Foundation pit settlement prediction method, device and equipment and storage medium

Abstract

The invention discloses a foundation pit settlement prediction method, device and equipment and a storage medium, and the method comprises the steps: inputting foundation pit settlement monitoring data into a preset foundation pit settlement reasoning model, so as to enable the foundation pit settlement reasoning model to output the foundation pit settlement data of a to-be-predicted measurement point; wherein the training of the foundation pit settlement reasoning model comprises the following steps: analyzing historical foundation pit settlement monitoring data, and training a preset neural network model according to an analysis result to obtain the foundation pit settlement reasoning model; wherein the foundation pit settlement reasoning model comprises a spatial information processing module and a representation sequence fusion module; the spatial information processing module is used for generating a corresponding representation sequence according to the time sequence of the key data; and the representation sequence fusion module is used for calculating the reasoning value of the foundation pit settlement data of the to-be-predicted measurement point. According to the invention, the foundation pit settlement monitoring data can be fully analyzed and utilized, and long-term global reasoning of the foundation pit settlement of the to-be-measured measurement point is realized.

Description

Technical Field

[0001] The present invention relates to the field of underground engineering technology, and in particular to a foundation pit settlement prediction method, device, equipment and storage medium. Background Art

[0002] At present, the construction of underground rail transit is in full swing, resulting in more and more foundation pit projects appearing in densely populated urban facilities, surrounded by various underground pipelines, buildings, traffic arteries and subway tunnels. Due to the tight construction site, complex working conditions and high environmental protection requirements for surrounding facilities, the safety situation of subway construction is severe and has received close attention from the public. How to accurately and efficiently analyze and predict the settlement monitoring data of the entire construction process to ensure the safety and reliability of the surrounding facilities during the foundation pit excavation and construction process has become an issue that must be paid attention to in rail transit construction.

[0003] Scholars at home and abroad have conducted a lot of research on the deformation of retaining structures and surrounding soil and structures caused by foundation pit excavation in complex environments. In recent years, data-driven machine learning models have gradually been applied to the prediction of foundation pit engineering settlement. Among them, deep neural networks have attracted the attention of researchers for their powerful fitting ability and have gradually become the key research direction of foundation pit deformation prediction. After training, the deep learning model can achieve approximate prediction and reasoning of foundation pit settlement in a very short time. By establishing various neural network models such as multi-layer perceptrons, long short-term memory networks, and convolutional neural networks, the construction conditions and processes, historical monitoring data, soil layer survey results, groundwater levels and other information are input into the model to achieve dynamic prediction of foundation pit engineering settlement and reduce the risk of major safety accidents such as foundation pit collapse.

[0004] However, the existing deep learning algorithm model architecture lacks theoretical support, and data analysis and utilization are insufficient; most of them focus on time series prediction, and there is little exploration of the spatial correlation of soil bodies, and global spatial deduction is urgently needed. Summary of the invention

[0005] The present invention provides a foundation pit settlement prediction method, device, equipment and storage medium to solve the technical problems of insufficient data analysis and utilization of existing deep learning algorithms and lack of mining of spatial correlation of soil.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a method for predicting foundation pit settlement, comprising:

[0007] Acquire foundation pit settlement monitoring data of a measurement point that has the most similar foundation pit settlement pattern to the measurement point to be predicted; wherein the foundation pit settlement monitoring data includes: the settlement conditions of the ground surface and buildings within a preset range of the foundation pit where the measurement point is located;

[0008] Inputting the foundation pit settlement monitoring data into a preset foundation pit settlement reasoning model, so that the foundation pit settlement reasoning model can reason the foundation pit settlement data of the predicted measurement point according to the foundation pit settlement monitoring data, and output the foundation pit settlement data of the predicted measurement point;

[0009] The training of the foundation pit settlement inference model includes:

[0010] Obtain historical foundation pit settlement monitoring data at each measuring point;

[0011] Performing time autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis, and spatiotemporal consistency analysis on the historical foundation pit settlement monitoring data of each measuring point, and then determining the measuring point that is most similar to the foundation pit settlement pattern of the measuring point to be predicted according to the analysis results as the key measuring point, and selecting the historical foundation pit settlement monitoring data that has a causal relationship with the historical foundation pit settlement monitoring data of the measuring point to be predicted from the historical foundation pit settlement monitoring data of the key measuring point as the key data;

[0012] Taking the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output, a preset neural network model is trained to obtain the foundation pit settlement inference model;

[0013] Wherein, the foundation pit settlement reasoning model includes: a spatial information processing module and a representation sequence fusion module;

[0014] The spatial information processing module is used to generate a corresponding representation sequence according to the time series of the key data;

[0015] The representation sequence fusion module is used to fuse the representation sequences and calculate the inference value of the foundation pit settlement data of the measurement point to be predicted according to the fused representation sequence.

