A method and device for predicting deformation of tunnel surrounding rock

Abstract

The present invention provides a method and device for predicting tunnel surrounding rock deformation. This solution uses a sparrow search algorithm to solve the optimal combination of VMD parameters τ and σ, avoiding the tedious manual parameter search and the subjective influence of empirical parameter setting, thereby improving the decomposition quality of the surrounding rock deformation sequence. The GRU model prediction method can fully mine the effective information in the surrounding rock deformation history data, and can better reflect the dynamic system nature of the surrounding rock deformation evolution compared with static models. The method has superior prediction accuracy and engineering practicality.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel surrounding rock deformation, and in particular to a tunnel surrounding rock deformation prediction method and device. Background Art

[0002] Excavating a tunnel through complex rock mass disrupts its initial equilibrium, altering its stress field and causing stress redistribution. During this process, the surrounding rock inevitably deforms. Excessive deformation can easily lead to major safety incidents such as instability and collapse, seriously endangering the lives and property of construction workers. Therefore, conducting research on surrounding rock deformation prediction and establishing accurate deformation prediction models are extremely urgent and important for timely understanding the evolution of surrounding rock deformation, ensuring construction safety, and reducing economic losses.

[0003] Decomposing surrounding rock deformation into components with practical physical significance based on time series analysis and predicting them separately is a very effective method. Classic time series decomposition methods include wavelet analysis (WA), empirical mode decomposition (EMD), variational mode decomposition (VMD), etc. Among them, VMD has a unique advantage. It can specify the number of components after decomposition, overcomes the shortcomings of WA's lack of adaptability, and to a certain extent alleviates problems such as EMD modal aliasing and endpoint effects. However, the effect of VMD decomposition depends on the selection of parameters. In the field of surrounding rock deformation research, scholars often obtain VMD parameters through continuous trial and error. This method is labor-intensive and it is difficult to obtain the optimal combination value, resulting in poor decomposition effect.

[0004] On the other hand, with the rapid development of artificial intelligence technology, a large number of researchers have introduced models such as SVM and BP neural networks into surrounding rock deformation prediction. However, these commonly used models are static models, whose output depends only on the current input, lack memory function, and cannot effectively adapt to the continuous changes of new samples. Obviously, for surrounding rock deformation sequences, these models ignore the dynamic system nature of surrounding rock deformation evolution to a certain extent, fail to fully explore its effective information, and thus restrict the improvement of prediction accuracy.

[0005] Therefore, in order to overcome the above shortcomings of the prior art, it is urgent to provide a solution. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention proposes a tunnel surrounding rock deformation prediction method to solve the technical problems existing in the existing technology, such as large workload and difficulty in obtaining the optimal combination value, resulting in poor decomposition effect, and ignoring the dynamic system nature of the surrounding rock deformation evolution, and failing to fully explore its effective information, which restricts the improvement of prediction accuracy.

[0007] A tunnel surrounding rock deformation prediction method, characterized by comprising: collecting monitoring data of surrounding rock deformation and determining whether the monitoring data are equally spaced; if so, dividing the monitoring data into a training set and a prediction set; defining a VMD decomposition evaluation criterion and solving the VMD parameters based on a sparrow search algorithm. and parameters The optimal combination of and parameters The optimal combination is used in VMD to obtain a target VMD; the target VMD is used to decompose the training set and the prediction set in the monitoring data into trend item displacement and random item displacement respectively; a GRU model is established, and the GRU model is trained with the training set to obtain a target GRU model; displacement prediction is performed according to the target GRU model, and the prediction result is denormalized to obtain a target displacement prediction result, wherein the target displacement prediction result includes a trend item displacement prediction result and a random item displacement prediction result; based on the target displacement prediction result, a predicted value of surrounding rock deformation is obtained by superposition calculation.

[0008] In one embodiment, after collecting monitoring data of surrounding rock deformation and determining whether the monitoring data is equidistant, the method further includes: if not, using a cubic spline interpolation method to perform equidistant processing, and dividing the processed monitoring data into a training set and a prediction set.

[0009] In one embodiment, the step of defining a VMD decomposition evaluation criterion includes: selecting a root mean square error as an indicator for measuring the completeness of information after decomposition; introducing sample entropy as an indicator for evaluating the decomposition effect; and constructing a VMD evaluation criterion by combining the root mean square error and the sample entropy.

