A method and system for predicting ground settlement during step-method construction of subway tunnels

Abstract

The present application discloses a method and system for predicting surface settlement during step method construction of subway tunnels, the method comprising the following steps: obtaining original surface settlement sequence data; processing the original settlement data; performing VMD decomposition on the processed original surface settlement sequence data to obtain subsequences of different central modes and a residual, wherein each subsequence is arranged in order from low to high according to the central frequency; setting and optimizing the hyperparameters of the GRU neural network according to each subsequence and the residual; training the optimized GRU neural network, and then inputting each subsequence and the residual to obtain the predicted component of each subsequence and the predicted component of the residual; superimposing the results of each predicted component to obtain the predicted value of surface settlement. The present application has good prediction accuracy for surface settlement with strong volatility and randomness, and provides guidance for subsequent step method construction of subway tunnels.

Description

Technical Field

[0001] The present application relates to the technical field of surface settlement prediction, and in particular to a method and system for predicting surface settlement during step-method construction of a subway tunnel. Background Art

[0002] At present, my country's urban underground space projects are developing rapidly, and the surface subsidence caused by underground engineering construction has an increasingly significant impact on the surrounding environment. Therefore, the prediction of the surface subsidence law caused by urban tunnel construction has become an important issue in underground engineering.

[0003] At present, there are two main methods for predicting surface subsidence, physical model method and statistical model method. The physical model method is to analyze the physical factors that cause subsidence, establish a physical mechanics model to calculate the surface subsidence value for prediction, and the statistical model method only uses mathematical statistical analysis methods to analyze and interpret the internal relationship and development law of a large amount of historical monitoring data, and then predict the subsidence value. Commonly used methods include Peck method and random medium theory method. Some scholars also predict foundation subsidence based on the Verhulst model. In view of the uncertainty of surface subsidence prediction, the above methods are often not ideal in prediction accuracy. With the rise of neural networks, neural networks such as back propagation, generalized regression, long short-term memory and other neural networks have also shown extremely high fitting degree for complex data and are widely used. However, traditional neural networks have gradient explosion or disappearance problems, and their computational efficiency is low due to their complex structure.

[0004] The step method construction of subway tunnels is a long process with a long settlement period, many influencing factors, and a complex settlement curve with many fluctuation segments. If the complex original data is directly trained and predicted, problems such as slow convergence, poor fitting of data fluctuation segments, and prediction distortion are likely to occur. If the original data is smoothed or denoised, the information carried by the data is easily lost, which reduces the practicality of the prediction results.

[0005] At present, some people use empirical mode decomposition combined with traditional neural networks to predict time series. However, the modal aliasing problem of empirical mode decomposition causes the components to lose the characteristics of a single feature scale, making feature extraction, model training, and pattern recognition difficult, which leads to low accuracy of prediction results. Summary of the invention

[0006] The present application proposes a method and system for predicting surface settlement during step-method construction of subway tunnels. The method performs VMD decomposition on the original surface settlement data to obtain multiple subsequences and a residual. Each subsequence and the residual are then input into a trained GUR network to obtain a prediction component. Finally, the prediction component results are superimposed to obtain a surface settlement prediction value.

[0007] To achieve the above-mentioned purpose, the present application provides a method for predicting ground settlement during construction of a subway tunnel using the step method, comprising the following steps:

[0008] Obtain original surface subsidence sequence data;

[0009] Processing the original settlement data;

[0010] Perform VMD decomposition on the processed original surface settlement sequence data to obtain subsequences with different central modes and a residual, wherein the subsequences are arranged in order from low to high according to the central frequency;

[0011] Setting and optimizing the hyperparameters of the GRU neural network according to each of the subsequences and the residuals;

[0012] The optimized GRU neural network is trained, and then each of the subsequences and the residual is input to obtain the predicted component of each subsequence and the predicted component of the residual; the results of each of the predicted components are superimposed to obtain the predicted value of surface settlement.

[0013] Optionally, the method for obtaining the original surface settlement sequence data includes: obtaining the data by monitoring the vertical displacement changes of each measuring point on the ground around the subway based on the elevation control network.

