CN-121980191-A - Membrane bag sand dike slope settlement prediction method and system based on neural network

CN121980191ACN 121980191 ACN121980191 ACN 121980191ACN-121980191-A

Abstract

The application discloses a membrane bag sand dike slope settlement prediction method and system based on a neural network, and belongs to the technical field of prediction systems; and based on the geological exploration data, predicting each compression amount by analyzing the pore drainage rate of each foundation soil layer, outputting a settlement prediction result of the membrane bag sand embankment slope, identifying the settlement abnormal risk of the membrane bag sand embankment slope, and outputting the settlement risk level of a preset construction point. Therefore, by implementing the method, the problem that the predicted total settlement of the membrane bag sand dike slope is not accurate enough due to the fact that the influence of the dike slope foundation structure on the settlement amount in the construction process is not considered in the prior art can be solved.

Inventors

WANG XIAOLIANG
TANG WENBIN
ZHOU MI
Hua Ruiqing
ZHANG XIAOFENG
CHEN ZHUO

Assignees

广东海洋大学
华南理工大学

Dates

Publication Date: 20260505
Application Date: 20260225

Claims (10)

1. The membrane bag sand dike slope settlement prediction method based on the neural network is characterized by comprising the following steps of: acquiring construction parameters, foundation monitoring data and geological exploration data of a membrane bag sand bank slope at a plurality of preset time points; Taking the compression amount of each foundation layer as a target feature, and carrying out feature fusion by analyzing the correlation between the construction parameters and the foundation monitoring data and the target feature in the time dimension to obtain a fusion time sequence feature; performing feature extraction and feature splicing on the fusion time sequence features and the geological exploration data, inputting a feature splicing result into a preset settlement prediction model, predicting the compression amount of each foundation soil layer by analyzing the pore drainage rate of each foundation soil layer, and outputting a settlement prediction result of the membrane bag sand embankment slope; and identifying abnormal sedimentation risks of the membrane bag sand embankment slope through a preset dynamic early warning mechanism according to the sedimentation prediction result, and outputting sedimentation risk grades of the membrane bag sand embankment slope at each preset construction point.
2. The neural network-based membrane bag sand bank slope settlement prediction method according to claim 1, wherein the method is characterized by obtaining construction parameters, foundation monitoring data and geological exploration data of the membrane bag sand bank slope at a plurality of preset time points, and specifically comprises the following steps: according to the construction form and geological profile result of the subgrade under the membrane bag sand bank slope, setting a plurality of preset measuring points on each foundation soil layer; For the preset measuring points, pore water pressure data and horizontal displacement data are acquired at each preset time point respectively to obtain foundation monitoring data, and the applied load intensity and the construction rate corresponding to each preset time point are acquired to obtain the construction parameters; and measuring the thickness, the initial pore ratio, the consolidation coefficient and the compression index of each foundation soil layer according to the geological section result to obtain the geological exploration data.
3. The neural network-based membrane bag sand dike slope settlement prediction method according to claim 1, wherein the method is characterized in that the compression amount of each foundation soil layer is taken as a target feature, and feature fusion is performed by analyzing the correlation between the construction parameters and the foundation monitoring data and the target feature in a time dimension, so as to obtain a fusion time sequence feature, which is specifically as follows: based on a preset domain knowledge base and the target features, feature screening is carried out on the construction parameters and the foundation monitoring data in a feature association mode to obtain core association features; Mapping each core associated feature to a preset target time axis for feature alignment to obtain a core feature matrix; Based on an attention mechanism, calculating a correlation score among all core correlation features through a core feature matrix, and carrying out feature fusion on the core correlation features to obtain the fusion time sequence features.
4. The neural network-based membrane bag sand dike slope settlement prediction method according to claim 3, wherein the construction parameters and the foundation monitoring data are subjected to feature screening in a feature association mode based on a preset domain knowledge base and the target features to obtain core association features, and specifically: extracting a settlement amount response curve related to a membrane bag sand bank slope from the domain knowledge base, and screening a first correlation characteristic which is directly correlated with the target characteristic from the construction parameter and the foundation monitoring data based on the variable of the settlement amount response curve; Extracting causal relation taking the first association feature as a variable from the domain knowledge base, and screening a second association feature indirectly associated with the target feature from the construction parameter and the foundation monitoring data based on the causal relation; and carrying out outlier rejection and continuity correction on the first association feature and the second association feature to obtain a core association feature.
5. The neural network-based membrane bag sand bank slope settlement prediction method according to claim 3, wherein the attention mechanism-based correlation score between each core correlation feature is calculated through a core feature matrix, and feature fusion is performed on the core correlation features to obtain the fused time sequence features, specifically: based on each time step of the target time axis, encoding a core feature matrix through an encoder to obtain an initial time sequence representation corresponding to each core associated feature; calculating a correlation score between each initial time sequence representation and the target feature, and normalizing the correlation score to obtain the attention weight of each core correlation feature; And if the attention weight is greater than a first threshold value, carrying out weighted summation on the corresponding initial time sequence representation to carry out feature fusion on the core associated features, and outputting the fused time sequence features.