[0010] In one embodiment, the parameters of VMD are solved based on the sparrow search algorithm. and parameters The optimal combination steps include: converting the evaluation criteria into a constrained optimization problem; using a sparrow search algorithm to solve the constrained optimization problem, and obtaining parameters and parameters The optimal combination.

[0011] In one embodiment, after the step of using the target VMD to decompose the training set and the prediction set in the monitoring data into trend item displacement and random item displacement respectively, the method further includes: normalizing the monitoring data decomposed into trend item displacement and random item displacement.

[0012] In one embodiment, the steps of establishing a GRU model and training the GRU model using a training set to obtain a target GRU model include: establishing a first GRU model, training the first GRU model using a training set in a trend item displacement, and obtaining a first target GRU model; establishing a second GRU model, training the second GRU model using a training set in a random item displacement, and obtaining a second target GRU model.

[0013] In one embodiment, displacement prediction is performed according to the target GRU model, and the prediction result is denormalized to obtain a target displacement prediction result, wherein the target displacement prediction result includes a trend item displacement prediction result and a random item displacement prediction result. The steps include: predicting the trend item displacement according to the first target GRU model, and performing a denormalization operation on the predicted trend item displacement to obtain the trend item displacement prediction result; predicting the random item displacement according to the second target GRU model, and performing a denormalization operation on the predicted random item displacement to obtain the random item displacement prediction result.

[0014] In one embodiment, the step of superimposing and calculating the surrounding rock deformation prediction value based on the target displacement prediction result includes: accumulating the trend item displacement prediction result and the random item displacement prediction result based on the time series superposition principle to calculate the surrounding rock deformation prediction value.

[0015] In one embodiment, after the step of superimposing and calculating the surrounding rock deformation prediction value based on the target displacement prediction result, the method further includes: obtaining the actual value of the surrounding rock deformation, and calculating the target determination coefficient, target mean absolute error and target root mean square error based on the surrounding rock deformation prediction value; and using the target determination coefficient, the target mean absolute error and the target root mean square error to quantitatively evaluate the surrounding rock deformation prediction value.

[0016] A tunnel surrounding rock deformation prediction device, characterized by comprising: a data acquisition and judgment module, a VMD decomposition module, a GRU prediction processing module and a prediction value calculation module, wherein the data acquisition and judgment module is used to collect monitoring data of surrounding rock deformation and determine whether the monitoring data is equidistant; if so, the monitoring data is divided into a training set and a prediction set; the VMD decomposition module is used to define a VMD decomposition evaluation standard and solve the VMD parameters based on a sparrow search algorithm and parameters The optimal combination of and parameters The optimal combination is used in VMD to obtain a target VMD; the target VMD is used to decompose the training set and the prediction set in the monitoring data into trend item displacement and random item displacement respectively; the GRU prediction processing module is used to establish a GRU model and train the GRU model with the training set to obtain a target GRU model; displacement prediction is performed according to the target GRU model, and the prediction result is denormalized to obtain a target displacement prediction result, and the target displacement prediction result includes a trend item displacement prediction result and a random item displacement prediction result; the calculation prediction value module is used to obtain a surrounding rock deformation prediction value by superimposing calculation based on the target displacement prediction result.

[0017] It can be seen from the above technical solution that the beneficial technical effects of the present invention are as follows:

[0018] 1. This method is based on the sparrow search algorithm to solve the parameters of VMD and parameters The optimal combination of the two models avoids the tedious manual search for parameters and the subjective influence of experience-based parameter setting, and improves the decomposition quality of the surrounding rock deformation sequence. The GRU model prediction method can fully mine the effective information in the surrounding rock deformation history data, and can better reflect the dynamic system nature in the surrounding rock deformation evolution than the static model. It has better prediction accuracy and engineering practicality.

[0019] 2. The non-uniformly spaced monitoring data are processed into equally spaced ones through the cubic spline interpolation method to ensure the rationality of the analyzed monitoring data, thereby preventing errors in subsequent analysis and prediction due to unreasonable monitoring data, and further improving the accuracy of the prediction.

[0020] 3. Quantitatively evaluate the prediction results using the target coefficient of determination, target mean absolute error, and root mean square error. The target coefficient of determination is set between 0 and 1, with larger values ​​indicating a higher degree of consistency between the predicted and actual values. The target mean absolute error reflects the actual level of error in the predicted value, while the target root mean square error is extremely sensitive to large errors and reflects the precision of the prediction. The smaller the target coefficient of determination, the better. This significantly improves the comprehensiveness and accuracy of the quantitative evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly describes the drawings required for the specific embodiments or the description of the prior art. Similar elements or parts are generally identified by similar reference numerals throughout the drawings. Elements or parts in the drawings are not necessarily drawn to scale.

