Method and device for judging key factors influencing soil settlement by tunnel excavation

Abstract

The invention provides a method and device for judging key factors influencing soil settlement by tunnel excavation, and the method is characterized in that the method comprises the following steps: a data set construction step: carrying out the classification permutation and combination of all influence factors to generate a plurality of combinations, and taking all working condition data corresponding to each combination as a corresponding data set; a model construction step: constructing an initial model; a model training step: training the initial model through each data set to obtain a corresponding prediction model and performance data; a performance analysis step: analyzing all performance data, and selecting n prediction models with optimal performance; and an intersection generation step: maximum intersection processing is carried out on the combinations corresponding to the n prediction models to obtain a maximum intersection, and each influence factor in the maximum intersection is a key factor. In short, the method can extract the corresponding key factors according to the existing working condition data.

Description

Technical Field

[0001] The invention belongs to the technical field of tunnel engineering, and in particular relates to a method and a device for judging key factors affecting soil settlement due to tunnel excavation. Background Art

[0002] In the process of tunnel excavation, the prediction of ground settlement is an extremely important link. Accurately predicting soil settlement helps to ensure project safety, improve construction efficiency, reduce construction costs, and also plays a vital role in protecting the environment and surrounding buildings.

[0003] Traditional soil settlement prediction methods include empirical methods, numerical simulation methods, theoretical analysis methods, machine learning methods, etc. Among them, machine learning methods have broad application prospects in the prediction of soil settlement during tunnel excavation. They can handle complex nonlinear relationships and high-dimensional data and have high prediction accuracy. However, these methods usually have poor model interpretability due to the complexity of the model and the existence of feature interactions and nonlinear relationships. In complex models, even if the model can capture the relationship between various features, it is difficult to explain the contribution of each feature separately. This will bring problems in engineering decision-making, risk management, and scientific research.

[0004] In summary, it is necessary to develop a method to select the features that have the greatest impact on soil settlement among various features in tunnel excavation under actual construction scenarios to solve practical problems. Summary of the invention

[0005] The present invention is made to solve the above-mentioned problem, and aims to provide a method and device for determining the key factors affecting soil settlement caused by tunnel excavation.

[0006] The present invention provides a method for judging the key factors affecting soil settlement due to tunnel excavation, which is used to obtain the key factors according to the working condition data under different existing influencing factors, and has the following characteristics, including the following steps: a data set construction step, in which all the influencing factors are arranged and combined into categories to generate multiple combinations, and all the working condition data corresponding to each combination are used as the corresponding data set; a model construction step, in which an initial model is constructed; a model training step, in which the initial model is trained through each data set respectively to obtain the corresponding prediction model and performance data; a performance analysis step, in which all the performance data are analyzed and n prediction models with the best performance are selected; an intersection generation step, in which the combinations corresponding to the n prediction models are subjected to maximum intersection processing to obtain the maximum intersection, wherein each influencing factor in the maximum intersection is a key factor.

[0007] In the method for determining the key factors affecting soil settlement due to tunnel excavation provided by the present invention, the following features may also be provided: wherein, the data set construction step includes the following sub-steps: step S1-1, extracting influencing factors of each category from all working condition data; step S1-2, arranging and combining the influencing factors of each category to obtain multiple combinations; step S1-3, for each combination, using all working condition data corresponding to only all influencing factors in the combination as the data set corresponding to the combination.

[0008] In the method for determining the key factors affecting soil settlement due to tunnel excavation provided by the present invention, the method may also have the following characteristics: wherein the initial model is a four-layer neural network model, the four-layer neural network model includes three hidden layers of 64 neurons, a Sigmoid activation function and a Dropout layer, and the expression of the Sigmoid activation function is: f(x)∈(0,1], where x is the independent variable of the activation function.

[0009] The method for determining the key factors affecting soil settlement due to tunnel excavation provided by the present invention may also have the following characteristics: in the model training step, the initial model is trained using K-Fold and the data set, as well as predicted and evaluated to obtain a prediction model and performance data.

