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion, characterized in that: include: Acquire collected data, wherein the collected data includes a landslide surface deformation time series data set and a landslide internal tunnel deformation time series data, wherein the landslide surface deformation time series data set includes a plurality of landslide surface deformation time series data; Based on the landslide surface deformation time series data set, screening one or more landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data; The screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data are used as sample data, a multivariate time series analysis model for predicting the landslide internal tunnel deformation data is established based on the sample data, and the multivariate time series analysis model is trained and verified to obtain a multivariate time series analysis optimization model; Updating the collected data, and predicting the deformation data of the tunnel inside the landslide based on the updated collected data and the multivariate time series analysis optimization model; The method of using the selected landslide surface deformation time series data and the landslide internal tunnel deformation time series data as sample data and establishing a multivariate time series analysis model for predicting landslide internal tunnel deformation data based on the sample data includes: According to the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data, the autoregressive terms of tunnel deformation at multiple first historical moments, the prediction error terms of tunnel deformation at multiple second historical moments, the landslide surface deformation prediction terms at future moments under multiple landslide surface deformation time series data, and the landslide surface deformation prediction error terms are determined, and based on the determination results, the multivariate time series analysis model is constructed.

2. The method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion according to claim 1 is characterized in that: The screening of one or more landslide surface deformation time series data sets that meet preset correlation conditions with the landslide internal tunnel deformation time series data includes: sequentially determining the correlation coefficient between each of the plurality of landslide surface deformation time series data and the landslide internal tunnel deformation time series data in the landslide surface deformation time series data set; The plurality of landslide surface deformation time series data corresponding to the correlation coefficients greater than a preset coefficient threshold are used as the screened landslide surface deformation time series data.

3. The method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion according to claim 1 is characterized in that: Before screening one or more of the landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data based on the landslide surface deformation time series data set, the method further includes: Determine a corresponding preprocessing method according to the type of the landslide surface deformation time series data and the landslide internal tunnel deformation time series data in the collected data, wherein the preprocessing method includes one or more of data anomaly elimination processing, data missing completion processing, data error smoothing processing, and data normalization processing; According to the determined preprocessing method, the landslide internal tunnel deformation time series data and the landslide surface deformation time series data are respectively preprocessed accordingly; The pre-processed collected data is used as screening data for screening the landslide surface deformation time series data.

4. The method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion according to claim 1 is characterized in that: The acquisition of collected data includes: Based on a plurality of deformation monitoring sensors in different acquisition modes, a plurality of landslide surface deformation time series data are obtained, wherein the plurality of deformation monitoring sensors include remote sensing radar, laser radar and navigation positioning monitor; The engineering survey data collected by the tunnel total station is obtained, and the engineering survey data is used as the time series data of the tunnel deformation inside the landslide.

5. The method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion according to claim 1 is characterized in that: The multivariate time series analysis model includes: ； Among them, the four terms on the right side of the equation are the autoregressive term, the prediction error term of the tunnel deformation, the landslide surface deformation prediction term and the landslide surface deformation prediction error term, is the tunnel deformation data at time t, is the tunnel deformation data of the tunnel deformation history dataset at time t-1, is the coefficient of the ith autoregressive term, p is the order of the autoregressive term, is the prediction error of the tunnel deformation time series data inside the landslide at time tj, q is the sliding mean term order, is the jth sliding average coefficient, is the Zth type of landslide surface deformation data at time t among the multiple types of landslide surface deformation time series data, m is the number of types of landslide surface deformation historical data sets, is the model coefficient corresponding to the historical data set of landslide surface deformation described in the Zth category, is the prediction error of the tunnel deformation time series data inside the landslide at time t.

6. The method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion according to claim 5 is characterized in that: The training and verification of the multivariate time series analysis model to obtain the multivariate time series analysis optimization model comprises: Obtaining a value range of the autoregressive term order and the sliding average term order, and determining multiple groups of autoregressive term orders and sliding average term orders according to specific values ​​in the value range; According to each group of the autoregressive term order and the moving average term order, a corresponding autoregressive moving average model is established; Evaluate each of the autoregressive moving average models using the Bayesian Information Criterion and determine the corresponding BIC value; All determined BIC values ​​are compared, and the autoregressive term order and the sliding average term order under the autoregressive moving average model corresponding to the minimum BIC value are selected as the autoregressive term order and the sliding average term order in the multivariate time series analysis optimization model.

7. The method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion according to any one of claims 1 to 5, characterized in that: The training and verification of the multivariate time series analysis model to obtain the multivariate time series analysis optimization model comprises: The accuracy is verified by the root mean square error between the predicted value and the measured value of the tunnel deformation data; If the accuracy verification is passed, the corresponding multivariate time series analysis model is used as the multivariate time series analysis optimization model; If the accuracy verification fails, the multivariate time series analysis model is retrained until the accuracy verification passes, thereby obtaining the multivariate time series analysis optimization model.

8. The method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion according to claim 6 is characterized in that: The updating of the collected data and predicting the deformation data of the tunnel inside the landslide based on the updated collected data and the multivariate time series analysis optimization model includes: Determine a new batch of sample data according to the latest collected data; Each time a new batch of sample data is determined, the earliest batch of sample data is eliminated, and the multivariate time series analysis optimization model is trained and verified, thereby updating the multivariate time series analysis optimization model; The latest landslide surface deformation time series data set in the collected data is substituted into the updated multivariate time series analysis optimization model to output the latest landslide internal tunnel deformation data.

9. A landslide tunnel deformation prediction system based on multi-source monitoring data fusion, characterized in that: The method for predicting deformation of tunnels in landslides based on multi-source monitoring data fusion as described in any one of claims 1 to 8 comprises: An acquisition module, used for acquiring collected data, wherein the collected data includes a landslide surface deformation time series data set and a landslide internal tunnel deformation time series data set, wherein the landslide surface deformation time series data set includes a plurality of landslide surface deformation time series data; A screening module, for screening one or more of the landslide surface deformation time series data that meet preset correlation conditions with the landslide internal tunnel deformation time series data based on the landslide surface deformation time series data set; A training and verification module, used to use the screened landslide surface deformation time series data and the landslide internal tunnel deformation time series data as sample data, establish a multivariate time series analysis model for predicting the landslide internal tunnel deformation data based on the sample data, and train and verify the multivariate time series analysis model to obtain a multivariate time series analysis optimization model; An updating and prediction module is used to update the collected data and predict the deformation data of the tunnel inside the landslide based on the updated collected data and the multivariate time series analysis optimization model.

Images