[0022] Figure 1 A schematic flow chart of a method for predicting tunnel surrounding rock deformation in one embodiment;

[0023] Figure 2 for Figure 1 A schematic diagram of the overall process of a tunnel surrounding rock deformation prediction method;

[0024] Figure 3 A schematic diagram of isometric data of vault settlement in one embodiment;

[0025] Figure 4 A schematic diagram of an optimization process for decomposing a deformation sequence using different algorithms in one embodiment;

[0026] Figure 5 A schematic diagram of the decomposition results of the vault settlement sequence in one embodiment;

[0027] Figure 6 A scatter plot showing the prediction of each component of a vault settlement sequence in one embodiment;

[0028] Figure 7 A schematic diagram of an overall prediction result of vault settlement in one embodiment;

[0029] Figure 8 A schematic diagram of the predicted results of vault settlement of different models in one embodiment;

[0030] Figure 9 The figure is a structural block diagram of a tunnel surrounding rock deformation prediction device in one embodiment. DETAILED DESCRIPTION

[0031] The following embodiments of the technical solution of the present invention will be described in detail with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention and are therefore only examples and are not intended to limit the scope of protection of the present invention.

[0032] It should be noted that, unless otherwise specified, the technical or scientific terms used in this application should have the common meanings understood by those skilled in the art to which the present invention belongs.

[0033] In one embodiment, Figure 1 and Figure 2 As shown, a method for predicting tunnel surrounding rock deformation is provided, comprising the following steps:

[0034] S1 collects monitoring data of surrounding rock deformation and determines whether the monitoring data are equidistant.

[0035] Specifically, in actual projects, the collected surrounding rock deformation monitoring data is not uniformly spaced, which makes it unreasonable to use it directly for analysis. Therefore, after obtaining the monitoring data, the first thing to do is to determine whether it is uniformly spaced.

[0036] In one embodiment, after step S1, the method further includes the following steps: if not, using the cubic spline difference method to perform equal spacing processing, and dividing the processed monitoring data into a training set and a prediction set.

[0037] Specifically, if the monitoring data is not equidistant, the cubic spline interpolation method is first used to equalize the intervals before the training and prediction sets are divided. Cubic spline interpolation (also known as spline interpolation) is a process that mathematically solves the three bending moment equations to obtain a set of curve functions by passing a smooth curve through a series of shape points.

[0038] If S2 is true, the monitoring data is divided into a training set and a prediction set.

[0039] Specifically, if the monitoring data is equally spaced, it is divided into a training set and a prediction set. The training set is used to train the model, and the prediction set is used to verify the generalization ability and applicability of the model.

[0040] S3 defines the decomposition evaluation criteria of VMD and solves the parameters of VMD based on the sparrow search algorithm and parameters The optimal combination.

[0041] In one embodiment, the step of defining the VMD decomposition evaluation criterion in step S3 includes: selecting the root mean square error as an indicator for measuring the completeness of information after decomposition; introducing sample entropy as an indicator for evaluating the decomposition effect; and constructing the VMD evaluation criterion by combining the root mean square error and sample entropy.

[0042] Specifically, VMD, or variational mode decomposition, is a completely non-recursive, adaptive signal decomposition method that can decompose the original sequence into a specified number of intrinsic mode components based on the time scale characteristics of the surrounding rock deformation sequence itself. , rising step length and the quadratic penalty factor The impact on the decomposition effect of the VMD algorithm is more significant. To preset the number of decomposed intrinsic mode components (IMFs), it is taken as 2 in the decomposition of surrounding rock deformation sequence; and It affects the loss degree and decomposition effect after decomposition, but its selection usually adopts empirical settings or continuous trial calculations. The empirical settings are highly random and lack theoretical basis. For complex surrounding rock deformation sequences, the continuous trial calculation method is not only computationally intensive and time-consuming, but also difficult to obtain the optimal combination value of the two parameters. Therefore, it is urgent to define an evaluation standard applicable to the field of surrounding rock deformation prediction to guide the selection of VMD parameters. So first define the VMD decomposition evaluation standard, and The two parameters affect each other. In order to fully consider their impact on the decomposition quality, the overall idea of ​​defining the evaluation criteria is as follows:

[0043] (1) In order to consider the loss degree after decomposition, the two eigenmode components obtained by decomposition are accumulated and reconstructed as , and calculate its difference with the original sequence The root mean square error RMSE is used as an indicator to measure the completeness of the decomposed information, that is:

[0044] (1)

[0045] Where: is the original value at time t, is the reconstruction value at time t, is the sequence length.