[0010] In the method for determining the key factors affecting soil settlement due to tunnel excavation provided by the present invention, the method may also have the following features: wherein, in the model training step, the model training is performed by using an Adam optimizer and a mean square error loss function, and the expression of the mean square error loss function is: Where MSE is the calculation result of the mean square error loss function, n is the number of samples used in training, and y i is the model prediction value corresponding to the i-th sample, is the true value corresponding to the i-th sample.

[0011] The method for determining key factors affecting soil settlement due to tunnel excavation provided by the present invention may also have the following characteristics: wherein the performance data includes MAE, MSE, RMSE and MAPE.

[0012] The present invention also provides a device for determining the key factors affecting soil settlement due to tunnel excavation, which is used to obtain the key factors based on the working condition data under different existing influencing factors, and has the following characteristics, including: a data set construction module, which is used to generate multiple combinations by categorizing and combining all influencing factors, and use all working condition data corresponding to each combination as the corresponding data set; a model construction module, which is used to construct an initial model; a model training module, which is used to train the initial model through each data set respectively, to obtain the corresponding prediction model and performance data; a performance analysis module, which is used to analyze all performance data and select n prediction models with the best performance; an intersection generation module, which is used to perform maximum intersection processing on the combinations corresponding to the n prediction models to obtain the maximum intersection, wherein each influencing factor in the maximum intersection is a key factor.

[0013] Functions and Effects of the Invention

[0014] According to the method and device for determining the key factors affecting soil settlement due to tunnel excavation involved in the present invention, the influencing factors are summarized for the working condition data of multiple different databases, and a model is trained by the data corresponding to the combination of various influencing factors, and a preferred combination is obtained according to the prediction performance of the trained model, and the key factors are extracted from the preferred combination by maximum intersection processing. Therefore, the method and device for determining the key factors affecting soil settlement due to tunnel excavation of the present invention can extract the corresponding key factors according to the existing working condition data. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a block diagram of a key factor determination device in an embodiment of the present invention;

[0016] Figure 2 is a schematic diagram of a process of constructing a data set in an embodiment of the present invention;

[0017] Figure 3 It is a schematic diagram of prediction results of prediction models corresponding to some combinations in the embodiments of the present invention;

[0018] Figure 4 is a schematic diagram of maximum intersection processing in an embodiment of the present invention;

[0019] Figure 5 is a schematic diagram of the prediction performance of the prediction model corresponding to the combination TP1 and the combination GC3 in the embodiment of the present invention;

[0020] Figure 6 It is a flow chart of a method for determining key factors affecting soil settlement due to tunnel excavation in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the following embodiments and the accompanying drawings specifically illustrate the method and device of the present invention for determining the key factors affecting soil settlement caused by tunnel excavation.

[0022] The present embodiment provides a device for determining key factors affecting soil settlement due to tunnel excavation, hereinafter referred to as a key factor determining device, which is used to obtain key factors based on existing working condition data under different influencing factors.

[0023] In this embodiment, the existing working condition data under different influencing factors come from three existing databases DB1, DB2 and DB3 respectively. Among them, DB1 contains two types of influencing factors, namely geological conditions GC and tunnel geometric parameters TP, DB2 contains three types of influencing factors, namely geological conditions GC, tunnel geometric parameters TP and construction methods CM, and DB3 contains one type of influencing factor, namely geological conditions GC.

[0024] Figure 1 It is a block diagram of a key factor determination device in an embodiment of the present invention.

[0025] like Figure 1 As shown, the key factor determination device 100 includes a data set construction module 10, a model construction module 20, a model training module 30, a performance analysis module 40, an intersection generation module 50 and a control module 60 for controlling the above modules.

[0026] The data set construction module 10 is used to generate multiple combinations by categorizing and combining all influencing factors, and to use all operating condition data corresponding to each combination as the corresponding data set.

[0027] Figure 2 It is a flowchart of data set construction in an embodiment of the present invention.

