Sandy soil earthquake liquefaction potential rapid evaluation method

Abstract

The invention relates to the technical field of data classification prediction, and discloses a sandy soil earthquake liquefaction potential rapid evaluation method, which comprises the following steps: S1, selecting data from a data set commonly used in the field of sandy soil liquefaction as a first data set; s2, constructing a sandy soil earthquake liquefaction potential evaluation model; s3, inputting the data of the first data set into the model, performing domain adaptive training by using the maximum mean value difference, correcting the model, and performing data prediction on the second data set by using the corrected model; and S4, comparing the predicted sandy soil liquefaction condition with the actual sandy soil liquefaction condition, and carrying out validity verification on the sandy soil seismic liquefaction potential. The scheme improves the adaptability of the model on different data sets, solves the problems of low prediction efficiency, easy occurrence of an overfitting phenomenon of the model and the like in an existing scheme, and improves the prediction accuracy of the sandy soil seismic liquefaction potential. And the method shows higher accuracy and reliability in sandy soil earthquake liquefaction potential evaluation, so that the method has better interpretability and applicability.

Description

Technical Field

[0001] The invention relates to the technical field of data classification prediction, and in particular to a method for quickly evaluating the earthquake liquefaction potential of sandy soil. Background Art

[0002] Sand liquefaction is one of the major natural disasters that has attracted widespread attention worldwide. It can easily cause serious structural damage to above-ground buildings and infrastructure. Therefore, establishing a new prediction model to evaluate sand liquefaction is an important task in the field of earthquake disaster prevention and reduction. However, in recent years, in the uncertainty of evaluating sand liquefaction, the existing methods have exposed the following major problems: (1) Soil liquefaction-related data are relatively scarce. Since soil liquefaction experimental data are basically small sample data sets, in the case of small samples, the accuracy and reliability of the results may be significantly affected due to insufficient data; (2) In the binary prediction problem of sand liquefaction, the accuracy of existing prediction methods is generally low. Therefore, it is necessary to correctly apply appropriate methods to avoid errors and biases and solve the problem of low accuracy.

[0003] In addition, existing prediction methods have certain limitations. For example, when the data set is insufficient or the data distribution changes, these methods need to retrain the model based on the specific data set, resulting in low prediction efficiency and the model is prone to overfitting. Summary of the invention

[0004] In order to overcome the problems in the prior art, the present invention provides a method for quickly evaluating the seismic liquefaction potential of sandy soil. To achieve the above purpose, the present invention provides the following scheme: S1. Select data from a commonly used data set in the field of sand liquefaction as the first data set; S2. Construct a sandy soil earthquake liquefaction potential assessment model; S3, inputting the data of the first data set into the model, performing domain adaptive training using the maximum mean difference, correcting the model, and using the corrected model to perform data prediction on the second data set; S4. Compare the predicted sand liquefaction situation with the actual sand liquefaction situation to verify the effectiveness of the sand seismic liquefaction potential.

[0005] A further technical solution is that the experimental data in step S1 includes the corrected shear wave velocity and cyclic stress ratio As the first data set and perform data processing.

[0006] A further technical solution is that in step S2, a convolutional neural network is used to construct a sand seismic liquefaction potential assessment model.

[0007] A further technical solution is to introduce transfer learning in step S2 to construct a sand seismic liquefaction potential assessment model.

[0008] A further technical solution is that the step S2 further includes: Using the 2008 Wenchuan earthquake dataset, the corrected shear wave velocity is also selected and cyclic stress ratio The data constructs the second data set.

[0009] A further technical solution is that step S3 further includes: Save the weights of the convolutional neural network constructed for the first data set, build a convolutional neural network model for the second data set, and migrate the weights to the latter network model.

[0010] A further technical solution is that step S3 further includes: Train the model on the second dataset and update as well as ,at this time Smaller than the previous value, 5000 pre-training iterations and 1000 training iterations; The modified shear wave velocity calculated from the first and second data sets is and cyclic stress ratio Normalization is performed to eliminate the adverse effects caused by singular sample data.

[0011] On the other hand, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of the sand seismic liquefaction potential evaluation method when executing the computer program.

[0012] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the sand seismic liquefaction potential evaluation method are implemented.

[0013] The technical effects of the present invention are as follows: In view of the above background technology, the technical solution of this application has significant technical effects and successfully solves several technical problems mentioned in the background technology. The comparison results with the existing solutions show that the sand seismic liquefaction potential rapid evaluation method proposed in this solution effectively overcomes the shortcomings of the existing solutions.

[0014] Specifically, existing solutions usually need to retrain the model when faced with insufficient data sets or changes in data distribution, resulting in poor flexibility and adaptability. This solution significantly improves the adaptability of the model on different data sets by optimizing the algorithm design, thereby avoiding the problem of frequent retraining. In addition, the prediction efficiency of existing solutions is low, and the model is prone to overfitting, which limits the effectiveness and reliability of its practical application. In contrast, this solution not only significantly improves the prediction efficiency but also effectively reduces the risk of overfitting by improving the model structure and introducing new regularization techniques. This solution significantly improves the prediction efficiency and the generalization ability of the model by optimizing the algorithm and model structure, and has higher practical value.