[0046] (2) To consider the decomposition effect, sample entropy (SE) is introduced as an indicator to evaluate the decomposition effect. Sample entropy is an algorithm that measures the regularity and complexity of time series. The larger the value, the higher the complexity, and vice versa.

[0047] (2)

[0048] Where: m is the dimension of the reconstructed sequence, which is 2; similarity tolerance , std is The standard deviation of It is the surrounding rock deformation sequence.

[0049] Among the two components decomposed from the surrounding rock deformation sequence, the displacement sequence of the trend term ( ) is to remove the deformation components caused by sudden factors and monitoring conditions, so the lower its complexity, the better. Therefore, this paper uses the sample entropy value of the trend term displacement sequence as an indicator to measure the decomposition effect.

[0050] (3) Combining RMSE with SE can not only ensure the integrity of the decomposed sequence information, but also reflect the decomposition effect of the original sequence. In summary, the evaluation criteria are constructed. for:

[0051] (3)

[0052] Where: To reconstruct the sequence and the original sequence The smaller the root mean square error, the higher the information integrity after decomposition; is the sample entropy value of the trend item displacement sequence after decomposition. The smaller the value, the more regular the trend item displacement sequence is and the better the decomposition effect is.

[0053] In one embodiment, the parameters of VMD are solved based on the sparrow search algorithm. and parameters The optimal combination steps include: transforming the evaluation criteria into a constrained optimization problem; using the sparrow search algorithm to solve the constrained optimization problem and obtain the parameters and parameters The optimal combination.

[0054] Specifically, and Parameter solution can convert formula (3) into a constrained optimization problem, that is, to find the parameters within a certain interval. and The value of Minimum, that is:

[0055]

[0056] (4)

[0057] The Sparrow Search Algorithm (SSA) is used to solve the above constrained optimization problem. SSA is a new meta-heuristic algorithm with good global optimization capabilities. Compared with the Grey Wolf Optimization Algorithm (GWO) and the Whale Optimization Algorithm (WOA) in terms of stability and accuracy, it has obvious advantages. The specific steps are as follows:

[0058] Input training set samples and initialize the sparrow search algorithm parameters;

[0059] Each sparrow individual represents the VMD algorithm and Two parameters. Use formula (3) as the fitness function of the sparrow search algorithm, calculate the fitness values ​​of all sparrows, find the current global optimal solution and determine its position;

[0060] Update the positions of the discoverer, joiner, and alerter, check the individual fitness values ​​after the position update, and determine the position of the best sparrow individual;

[0061] Determine whether the termination criteria are met, that is, whether the global optimal solution is obtained or the maximum number of iterations is met. If so, the optimal sparrow position is output as a parameter and Otherwise, continue to execute step 、 .

[0062] S4 will get the parameters and parameters The optimal combination of is used in VMD to obtain the target VMD.

[0063] Specifically, the obtained and The best combination value is used in the VMD algorithm to obtain the target VMD algorithm.

[0064] S5 uses target VMD to decompose the training set and prediction set in the monitoring data into trend term displacement and random term displacement, respectively.

[0065] Specifically, the training set is decomposed into a training set of trend item displacement and a training set of random item displacement; the prediction set is decomposed into a prediction set of trend item displacement and a prediction set of random item displacement.

[0066] In one embodiment, after step S5, the method further includes normalizing the monitoring data of the decomposed trend term displacement and random term displacement. Specifically, to facilitate convergence of the prediction algorithm, the data are normalized to be between [0, 1].

[0067] S6 establishes a GRU model and uses the training set to train the GRU model to obtain the target GRU model.

[0068] In one embodiment, step S6 also includes: establishing a first GRU model, training the first GRU model using the training set in the trend item displacement to obtain a first target GRU model; establishing a second GRU model, training the second GRU model using the training set in the random item displacement to obtain a second target GRU model.