[0014] Optionally, the method for processing the original settlement data includes: using the mean of the data at both ends to supplement the interval data in the original surface settlement sequence data, so that the frequency of the entire settlement data time series is 1 Hz.

[0015] Optionally, the method of performing VMD decomposition on the original surface settlement sequence data includes: trial decomposing the data according to K numbers respectively to obtain center frequency distribution diagrams corresponding to different K numbers, and obtaining the optimal number of intrinsic mode components K based on the center frequency distribution diagram; decomposing the original settlement data in combination with the optimal number of intrinsic mode components K to obtain K subsequences of different center modes and a residual.

[0016] Optionally, the GRU neural network includes an update gate, a reset gate and an output gate;

[0017] The GRU neural network merges the forget gate and input gate in the LSTM network into an update gate, retains the original reset gate, and learns the advantages of the LSTM gated network to update the cell state and hidden state; the update gate is used to describe the influence of the feature information of the past moment on the feature information of the present moment. The larger the threshold, the greater the influence of the feature information of the previous moment on the present moment; the reset gate is used to describe the degree to which the state feature information of the past moment is discarded. The smaller the threshold, the more the past information is discarded.

[0018] Optionally, the method of training each of the subsequences and the residuals by using a GRU neural network includes:

[0019] Forward propagation, performing weight calculation on each of the input subsequences in sequence propagation direction and outputting a target vector;

[0020] After comparison with the test set, the error is output through the objective function;

[0021] Back propagation changes the weights to extract time series features;

[0022] The forward propagation mathematical operation process is as follows:

[0023]

[0024] Where: h t-1 With h t is the hidden layer vector at different times, X t is the input data, U and w are weight parameters, is the gated activation function sigmoid, the output value is 0 to 1, tanh is used as the activation function when generating candidate memories, ⊙ is the Hadamard product symbol, z t 、r t 、h t ′ are update gate, reset gate and candidate state respectively.

[0025] The present application also provides a surface settlement prediction system for subway tunnel step method construction, including a data collection module, a data processing module, a decomposition module, an optimization module and a prediction module;

[0026] The data collection module is used to obtain original surface settlement sequence data;

[0027] The data processing module is used to process the original settlement data;

[0028] The decomposition module is used to perform VMD decomposition on the processed original surface settlement sequence data to obtain subsequences with different central modes and a residual, and each subsequence is arranged in order from low to high according to the central frequency;

[0029] The optimization module is used to set and optimize the hyperparameters of the GRU neural network according to the subsequence and the residual;

[0030] The prediction module is used to train the optimized GRU neural network, and then input each subsequence and the residual to obtain the prediction component of each subsequence and the prediction component of the residual; the prediction module is also used to superimpose the results of each prediction component to obtain the surface settlement prediction value.

[0031] Preferably, the workflow of the data processing module includes: using the mean of the data at both ends to supplement the interval data in the acquired original surface settlement sequence data, so that the frequency of the entire settlement data time series is 1 Hz.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] The present application obtains the prediction result by superimposing multiple prediction components obtained by the GUR network, which can effectively improve the prediction accuracy compared with the traditional surface settlement prediction method; it also effectively solves the problem of mode aliasing in empirical mode decomposition causing the component to lose its single characteristic scale. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solution of the present application, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0035] Figure 1 Schematic diagram of the technical process of the method in this embodiment;

[0036] Figure 2 This is a schematic diagram of the measured settlement curve in this embodiment;

[0037] Figure 3 This is a schematic diagram of VMD decomposition results in this embodiment;

[0038] Figure 4 This is a schematic diagram comparing the predicted sedimentation curves of the GRU and VMD-GRU models in this embodiment;

[0039] Figure 5 This is a structural schematic diagram of a surface settlement prediction system for step-method construction of subway tunnels. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0041] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0042] Embodiment 1

[0043] like Figure 1 The figure shows a schematic diagram of the technical process of the method of this embodiment, taking the surface settlement data of the step method construction of the west section of Line 11 in Beijing as an example. The surrounding rock of the tunnel is mainly clay and clayey silt, with a lot of interlayer groundwater, complex geological conditions, and variable surface settlement patterns.