6. The neural network-based membrane bag sand dike slope settlement prediction method according to claim 1, wherein the characteristic extraction and characteristic splicing are performed on the fusion time sequence characteristic and the geological exploration data, the characteristic splicing result is input into a preset settlement prediction model, the compression amount of each foundation soil layer is predicted by analyzing the pore drainage rate of each foundation soil layer, and the settlement prediction result of the membrane bag sand dike slope is output, specifically: Coding the geological exploration data into semantic vectors with preset dimensions, and dividing the semantic vectors according to the types of foundation layers to obtain static geological features of each foundation layer; Extracting response characteristics corresponding to the static geological characteristics from the fusion time sequence characteristics by analyzing the influence of the static geological characteristics on the change trend of the foundation monitoring data, and performing characteristic splicing on the static geological characteristics and the response characteristics based on a preset time step to obtain characteristic splicing results; Inputting the characteristic splicing result into the settlement prediction model based on the long-short-period memory network, predicting the compression amount of each foundation layer by analyzing the pore drainage rate of each foundation layer, and outputting the settlement prediction result, wherein the static geological characteristic is taken as a constraint condition, and the prediction process is adjusted.
7. The neural network-based membrane bag dyke slope settlement prediction method according to claim 6, wherein the prediction of the compression amount of each foundation layer by analyzing the pore drainage rate of each foundation layer is performed, and the settlement prediction result is output specifically as follows: based on characteristic splicing results, predicting groundwater level data in a future time period according to the pore ratio of each foundation soil layer and the applied load intensity of the film bag sand embankment slope to obtain a predicted pore drainage rate; extracting a construction rate from a characteristic splicing result, and comparing the construction rate with the predicted pore drainage rate to obtain effective stress distribution of each foundation layer in the vertical direction; and predicting the compression amount distribution of each foundation soil layer according to the effective stress distribution and the actual thickness of each foundation soil layer, and outputting the settlement prediction result.
8. The neural network-based membrane bag dyke slope settlement prediction method according to claim 6, wherein the prediction process is adjusted by taking the static geological feature as a constraint condition, specifically: in the prediction process, using the static geological feature to carry out boundary detection on a predicted value of the response feature in a future time period; if the boundary detection is unqualified, adjusting the predicted value to be within a preset safety range, and predicting the compression amount distribution of each foundation layer by using the adjusted predicted value; the safety range is a set value range of the response characteristic.
9. The neural network-based membrane bag sand dike slope settlement prediction method according to claim 1, wherein the identifying the abnormal settlement risk of the membrane bag sand dike slope through a preset dynamic early warning mechanism according to the settlement prediction result, and outputting the settlement risk level of the membrane bag sand dike slope at each preset construction point is specifically as follows: drawing a settlement continuous distribution map of the predicted settlement amount of each preset construction point in space according to the settlement prediction result; Superposing a preset predicted settlement cloud chart and the settlement continuous distribution chart, and identifying a first preset construction point and an abnormal mode thereof, wherein the settlement gradient exceeds a first threshold value, and a second preset construction point and an abnormal mode thereof with asymmetric settlement amount at two sides of the axis of the dike according to superposition results, wherein the settlement gradient is the difference of the settlement amounts in a fixed horizontal distance; and respectively evaluating the structural hazard degrees at the first preset construction point and the second preset construction point according to the dynamic early warning mechanism and the abnormal mode to obtain the corresponding sedimentation risk level.
10. The membrane bag sand dike slope settlement prediction system based on the neural network is characterized by comprising a data acquisition module, a characteristic fusion module, a settlement prediction module and a risk prediction module; The data acquisition module is used for acquiring construction parameters, foundation monitoring data and geological exploration data of the membrane bag sand bank slope at a plurality of preset time points; The feature fusion module is used for carrying out feature fusion on the compression amount of each foundation soil layer serving as a target feature by analyzing the correlation between the construction parameters and the foundation monitoring data and the target feature in the time dimension to obtain a fusion time sequence feature; the settlement prediction module is used for carrying out feature extraction and feature splicing on the fusion time sequence features and the geological exploration data, inputting a feature splicing result into a preset settlement prediction model, predicting the compression amount of each foundation soil layer by analyzing the pore drainage rate of each foundation soil layer, and outputting a settlement prediction result of the membrane bag sand embankment slope; And the risk prediction module is used for identifying abnormal sedimentation risks of the membrane bag sand embankment slope through a preset dynamic early warning mechanism according to the sedimentation prediction result and outputting sedimentation risk grades of the membrane bag sand embankment slope at each preset construction point.