[0069] Specifically, the GRU dynamic neural network is trained using the training set from the trend term shift to obtain the first target GRU model; the GRU dynamic neural network is trained using the training set from the random term shift to obtain the second target GRU model. The GRU (Gated Recurrent Unit) is a highly effective variant of the LSTM network. Its structure is simpler than the LSTM network, yet it also performs well, making it a very popular network currently.

[0070] S7 performs displacement prediction according to the target GRU model and performs denormalization processing on the prediction result to obtain a target displacement prediction result. The target displacement prediction result includes a trend item displacement prediction result and a random item displacement prediction result.

[0071] In one embodiment, step S7 also includes: predicting the trend item displacement according to the first target GRU model, and performing a denormalization operation on the predicted trend item displacement to obtain a trend item displacement prediction result; predicting the random item displacement according to the second target GRU model, and performing a denormalization operation on the predicted random item displacement to obtain a random item displacement prediction result.

[0072] Specifically, the first target GRU network after training is used to predict the displacement of the trend item, and an inverse normalization operation is performed to obtain the trend item displacement prediction result. The second target GRU network after training is used to predict the displacement of the random item, and an inverse normalization operation is performed to obtain the random item displacement prediction result.

[0073] S8 obtains the predicted value of surrounding rock deformation through superposition calculation based on the target displacement prediction results.

[0074] In one embodiment, step S8 includes: accumulating the trend item displacement prediction result and the random item displacement prediction result based on the time series superposition principle to calculate the surrounding rock deformation prediction value.

[0075] In one embodiment, after step S8, the method further includes: obtaining the actual value of surrounding rock deformation, and calculating the target determination coefficient, target mean absolute error, and target root mean square error based on the predicted value of surrounding rock deformation; and quantitatively evaluating the predicted value of surrounding rock deformation using the target determination coefficient, target mean absolute error, and target root mean square error.

[0076] Specifically, the coefficient of determination , mean absolute error and root mean square error Conduct quantitative evaluation of the prediction results. The value is (0,1), and the larger the value, the higher the consistency between the predicted value and the actual value; Can reflect the actual situation of the prediction value error, It is extremely sensitive to large errors and can reflect the precision of the prediction. The smaller the two indicators are, the better. The calculation formula is as follows:

[0077] (5)

[0078] (6)

[0079] (7)

[0080] Where: is the predicted value of surrounding rock deformation, is the actual value of surrounding rock deformation, is the actual value of surrounding rock deformation, is the number of data groups.

[0081] In one embodiment, the results obtained by using the above tunnel surrounding rock deformation prediction method are verified. First, Figure 3 As shown in the figure, the 60-day equidistant data of the arch settlement after processing are shown. Then the surrounding rock deformation sequence is decomposed. In order to verify the effectiveness and superiority of the sparrow search algorithm (SSA) to optimize VMD, the empirical setting value, the whale optimization algorithm (WOA) (an existing technology) optimization value and the SSA optimization value are used as the decomposition parameters of VMD, respectively, to decompose the arch settlement training set sequence into corresponding modal components. For fairness, the population parameters of the WOA and SSA algorithms are kept consistent: the maximum number of iterations is 40, the population size is 60, and the optimization range is [1,10000], [0,1]. In addition, the algorithm parameters are set as follows: the proportion of SSA discoverers is 0.7, the proportion of warnings is 0.2, and the safety value is 0.6; the WOA spiral shape parameter is 1. Formula (3) is used as the fitness function, and the evaluation criteria are: The value is the fitness value. The optimization process of the three algorithms is as follows Figure 4 As shown in Table 1, the optimization results are shown in Table 1.

[0082] Table 1 Optimization results of different algorithms

[0083]

[0084] Depend on Figure 4 It can be clearly seen that compared with the empirical setting, after using WOA and SSA to optimize the VMD parameters, the EI value of the evaluation standard is greatly reduced, which shows the effectiveness of the swarm intelligence algorithm in optimizing VMD. Compared with WOA-VMD, SSA-VMD has the smallest EI value and the best decomposition effect. This is because the SSA algorithm has a strong optimization ability and can more accurately find the optimal solution. and The optimal combination value; combined with Table 1, we can see that the minimum EI value is 0.19 when using SSA to optimize the parameters, which is a 92.96% and 72.06% reduction compared to the empirical settings and WOA optimization parameters, respectively. In summary, SSA can adaptively find the optimal combination of two parameters, which not only improves the integrity of the decomposed sequence information and the decomposition effect, but also avoids the tedious manual parameter search and reduces the subjective influence of the empirical settings.