[0044] The data were collected from point DB1-17 of section 17, and the monitoring period was from June 4, 2020 to February 11, 2021. For the interval data, the mean of the data at both ends was used to make up for it, so that the frequency of the entire settlement data time series was 1Hz, with a total of 248 data settlement curves as follows Figure 2 shown.

[0045] When the number of variational modes is 6, the smaller the overlap of the center frequencies, the more thorough the time series decomposition is, and the optimal number of modes is obtained. The surface settlement time series is decomposed by VMD to obtain 6 subsequences with different center modes and a residual. VMD decomposition methods include:

[0046] S1. Assume that the decomposed mode function u k (t), and then transform it into a one-sided spectrum:

[0047]

[0048] Where δ(t) is the unit impulse function, δ(t) is infinite when t is equal to 0, δ(t) is 0 when t is not equal to , and j is an imaginary unit;

[0049] S2. Then compare the spectrum with the center frequency ω k Exponential term e -j ω k t Aliasing causes it to be translated to baseband, and the expression of the analytical signal is as follows:

[0050]

[0051] S3. To ensure that when the original sequence is decomposed into K modal components, the sum of the estimated bandwidths of the modal components with the center frequency is the smallest, the sum of the modal components should be equal to the original sequence, and a constrained variational problem is obtained, which is expressed as:

[0052]

[0053] Where: u k are the K mode functions after decomposition, ω k are the center frequencies corresponding to the K mode functions, To find the partial derivative with respect to t, |||| is the norm, * is the convolution operation, and f(t) is the original time series signal;

[0054] S4. The penalty factor α and the Lagrange multiplication operator λ are introduced to transform the constrained variational problem into an unconstrained variational problem for solution. The corresponding equation is:

[0055]

[0056] Where ﹤﹥ is the dot product operation of two vectors. Then the alternating direction multiplier iteration algorithm is used to solve the unconstrained variational problem, and the results show that λ and u k ,ω k Iterative formula in the corresponding frequency domain;

[0057] After S5, enter the initial u k (1) ,ω k (1) , (1) The Fourier transform form of is iterated to obtain equations (1) and (2);

[0058]

[0059]

[0060] Repeat the iterations of (1) and (2) until the following equation is satisfied:

[0061]

[0062] In the formula, the upper right subscripts (n) and (n+1) represent the number of iterations, the superscript ^ represents the Fourier transform, the lower right subscript k represents the kth decomposition mode, ε is the allowable error, τ is the noise tolerance, and both ε and τ are constants greater than zero; the decomposed components are arranged in order from low to high according to the center frequency, and then the total data is subtracted from each modal component to obtain the residual err.

[0063] The final settlement data components are as follows Figure 3 As shown, component 1 is the main component, which is basically consistent with the trend and size of the total settlement data. It can be seen that it has a more obvious "three-stage" feature, which is strongly correlated with the double-step step method excavation process on site. The remaining components are smaller in value and should be composed of other minor factors and errors that affect settlement. In terms of data, they mainly affect the short-term construction settlement early warning judgment.

[0064] The GRU neural network uses the Adam optimizer with a learning rate set to 0.001. The memory has two layers with 80 and 100 respectively. The dropout rate is set to 0.2, the batch size is set to 64, and the first 78 periods of sedimentation monitoring data are selected as the training set, and the last 170 periods of data are selected as the test set. MSE is also used as the loss function, and MAE and root mean square error are used as auxiliary observation parameters.

[0065] The components obtained by variational mode decomposition are arranged from large to small and input into the GRU neural network in sequence. The GRU neural network trains each component and the residual, performs single-step prediction superposition to generate 238 single-step prediction data, and then generates a data group with 9 measured data plus one prediction data. Similarly, 238 double-step prediction data are superimposed and repeated until the seventh-step prediction data is generated, that is, one week of prediction data, to provide a longer-term prediction for construction.

[0066] Figure 4 The single-step prediction effects of the GRU prediction model and the VMD-GRU prediction model at the DB1-17 point and the comparison of the actual settlement values ​​are shown. It can be found that the combined model fits the nonlinear data significantly better than a single neural network, and the data predicted by the combined model is more "rough" and more consistent with the measured data.