[0028] like Figure 2 As shown, in the data set construction module 10, the data set construction includes the following steps:

[0029] Step S1-1, extracting influencing factors of each category from all operating condition data.

[0030] In this embodiment, through step S1-1, the geological condition GC in DB1 is recorded as GC1, the tunnel geometric parameter TP in DB1 is recorded as TP1, the geological condition GC in DB2 is recorded as GC2, the tunnel geometric parameter TP in DB2 is recorded as TP2, the construction method CM in DB2 is recorded as CM2, and the geological condition GC in DB3 is recorded as GC3, that is, a total of 6 categories of influencing factors are obtained.

[0031] Step S1-2, arranging and combining the influencing factors of each category to obtain multiple combinations.

[0032] In step S1-2 of this embodiment, the operating condition data corresponding to the combination is not empty, that is, there is at least one operating condition data corresponding to the combination. Therefore, all categories of influencing factors can be exhaustively combined, and then the combination without corresponding operating condition data can be deleted to obtain the final valid combination. The valid combinations in this embodiment are: GC1, TP1, GC1+TP1, GC2, TP2, CM2, GC2+TP2, GC2+CM2, TP2+CM2, CM1+TP2+GC2 and GC3, a total of 11.

[0033] Step S1-3: for each combination, all the corresponding working condition data of all the influencing factors in the combination are taken as the data set corresponding to the combination.

[0034] The model building module 20 is used to build an initial model.

[0035] The initial model is a four-layer neural network model, which includes three hidden layers of 64 neurons, a Sigmoid activation function and a Dropout layer. In this embodiment, the Dropout layer is set with a dropout probability of 0.3, and the Sigmoid activation function and the Dropout layer are used to prevent overfitting.

[0036] The expression of the Sigmoid activation function is:

[0037]

[0038] Where x is the independent variable of the activation function. In the network structure of this embodiment, x is the input value of the neuron, which is obtained by linearly combining the output value of the previous layer and the weight matrix of the current layer.

[0039] The model training module 30 is used to train the initial model through each data set to obtain the corresponding prediction model and performance data, wherein the performance data includes MAE, MSE, RMSE and MAPE.

[0040] In this embodiment, each operating condition data in the data set includes corresponding influencing factor data and surface settlement data, wherein the influencing factor data is used as the input data of the model, and the surface settlement data is used as the true value, which is used to compare with the predicted value of the model and assist in training the model.

[0041] Among them, in the model training module 30, the initial model is trained, predicted and evaluated by K-Fold and the data set, and then the prediction model and performance data are obtained. In this embodiment, 10-fold cross validation is performed by K-Fold, and the data set is divided into a training set and a test set. In each cross validation, the initial model is trained by the training set, and the test set is used for prediction and evaluation, thereby reducing the model prediction variance and improving the generalization ability of the model.

[0042] Furthermore, the model is trained using the Adam optimizer and the mean square error loss function. The expression of the mean square error loss function is:

[0043]

[0044] Where MSE is the calculation result of the mean square error loss function, n is the number of samples used in training, and y i is the model prediction value corresponding to the i-th sample, is the true value corresponding to the i-th sample.

[0045] The error index data of the models corresponding to each combination of DB2 in this embodiment in terms of MAE, MSE, RMSE, MAPE and loss are shown in the following table:

[0046]

[0047]

[0048] The first column in the table above shows the various combinations, and the second to sixth columns show the loss, MAE, MSE, RMSE, and MAPE values ​​of the prediction models corresponding to the various combinations. For example, the cell in the third column of the second row shows that the error indicator MAE value of the prediction model corresponding to the combination GC2 is 0.511161625.

[0049] The performance analysis module 40 is used to analyze all performance data and select n prediction models with the best performance. In this embodiment, n prediction models with the best performance are selected according to the relative sizes of MAE, MSE, RMSE and MAPE of different performance data predicted by different data sets.

[0050] In this embodiment, the performance data also includes prediction accuracy, that is, the error result between the predicted value of the prediction model and the corresponding true value.