[0015] In summary, this scheme is superior to the existing schemes in terms of technical effect, and has better interpretability and applicability, and can be more effectively applied to the actual evaluation of sand seismic liquefaction potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to clearly explain the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly described below. The drawings described below are only schematic diagrams of some embodiments of the present invention. For ordinary technicians in this field, other related drawings can be derived from these drawings without creative work.

[0017] Figure 1 It is a schematic diagram of the method flow of the present invention; Figure 2 It is a flow chart of the sand soil earthquake liquefaction potential analysis method of the present invention; Figure 3 It is a flow chart of sand liquefaction assessment model training of the present invention; Figure 4 The first data model of the embodiment liquefies the boundary map in the first data set; Figure 5 The second data model of the embodiment liquefies the boundary map in the second data set. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions of the embodiments of the present invention. It should be noted that the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0019] The object of the present invention is to provide a method for rapid evaluation of sand liquefaction potential during earthquakes. To make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0020] This embodiment mainly uses a commonly used data set in the field of sand liquefaction as the first data set, builds a trained and improved CNN neural network based on the data set as a pre-training model, and migrates this model to the 2008 Wenchuan earthquake data as the second data set model training. On the basis of retaining certain pre-training data information, the network model reconciles the Wenchuan earthquake data information, so that the revised model has more generalization ability in the Wenchuan earthquake data sand liquefaction prediction problem, and effectively addresses the problem of model overfitting.

[0021] like Figure 1 As shown, the present invention provides a method for quickly evaluating the seismic liquefaction potential of sandy soil, comprising the following steps: S1. Select data from a commonly used data set in the field of sand liquefaction as the first data set; In data processing, the present invention mainly adopts The method has the following significant advantages: (1) It is applicable to complex strata such as gravel sand, especially in areas where the reliability of standard penetration test (SPT) and cone penetration test (CPT) is low, and can provide more accurate evaluation of mechanical properties; (2) It can be used not only for field tests but also for laboratory samples. By comparing the results of laboratory and field tests, the consistency and reliability of the method can be verified; (3) The method is directly related to the small strain shear modulus, which is a key engineering property that describes the dynamic mechanical behavior of sand in the small strain range.

[0022] This embodiment uses the modified shear wave velocity and cyclic stress ratio Recently, as the input node of the CNN neural network model, define is the amplitude proportional factor, and the calculation expression is: (1) in, is the earthquake magnitude; definition is the shear stress reduction coefficient, and its expression is: (2) in, is the formation depth; definition is the cycle force ratio, and the calculation formula is: (3) in, is the peak ground acceleration, is the acceleration due to gravity, is the total overburden stress, is the initial effective overburden stress; definition is the net sand correction adjustment value, and its expression is: (4) in, is the fine sand content, is the stress-corrected shear wave velocity, The calculation formula is: (5) definition To correct the shear wave velocity, the calculation formula is: (6) in, is the small strain shear wave velocity, is the reference stress of 100 kPa.

[0023] S2. Construct a sandy soil earthquake liquefaction potential assessment model; The model is based on the constructed CNN convolutional neural network and updates the weights in the network through the back propagation algorithm. Figure 2 As shown, the first data set is taken as input, and the neuron input values ​​of the hidden layer and the output layer (except the bias term) are calculated through forward propagation ( )、output value( ) and the final output value ( ). Then, according to the objective function, the weights of the neurons in the input layer and the hidden layer corresponding to the next layer are adjusted by back propagation. The objective function of the present invention is defined as: (7) in, is the sample size, It is a sample The real label, The labels predicted by the model CNN model. At the end of the training phase, the objective function value is infinitely close to the minimum, indicating that the neural network prediction results are as close as possible to the target output of the training data.

[0024] According to formula (7), the loss of each neuron in the output layer is calculated and the weight value of each neuron is updated. Each data is passed into the network to perform the above calculation and update the weight, and finally repeat the above operation according to the number of iterations. For example, data, and the number of iterations is , then ( ) weight updates. The model algorithm flow chart is as follows Figure 2 .

[0025] S3, inputting the data of the first data set into the model, performing domain adaptive training using the maximum mean difference, correcting the model, and using the corrected model to perform data prediction on the second data set; First, train the improved CNN neural network in the pre-trained dataset and save this network for weight migration. Then build a CNN neural network model for the second dataset, migrate the weights to this network model, and finally train this model in the second dataset and update as well as ,at this time Smaller than the previous values ​​(5000 pre-training iterations and 1000 training iterations).

[0026] For model training, the first data set and the second data set need to be calculated and Normalization is performed to eliminate the adverse effects caused by singular sample data: (8) (9) in, , Respectively, in the two data sets The minimum and maximum values ​​of , Respectively, in the two data sets The minimum and maximum values ​​of . The model training process is as follows Figure 3 .