[0067] Table 1 shows the prediction value accuracy evaluation indicators of each prediction model at DB1-17 points. It can be seen that whether it is single-step prediction or multi-step prediction, the MSE, RMSE, and MAE of the combined model are better than those of a single neural network, among which the VMD-GRU combined model has the best accuracy evaluation indicators.

[0068] Table 1

[0069]

[0070] The above shows the principle of the present application and verifies the effectiveness of the present application. The trend of the principal component obtained by variational mode decomposition is basically similar to the trend and size of the total settlement data under the step method construction. The VMD-GRU combined model has good accuracy. When the predicted value exceeds the standard, it can issue an early warning to guide subsequent construction, and has good application prospects.

[0071] Embodiment 2

[0072] like Figure 5 As shown, it is a structural schematic diagram of the surface settlement prediction system for subway tunnel step method construction in this embodiment, which includes a data collection module, a data processing module, a decomposition module, an optimization module and a prediction module.

[0073] The data collection module is used to collect the original surface settlement sequence. Taking the surface settlement data of the step method construction of the west section of Beijing Line 11 subway tunnel as an example, the surrounding rock of the tunnel is mainly clay and clayey silt, with a lot of interlayer groundwater, complex geological conditions, and variable surface settlement patterns.

[0074] The data was collected from point DB1-17 of section 17, and the monitoring period was from June 4, 2020 to February 11, 2021. For the interval data, the data processing module uses the mean of the data at both ends to make it complete, so that the frequency of the entire settlement data time series is 1Hz, with a total of 248 data settlement curves as follows Figure 2 shown.

[0075] At this time, the decomposition module knows that when the number of intrinsic mode components is 6, the smaller the overlap of the center frequency, the more thorough the time series decomposition is, and the optimal number of modes is obtained. The surface settlement time series is decomposed by VMD through the decomposition module to obtain 6 subsequences of different center modes and a residual. VMD decomposition methods include:

[0076] S1. Assume that the decomposed mode function u k (t), and then transform it into a one-sided spectrum:

[0077]

[0078] Where δ(t) is the unit impulse function, δ(t) is infinite when t is equal to 0, δ(t) is 0 when t is not equal to , and j is an imaginary unit;

[0079] S2. Then compare the spectrum with the center frequency ω k Exponential term e -jω k t Aliasing causes it to be translated to baseband, and the expression of the analytical signal is as follows:

[0080]

[0081] S3. To ensure that when the original sequence is decomposed into K modal components, the sum of the estimated bandwidths of the modal components with the center frequency is the smallest, the sum of the modal components should be equal to the original sequence, and a constrained variational problem is obtained, which is expressed as:

[0082]

[0083] Where: u k are the K mode functions after decomposition, ω k are the center frequencies corresponding to the K mode functions, To find the partial derivative with respect to t, |||| is the norm, * is the convolution operation, and f(t) is the original time series signal;

[0084] S4. The penalty factor α and the Lagrange multiplication operator λ are introduced to transform the constrained variational problem into an unconstrained variational problem for solution. The corresponding equation is:

[0085]

[0086] Where ﹤﹥ is the dot product operation of two vectors. Then the alternating direction multiplier iteration algorithm is used to solve the unconstrained variational problem, and the results show that λ and u k ,ω k Iterative formula in the corresponding frequency domain;

[0087] After S5, enter the initial u k (1) ,ω k (1) , (1) The Fourier transform form of is iterated to obtain equations (1) and (2);

[0088]

[0089]

[0090] Repeat the iterations of (1) and (2) until the following equation is satisfied:

[0091]

[0092] In the formula, the upper right subscripts (n) and (n+1) represent the number of iterations, the superscript ^ represents the Fourier transform, the lower right subscript k represents the kth decomposition mode, ε is the allowable error, τ is the noise tolerance, and both ε and τ are constants greater than zero; the decomposed components are arranged in order from low to high according to the center frequency, and then the total data is subtracted from each modal component to obtain the residual err.