[0051] In this embodiment, the prediction models corresponding to combination GC2, combination GC2+TP2, combination GC2+CM2 and combination CM1+TP2+GC2 have better prediction accuracy than other combinations, so the prediction models corresponding to the above four combinations are selected as the n=4 prediction models with the best performance.

[0052] Figure 3 It is a schematic diagram of prediction results of prediction models corresponding to some combinations in an embodiment of the present invention.

[0053] like Figure 3 As shown, (a) is a schematic diagram of the predicted values ​​and true values ​​of the prediction model corresponding to the combination GC2 on each input, (b) is a schematic diagram of the predicted values ​​and true values ​​of the prediction model corresponding to the combination GC2+TP2 on each input, (c) is a schematic diagram of the predicted values ​​and true values ​​of the prediction model corresponding to the combination GC2+CM2 on each input, (d) is a schematic diagram of the predicted values ​​and true values ​​of the prediction model corresponding to the combination CM1+TP2+GC2 on each input. In each figure, the horizontal axis is the group, the vertical axis is the soil settlement value, Measured is the curve composed of the true value, and Predicted is the curve composed of the predicted value.

[0054] The intersection generation module 50 is used to perform maximum intersection processing on the combinations corresponding to the n prediction models to obtain the maximum intersection, wherein each influencing factor in the maximum intersection is a key factor.

[0055] Figure 4 4 is a schematic diagram of maximum intersection processing in an embodiment of the present invention.

[0056] like Figure 4 As shown, in this embodiment, the influencing factors of each combination are restored, that is, the same influencing factors from different databases are no longer distinguished, and the above four combinations are respectively converted into corresponding working conditions, that is, working condition 4: GC, working condition 7: GC+TP, working condition 8: GC+CM and working condition 10: CM+TP+GC. The maximum intersection of the above four working conditions is removed, and GC, that is, the geological condition, is the maximum intersection, so the geological condition GC is the key factor.

[0057] The above key factors are further verified below, specifically, a combination that does not include geological condition GC1, namely combination TP1, is selected from database DB1, and combination GC3 is selected from database DB3. The prediction performance of the prediction model corresponding to combination TP1 and the prediction model corresponding to combination GC3 are analyzed respectively.

[0058] Figure 5 It is a schematic diagram of the prediction performance of the prediction models corresponding to the combination TP1 and the combination GC3 in the embodiment of the present invention.

[0059] like Figure 5As shown, (a) is a schematic diagram of the prediction performance of the prediction model corresponding to the combination TP1, and (b) is a schematic diagram of the prediction performance of the prediction model corresponding to the combination GC3. In each figure, the horizontal axis is the group, the vertical axis is the soil settlement value, Measured is the curve composed of the true value, and Predicted is the curve composed of the predicted value. It can be seen that the prediction model corresponding to the combination TP1 has poor prediction ability, and the prediction model corresponding to the combination GC3 has good prediction ability. Therefore, the geological condition GC is the key factor affecting the soil settlement caused by tunnel excavation.

[0060] The control module 60 stores a control program for controlling the operation of each module.

[0061] The following describes the process of using the key factor determination device 100 to determine the key factors affecting soil settlement due to tunnel excavation in conjunction with the accompanying drawings.

[0062] Figure 6 It is a flow chart of a method for determining key factors affecting soil settlement due to tunnel excavation in an embodiment of the present invention.

[0063] like Figure 6 As shown in FIG. 1 , the method for determining the key factors affecting soil settlement due to tunnel excavation includes the following steps:

[0064] Step S1, a data set construction step, uses the data set construction module 10 to arrange and combine all influencing factors into categories to generate multiple combinations, and uses all operating condition data corresponding to each combination as the corresponding data set.

[0065] Step S2, a model building step, uses the model building module 20 to build an initial model.

[0066] Step S3, a model training step, using the model training module 30 to train the initial model through each data set to obtain the corresponding prediction model and performance data;

[0067] Step S4, a performance analysis step, using the performance analysis module 40 to analyze all performance data and select n prediction models with the best performance;

[0068] Step S5, an intersection generation step, uses the intersection generation module 50 to perform maximum intersection processing on the combinations corresponding to the n prediction models to obtain the maximum intersection.