[0085] The optimal parameter combination value is applied to the VMD algorithm, and the vault settlement test set sequence is decomposed into trend item displacement and random item displacement, as shown in Figure 5 shown.

[0086] To predict surrounding rock deformation, we used Keras 2.4.3 (a high-level neural network API) on the backend of TensorFlow 2.3.0 (an open-source software library for high-performance numerical computing) running Python 3.8. We constructed a GRU neural network (GRU, or Gate Recurrent Unit, is a highly effective variant of the LSTM network. GRUs are simpler and perform equally well than LSTMs, making them a very popular network currently available). The number of neurons in each GRU layer was 80 and 100, respectively. The optimizer used the Adam algorithm (Adam is a first-order optimization algorithm that can replace the traditional stochastic gradient descent process and iteratively updates the neural network weights based on training data). The loss function was the Mean Sequence (MSE), and the number of iterations was 100. After multiple trial calculations, we determined the input sequence length to be 3, meaning that the displacement prediction for the current stage is determined by the displacement data from the previous 3 days.

[0087] SVR (SVM regression algorithm is also called support vector regression or SVR. Support vector regression is a supervised learning algorithm used to predict discrete values.) and BP neural network (BP network (Back Propagation) was proposed by a team of scientists led by Rumelhart and McCelland in 1986. It is a multi-layer feedforward network trained by the error back propagation algorithm and is one of the most widely used neural network models.) were used to verify the superiority of GRU neural network. The prediction results of each component are as follows Figure 6 The overall prediction results are shown as Figure 7 As shown in Table 2, the quantitative evaluation index values ​​are shown in Table 2.

[0088] Table 2 Evaluation index values ​​of prediction results of different models

[0089]

[0090] Comparisons with commonly used SVM and BP neural network models show that the SSA-VMD-GRU model achieves the highest prediction accuracy and the lowest error. This demonstrates that the GRU neural network-based prediction model fully exploits the deep and effective information of the surrounding rock deformation sequence and better reflects the dynamic nature of the surrounding rock deformation system development process than static models.

[0091] In order to explore the influence of decomposition on the prediction results, three models, GRU, VMD-GRU and WOA-VMD-GRU, were established to predict the surrounding rock deformation sequence. Figure 8 As shown in Table 3, the quantitative evaluation index values ​​are shown in Table 3.

[0092] Table 3 Quantitative evaluation index values

[0093]

[0094] By considering the impact of undecomposed and decomposed algorithms on the prediction results, the necessity of optimizing VMD parameters and the superiority of SSA are fully demonstrated, and the rationality of the proposed evaluation criteria in the decomposition of surrounding rock deformation sequence is also proved.

[0095] In one embodiment, Figure 9 As shown, a tunnel surrounding rock deformation prediction device is provided, which is characterized by comprising: a data acquisition and judgment module 210, a VMD decomposition module 220, a GRU prediction processing module 230 and a calculation prediction value module 240, wherein:

[0096] The data collection and judgment module 210 is used to collect monitoring data of surrounding rock deformation and determine whether the monitoring data is equidistant; if so, the monitoring data is divided into a training set and a prediction set;

[0097] The VMD decomposition module 220 is used to define the VMD decomposition evaluation criteria and solve the VMD parameters. and parameters The optimal combination of and parameters The optimal combination of is used in VMD to obtain the target VMD; the target VMD is used to decompose the training set and prediction set in the monitoring data into trend item displacement and random item displacement respectively;

[0098] The GRU prediction processing module 230 is used to establish a GRU model and train the GRU model using a training set to obtain a target GRU model; perform displacement prediction based on the target GRU model and perform denormalization processing on the prediction result to obtain a target displacement prediction result, which includes a trend item displacement prediction result and a random item displacement prediction result;