[0093] The final settlement data components are as follows Figure 3 As shown, component 1 is the main component, which is basically consistent with the trend and size of the total settlement data. It can be seen that it has a more obvious "three-stage" feature, which is strongly correlated with the double-step step method excavation process on site. The remaining components are smaller in value and should be composed of other minor factors and errors that affect settlement. In terms of data, they mainly affect the short-term construction settlement early warning judgment.

[0094] The optimization module controls the GRU neural network to use the Adam optimizer with a learning rate set to 0.001. The memory has two layers with 80 and 100 respectively, the dropout rate is set to 0.2, the batch size is set to 64, and the first 78 periods of sedimentation monitoring data are selected as the training set, and the last 170 periods of data are selected as the test set. The optimization module also uses MSE as the loss function, and MAE and root mean square error as auxiliary observation parameters.

[0095] The components obtained by variational modal decomposition are arranged from large to small and input into the prediction module in sequence to train each component and residual. Single-step prediction is superimposed to generate 238 single-step prediction data. Then, 9 measured data plus one prediction data are used to generate a data group. Similarly, 238 double-step prediction data are superimposed and repeated until the seventh-step prediction data is generated, that is, one week of prediction data, to provide a longer-term prediction for construction.

[0096] Figure 4 The comparison between the single-step prediction effect of the GRU prediction model and the VMD-GRU prediction model at point DB1-17 and the actual settlement value shows that the combined model fits the nonlinear data significantly better than a single neural network, and the data predicted by the combined model is more "rough" and more consistent with the measured data.

[0097] Table 1 shows the prediction value accuracy evaluation indicators of each prediction model at DB1-17 points. It can be seen that whether it is single-step prediction or multi-step prediction, the MSE, RMSE, and MAE of the combined model are better than those of a single neural network, among which the VMD-GRU combined model has the best accuracy evaluation indicators.

[0098] Table 1

[0099]

[0100] The above shows the principle of the present application and verifies the effectiveness of the present application. The trend of the principal component obtained by variational mode decomposition is basically similar to the trend and size of the total settlement data under the step method construction. The VMD-GRU combined model has good accuracy. When the predicted value exceeds the standard, it can issue an early warning to guide subsequent construction, and has good application prospects.

[0101] The embodiments described above are only descriptions of the preferred methods of the present application, and are not intended to limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements made to the technical solutions of the present application by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present application.

Claims

1. A method for predicting ground settlement during construction of a subway tunnel using the step method, characterized in that: The following steps are involved: Obtain original surface subsidence sequence data; Processing the original settlement data; Perform VMD decomposition on the processed original surface settlement sequence data to obtain subsequences with different central modes and a residual, wherein the subsequences are arranged in order from low to high according to the central frequency; Setting and optimizing the hyperparameters of the GRU neural network according to each of the subsequences and the residuals; Training the optimized GRU neural network, and then inputting each of the subsequences and the residual to obtain a predicted component of each of the subsequences and a predicted component of the residual; Superimposing the prediction results of each of the prediction components to obtain a predicted value of surface subsidence; The method for training each of the subsequences and the residuals by a GRU neural network includes: Forward propagation, performing weight calculation on each of the input subsequences in sequence propagation direction and outputting a target vector; After comparison with the test set, the error is output through the objective function; Back propagation changes the weights to extract time series features; The forward propagation mathematical operation process is as follows: Where: h t-1 With h t is the hidden layer vector at different times, X t is the input data, U and w are weight parameters, is the gated activation function sigmoid, the output value is 0 to 1, tanh is used as the activation function when generating candidate memories, ⊙ is the Hadamard product symbol, z t 、r t 、h t ′ are update gate, reset gate and candidate state respectively; The GRU neural network uses the Adam optimizer with a learning rate set to 0.

001. The memory has two layers with 80 and 100 respectively, the dropout rate is set to 0.2, the batch size is set to 64, and the first 78 periods of sedimentation monitoring data are selected as the training set, and the last 170 periods of data are selected as the test set. MSE is also used as the loss function, and MAE and root mean square error are used as auxiliary observation parameters. The components obtained by variational mode decomposition are arranged from large to small and input into the GRU neural network in sequence to train each component and residual, and single-step prediction is superimposed to generate 238 single-step prediction data. Then, a data group is generated with 9 measured data plus one prediction data, and 238 double-step prediction data are also superimposed.