Description

Membrane bag sand dike slope settlement prediction method and system based on neural network Technical Field The application belongs to the technical field of prediction systems, and particularly relates to a membrane bag sand bank slope settlement prediction method and system based on a neural network. Background The membrane bag sand embankment slope is a flexible structure embankment slope formed by filling sand materials with geomembrane bags, has the advantages of high mechanical degree, high construction speed, good overall stability and strong adaptability, and is common in embankment slope projects such as coast protection and river bank reinforcement. On a soft soil foundation, the membrane bag sand embankment slope can serve as a load dispersion layer to strengthen the foundation and improve the bearing capacity of the foundation, so that the settlement of the membrane bag sand embankment slope is a core index for evaluating the stability of the embankment slope and engineering safety, and the settlement is large and has long duration time, so that the risk of cracking, uneven settlement and the like of the embankment slope can be caused. The current layering summation method is to perform geological section on the foundation to obtain a plurality of foundation layers, calculate the compression amount of each foundation layer respectively through construction parameters, and then accumulate to obtain the total settlement of the embankment slope. The method can accumulate the load of the whole foundation on the premise of being uniformly distributed, but in the actual construction process, because of structural differences of foundation layers, the drainage rates of all spaces are different under the same construction parameters, and the load is unevenly distributed, so that the total settlement of the membrane bag sand dike slope calculated by the method has larger error, and effective data support cannot be provided for the subsequent dike slope reinforcement and the construction rate adjustment. Disclosure of Invention The application provides a membrane bag sand dike slope settlement prediction method and system based on a neural network, which can solve the problem that the total settlement of a predicted membrane bag sand dike slope is not accurate enough because the influence of a dike slope foundation structure on the settlement is not fully considered in the construction process in the prior art. The first aspect of the application provides a membrane bag sand bank slope settlement prediction method based on a neural network, which comprises the following steps: acquiring construction parameters, foundation monitoring data and geological exploration data of a membrane bag sand bank slope at a plurality of preset time points; Taking the compression amount of each foundation layer as a target feature, and carrying out feature fusion by analyzing the correlation between the construction parameters and the foundation monitoring data and the target feature in the time dimension to obtain a fusion time sequence feature; Inputting the fusion time sequence characteristics into a preset settlement prediction model by taking the geological exploration data as constraint conditions, predicting the compression amount of each foundation soil layer by analyzing the pore drainage rate of each foundation soil layer, and outputting the settlement prediction result of the membrane bag sand embankment slope; and identifying abnormal sedimentation risks of the membrane bag sand embankment slope through a preset dynamic early warning mechanism according to the sedimentation prediction result, and outputting sedimentation risk grades of the membrane bag sand embankment slope at each preset construction point. According to the scheme, the settlement of the membrane bag sand embankment slope is difficult to accurately predict only by a traditional empirical formula and a single model under complex geological conditions, so that historical construction parameters, foundation monitoring data and other data related to the embankment slope settlement are adopted, the data from different sources are combined together through feature fusion, fusion time sequence features closely related to the embankment slope settlement are obtained, and the accuracy of subsequent prediction is effectively improved. And then using a settlement prediction model, knowing the load intensity applied to the embankment slope through construction parameters, and calculating the pore drainage rate of the foundation soil layer through foundation monitoring data, so as to analyze the effective stress distribution directly related to the settlement after eliminating the pore water stress, and further obtain the accurate compression quantity of each foundation soil layer after the pore water is discharged. And finally, judging a settlement prediction result by using a dynamic early warning mechanism, gradually identifying settlement anomalies at