[0099] The prediction value calculation module 240 is used to obtain the surrounding rock deformation prediction value by superimposing the target displacement prediction result.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A method for predicting tunnel surrounding rock deformation, characterized in that: include: Collecting monitoring data of surrounding rock deformation and determining whether the monitoring data are equally spaced; If yes, the monitoring data is divided into a training set and a prediction set; The root mean square error is selected as an indicator to measure the completeness of the information after decomposition; the sample entropy is introduced as an indicator to evaluate the decomposition effect; the VMD evaluation standard is constructed by combining the root mean square error and the sample entropy, and the VMD parameters are solved based on the sparrow search algorithm. and parameters The optimal combination of The parameters obtained and parameters The optimal combination of is used in VMD to obtain the target VMD; The target VMD is used to decompose the training set and the prediction set in the monitoring data into trend item displacement and random item displacement respectively; Establish a GRU model and train the GRU model using a training set to obtain a target GRU model; Performing displacement prediction according to the target GRU model and performing denormalization processing on the prediction result to obtain a target displacement prediction result, wherein the target displacement prediction result includes a trend item displacement prediction result and a random item displacement prediction result; Based on the target displacement prediction results, the surrounding rock deformation prediction value is obtained by superimposing the calculation.

2. The method according to claim 1, characterized in that After the steps of collecting monitoring data of surrounding rock deformation and determining whether the monitoring data are equally spaced, the method further includes: If not, the cubic spline difference method is used for equal spacing processing, and the processed monitoring data are divided into a training set and a prediction set.

3. The method according to claim 1, characterized in that The parameters of VMD are solved based on the sparrow search algorithm and parameters The optimal combination steps include: Converting the evaluation criteria into a constrained optimization problem; The sparrow search algorithm is used to solve the constraint optimization problem and obtain the parameters and parameters The optimal combination.

4. The method according to claim 1, wherein After the step of using the target VMD to decompose the training set and the prediction set in the monitoring data into trend term displacement and random term displacement, the method further includes: The monitoring data decomposed into trend item displacement and random item displacement are normalized.

5. The method according to claim 1, characterized in that The steps of establishing a GRU model and training the GRU model using a training set to obtain a target GRU model include: Establishing a first GRU model, and training the first GRU model using the training set in the trend item shift to obtain a first target GRU model; A second GRU model is established, and the second GRU model is trained using the training set in the random item shift to obtain a second target GRU model.

6. The method according to claim 5, characterized in that The steps of performing displacement prediction according to the target GRU model and performing denormalization processing on the prediction result to obtain a target displacement prediction result, wherein the target displacement prediction result includes a trend item displacement prediction result and a random item displacement prediction result include: The trend item displacement is predicted according to the first target GRU model, and the predicted trend item displacement is denormalized to obtain the trend item displacement prediction result; The random item displacement is predicted according to the second target GRU model, and the predicted random item displacement is denormalized to obtain the random item displacement prediction result.

7. The method according to claim 1, characterized in that The step of obtaining the surrounding rock deformation prediction value by superimposing and calculating the target displacement prediction result includes: Based on the time series superposition principle, the trend item displacement prediction result and the random item displacement prediction result are accumulated to calculate the surrounding rock deformation prediction value.

8. The method according to claim 1, characterized in that After the step of obtaining the surrounding rock deformation prediction value by superimposing the target displacement prediction result, the method further includes: Obtaining actual values ​​of surrounding rock deformation, and calculating target determination coefficient, target mean absolute error, and target root mean square error based on the predicted values ​​of surrounding rock deformation; The target determination coefficient, the target mean absolute error and the target root mean square error are used to quantitatively evaluate the predicted value of surrounding rock deformation.

9. A tunnel surrounding rock deformation prediction device, characterized in that: include: Data collection and judgment module, VMD decomposition module, GRU prediction processing module and calculation prediction value module, among which, The data acquisition and judgment module is used to collect monitoring data of surrounding rock deformation and determine whether the monitoring data is equidistant; if so, the monitoring data is divided into a training set and a prediction set; The VMD decomposition module is used to select the root mean square error as an indicator to measure the completeness of the decomposed information; introduce sample entropy as an indicator to evaluate the decomposition effect; combine the root mean square error and the sample entropy to construct the VMD evaluation standard and solve the VMD parameters and parameters The optimal combination of and parameters The optimal combination of is used in VMD to obtain a target VMD; the target VMD is used to decompose the training set and the prediction set in the monitoring data into trend item displacement and random item displacement respectively; The GRU prediction processing module is used to establish a GRU model and train the GRU model using a training set to obtain a target GRU model; perform displacement prediction based on the target GRU model and perform denormalization processing on the prediction result to obtain a target displacement prediction result, wherein the target displacement prediction result includes a trend item displacement prediction result and a random item displacement prediction result; The calculation prediction value module is used to obtain the surrounding rock deformation prediction value by superimposing the target displacement prediction result.

Images