[0069] In this embodiment, the order of step S1 and step S2 is to execute step S1 first and then execute step S2. In other embodiments, as an alternative, the order of the above steps S1 and S2 can also be swapped according to actual needs, that is, to execute step S2 first and then execute step S1, or to execute step S1 and step S2 together as parallel steps.

[0070] Functions and Effects of the Embodiments

[0071] According to the method and device for determining the key factors affecting soil settlement caused by tunnel excavation involved in this embodiment, the working condition data of multiple different databases are summarized for the influencing factors, a model is trained by the data corresponding to the combination of various influencing factors, a preferred combination is obtained according to the prediction performance of the trained model, and the key factors are extracted from the preferred combination by maximum intersection processing. In short, this method can extract the corresponding key factors according to the existing working condition data.

[0072] Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for determining the key factors affecting soil settlement caused by tunnel excavation, which is used to obtain the key factors based on the existing working condition data under different influencing factors, and is characterized in that: The following steps are involved: A data set construction step, arranging and combining all the influencing factors by category to generate multiple combinations, and taking all the working condition data corresponding to each combination as the corresponding data set; Model building step, building the initial model; A model training step, training the initial model using each of the data sets to obtain a corresponding prediction model and performance data; A performance analysis step, analyzing all the performance data and selecting n prediction models with the best performance; The intersection generation step is to perform maximum intersection processing on the combinations corresponding to the n prediction models to obtain the maximum intersection. Among them, each of the influencing factors in the maximum intersection is the key factor.

2. The method for determining key factors affecting soil settlement caused by tunnel excavation according to claim 1, characterized in that: in, The data set construction step includes the following sub-steps: Step S1-1, extracting the influencing factors of each category from all the operating condition data; Step S1-2, arranging and combining the influencing factors of each category to obtain a plurality of combinations; Step S1-3: for each of the combinations, all the corresponding operating condition data only including all the influencing factors in the combination are taken as the data set corresponding to the combination.

3. The method for determining key factors affecting soil settlement caused by tunnel excavation according to claim 1, characterized in that: in, The initial model is a four-layer neural network model. The four-layer neural network model includes three hidden layers of 64 neurons, a Sigmoid activation function and a Dropout layer. The expression of the Sigmoid activation function is: Where x is the independent variable of the activation function.

4. The method for determining key factors affecting soil settlement caused by tunnel excavation according to claim 1, characterized in that: in, In the model training step, the initial model is trained, predicted and evaluated using K-Fold and the data set to obtain the prediction model and the performance data.

5. The method for determining key factors affecting soil settlement caused by tunnel excavation according to claim 1, characterized in that: in, In the model training step, the model training is performed using the Adam optimizer and the mean square error loss function. The expression of the mean square error loss function is: Where MSE is the calculation result of the mean square error loss function, n is the number of samples used in training, and y i is the model prediction value corresponding to the i-th sample, is the true value corresponding to the i-th sample.

6. The method for determining key factors affecting soil settlement caused by tunnel excavation according to claim 1, characterized in that: in, The performance data include MAE, MSE, RMSE and MAPE.

7. A device for determining the key factors affecting soil settlement due to tunnel excavation, used to obtain the key factors based on the existing working condition data under different influencing factors, characterized in that: include: A data set construction module, used to arrange and combine all the influencing factors into categories to generate multiple combinations, and use all the working condition data corresponding to each combination as the corresponding data set; Model building module, used to build the initial model; A model training module, used to train the initial model through each of the data sets to obtain corresponding prediction models and performance data; A performance analysis module, used to analyze all the performance data and select n prediction models with the best performance; An intersection generation module is used to perform maximum intersection processing on the combinations corresponding to the n prediction models to obtain a maximum intersection, Among them, each of the influencing factors in the maximum intersection is the key factor.

Images