2. The method for predicting ground settlement during construction of a subway tunnel using the step method according to claim 1, characterized in that: The method for obtaining the original surface settlement sequence data includes: obtaining the data by monitoring the vertical displacement changes of each measuring point on the ground around the subway on the basis of the elevation control network.

3. The method for predicting ground settlement during construction of a subway tunnel using the step method according to claim 1, characterized in that: The method for processing the original settlement data includes: using the mean of the data at both ends to supplement the interval data in the original surface settlement sequence data, so that the frequency of the entire settlement data time series is 1 Hz.

4. The method for predicting ground settlement during construction of a subway tunnel using the step method according to claim 1, characterized in that: The method for performing VMD decomposition on the original surface settlement sequence data includes: trial decomposing the data according to K numbers respectively to obtain center frequency distribution diagrams corresponding to different K numbers, and obtaining the optimal number of intrinsic mode components K based on the center frequency distribution diagram; decomposing the original settlement data in combination with the optimal number of intrinsic mode components K to obtain K subsequences with different center modes and a residual.

5. The method for predicting ground settlement during construction of a subway tunnel using the step method according to claim 1, characterized in that: The GRU neural network includes an update gate, a reset gate and an output gate; The GRU neural network merges the forget gate and input gate in the LSTM network into an update gate, retains the original reset gate, and learns the advantages of the LSTM gated network to update the cell state and hidden state; the update gate is used to describe the influence of the feature information of the past moment on the feature information of the present moment. The larger the threshold, the greater the influence of the feature information of the previous moment on the present moment; the reset gate is used to describe the degree to which the state feature information of the past moment is discarded. The smaller the threshold, the more the past information is discarded.

6. A surface settlement prediction system for subway tunnel step method construction, characterized in that: It includes data collection module, data processing module, decomposition module, optimization module and prediction module; The data collection module is used to obtain original surface settlement sequence data; The data processing module is used to process the original settlement data; The decomposition module is used to perform VMD decomposition on the processed original surface settlement sequence data to obtain subsequences with different central modes and a residual, and each subsequence is arranged in order from low to high according to the central frequency; The optimization module is used to set and optimize the hyperparameters of the GRU neural network according to the subsequence and the residual; The prediction module is used to train the optimized GRU neural network, and then input each of the subsequences and the residual to obtain the prediction component of each of the subsequences and the prediction component of the residual; the prediction module is also used to superimpose the results of each of the prediction components to obtain the predicted value of surface settlement; The method for training each of the subsequences and the residuals by a GRU neural network includes: Forward propagation, performing weight calculation on each of the input subsequences in sequence propagation direction and outputting a target vector; After comparison with the test set, the error is output through the objective function; Back propagation changes the weights to extract time series features; The forward propagation mathematical operation process is as follows: Where: h t-1 With h t is the hidden layer vector at different times, X t is the input data, U and w are weight parameters, is the gated activation function sigmoid, the output value is 0 to 1, tanh is used as the activation function when generating candidate memories, ⊙ is the Hadamard product symbol, z t 、r t 、h t ′ are update gate, reset gate and candidate state respectively; The GRU neural network uses the Adam optimizer with a learning rate set to 0.

001. The memory has two layers with 80 and 100 respectively, the dropout rate is set to 0.2, the batch size is set to 64, and the first 78 periods of sedimentation monitoring data are selected as the training set, and the last 170 periods of data are selected as the test set. The optimization module also uses MSE as the loss function, MAE and root mean square error as auxiliary observation parameters. The components obtained by variational mode decomposition are arranged from large to small and input into the prediction module in sequence to train each component and residual, and single-step prediction is superimposed to generate 238 single-step prediction data. Then, a data group is generated by adding 9 measured data to one predicted data, and 238 double-step prediction data are also superimposed.

7. A surface settlement prediction system for subway tunnel step method construction according to claim 6, characterized in that: The workflow of the data processing module includes: using the mean of the data at both ends to supplement the interval data in the original surface settlement sequence data, so that the frequency of the entire settlement data time series is 1 Hz.

Images