[0027] In order to help the model better adapt to the target domain data and achieve better migration performance between the source domain and the target domain, the maximum mean discrepancy (MMD) is used for domain adaptation training to reduce the distribution difference between the source domain and the target domain data by minimizing MMD.

[0028] S4. Compare the predicted sand liquefaction situation with the actual sand liquefaction situation to verify the effectiveness of the sand seismic liquefaction potential.

[0029] To evaluate the model performance and compare the developed pre-trained model with other models, three widely used classification evaluation metrics were applied, including the classification accuracy calculation of the model and the probability prediction error . It shows how good the model classification is, and It indicates the probability that the model will classify the sample correctly. The smaller the probability prediction error, the greater the probability that the model predicts the sample correctly, and the more confident that the predicted result is correct. The classification accuracy of the model is calculated as follows: (10) in, Indicates the number of samples that the model predicts correctly. is the total number of samples; The uncertainty of the model can be expressed by the probabilistic prediction error, which is calculated as follows: (11) Among them, the probability prediction error of the model The Euclidean norm sum of the differences between the target output and the predicted value for all data. For the The target output vector of data, A probability vector representing the output of the model.

[0030] Table 1 Performance comparison of different models on the first dataset Table 2 Performance comparison of different models on the second dataset Tables 1 and 2 show the performance comparison of different models on the first and second datasets, respectively. As can be seen from Table 1, on the first dataset, the CNN neural network model has the highest classification accuracy, followed by the SVM model, while the Logistic probability model has the lowest classification accuracy. From Table 2, it can be concluded that on the second dataset, the CNN neural network model also performs better than the Logistic probability model; in terms of accuracy, the Logistic probability model is consistent with the SVM model, but the loss value of the CNN neural network model is significantly lower than that of the Logistic probability model and the SVM model. Therefore, in terms of evaluating sand liquefaction, the pre-trained CNN neural network model has less uncertainty.

[0031] Figure 4 and Figure 5 The liquefaction boundary diagrams of the first data model and the second data model in the embodiment are shown respectively. The sand liquefaction problem can be regarded as a binary classification problem. Based on the trained CNN neural network model, the liquefaction boundary of the model of the method can be drawn. The performance of the CNN neural network model trained on the first data set (as a pre-trained model) is shown in FIG. Figure 4 As shown in Figure 2, for sample data distributed in the low modified shear wave velocity area, the CNN neural network model can distinguish different sample data well. The pre-trained model was transferred to the second data set, and after a small amount of training, the CNN neural network model for Wenchuan sandy soil earthquake liquefaction assessment was obtained, as shown in Figure 2. Figure 5 As shown in Figure 2, it can be found that the model after a small amount of training is similar to Figure 4The segmentation line shape of the pre-trained model in is basically the same, which retains a lot of information of the pre-training dataset and also shows good performance on the second dataset. As shown in Table 3, the accuracy of the model on the second dataset reaches 84.44%.

Claims

1. A method for rapid evaluation of sand liquefaction potential during earthquakes, characterized in that: The following steps are involved: S1. Select data from a commonly used data set in the field of sand liquefaction as the first data set; S2. Construct a sandy soil earthquake liquefaction potential assessment model; S3, inputting the data of the first data set into the model, performing domain adaptive training using the maximum mean difference, correcting the model, and using the corrected model to perform data prediction on the second data set; S4. Compare the predicted sand liquefaction situation with the actual sand liquefaction situation to verify the effectiveness of the sand seismic liquefaction potential.

2. The method for rapid evaluation of sand liquefaction potential according to claim 1, characterized in that: In step S1, the corrected shear wave velocity is selected and cyclic stress ratio As the core parameter of the first data set, and perform data processing on it.

3. The method for rapid evaluation of sand liquefaction potential according to claim 1, characterized in that: In step S2, a convolutional neural network is used to construct a sand seismic liquefaction potential assessment model.

4. The method for rapid evaluation of sand liquefaction potential according to claim 1, characterized in that: In step S2, transfer learning is introduced to construct a sand seismic liquefaction potential assessment model.

5. The method for rapid evaluation of sand liquefaction potential according to claim 1, characterized in that: In step S3, the 2008 Wenchuan earthquake data set is used to select the corrected shear wave velocity and cyclic stress ratio The data constructs the second data set.

6. The method for rapid evaluation of sand liquefaction potential according to claim 1, characterized in that: Step S3 also includes: saving the weights of the convolutional neural network constructed by the first data set, constructing a convolutional neural network model for the second data set, and migrating the weights to the latter network model.

7. The method for rapid evaluation of sand liquefaction potential according to claim 1, characterized in that: The step S3 further comprises: Train the model on the second dataset and update as well as ,at this time Smaller than the previous value, 5000 pre-training iterations and 1000 training iterations; The modified shear wave velocity calculated from the first and second data sets is and cyclic stress ratio Normalization is performed to eliminate the adverse effects caused by singular sample data.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method for rapid evaluation of sand seismic liquefaction potential described in any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, a method for quickly evaluating the seismic liquefaction potential of sand soil as described in any one of claims 1 to 7 is implemented.

Images