[0016] As a preferred solution, the temporal autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis and spatiotemporal consistency analysis of the historical foundation pit settlement monitoring data of each measuring point are performed, including:

[0017] According to the autocorrelation function and partial autocorrelation function, the temporal autocorrelation analysis of the historical foundation pit settlement monitoring data of each measuring point is carried out to obtain the autocorrelation between the historical foundation pit settlement monitoring data of each measuring point;

[0018] Based on dynamic time warping and clustering, the spatial similarity analysis of the historical foundation pit settlement monitoring data of each measuring point is carried out to obtain the similarity of the foundation pit settlement patterns between different measuring points;

[0019] According to the Granger causality test method, the spatiotemporal causality analysis of the historical foundation pit settlement monitoring data of each measuring point is carried out to obtain the causal relationship between the variables in the historical foundation pit settlement monitoring data of each measuring point;

[0020] According to the Spearman correlation coefficient matrix of the same measuring point in different periods, the spatiotemporal consistency analysis of the historical foundation pit settlement monitoring data of each measuring point was carried out to determine whether the settlement patterns of the same measuring point in different periods were consistent.

[0021] As a preferred solution, the corresponding characterization sequence is generated according to the following formula:

[0022] r t = relu(W r ·s t +b r );

[0023] Among them, r t is the representation sequence corresponding to the time series of key data, W r is the weight matrix, b r is the bias matrix, relu is the nonlinear activation function, s t is the foundation pit settlement value of the measuring point at historical time t.

[0024] As a preferred solution, the inference value of the foundation pit settlement data of the measurement point to be predicted is calculated by the following formula:

[0025] e=W e ·[r 1 ,…,r n ]+b e ;

[0026] Among them, e is the inference value of the foundation pit settlement data of the measurement point to be predicted, W e is the weight matrix; b e is the bias matrix; [r 1 ,…,r n ] means to merge and connect all representations.

[0027] As a preferred solution, the key data is used as input, and the historical foundation pit settlement data corresponding to the measurement point to be predicted is used as output to train a preset neural network model to obtain the foundation pit settlement inference model, including:

[0028] Get the preset hyperparameters;

[0029] According to the hyperparameters, a preset neural network model is trained with the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output. During the training process, the hyperparameters are optimized according to the asynchronous continuous halving algorithm and the grid search method, and finally the foundation pit settlement inference model is obtained through training.

[0030] As a preferred solution, after the foundation pit settlement inference model infers the foundation pit settlement data of the predicted measurement point based on the foundation pit settlement monitoring data and outputs the foundation pit settlement data of the predicted measurement point, it also includes:

[0031] According to the preset interpretable machine learning model, Shapley additive interpretation analysis is performed on the foundation pit settlement data of the measurement point to be predicted output by the foundation pit settlement reasoning model, and the influence of the input features of the foundation pit settlement reasoning model on the foundation pit settlement data of the measurement point to be predicted is calculated;

[0032] According to the influence of each input feature on the inferred foundation pit settlement data of the measurement point to be predicted when the input feature changes, the importance of each input feature is statistically obtained.

[0033] As a preferred solution, Shapley additive interpretation analysis is performed on the foundation pit settlement data of the measurement points to be predicted output by the foundation pit settlement inference model in the following manner:

[0034]

[0035] Where g is the interpretable machine learning model, x′ is the simplified data input that can be mapped to the original input data point, m is the number of input features, and φ j is the attribute value of each input feature, φ 0 is the mean of all sample prediction values, {x 1 ,…,x p} represents the set of all input features, {x 1 ,…,x p}\{x j} means division by x j The possible set of all input features except x (S) is the prediction for feature subset S.

[0036] Based on the above embodiment, another embodiment of the present invention provides a foundation pit settlement prediction device, including: a data acquisition module, a data prediction module and a model training module;

[0037] The data acquisition module is used to acquire the foundation pit settlement monitoring data of the measurement point that has the most similar foundation pit settlement pattern to the measurement point to be predicted; wherein the foundation pit settlement monitoring data includes: the settlement conditions of the ground surface and buildings within a preset range of the foundation pit where the measurement point is located;

[0038] The data prediction module is used to input the foundation pit settlement monitoring data into a preset foundation pit settlement reasoning model, so that the foundation pit settlement reasoning model can reason the foundation pit settlement data of the measurement point to be predicted based on the foundation pit settlement monitoring data, and output the foundation pit settlement data of the measurement point to be predicted;

[0039] The model training module is used to obtain the historical foundation pit settlement monitoring data of each measuring point; perform time autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis and spatiotemporal consistency analysis on the historical foundation pit settlement monitoring data of each measuring point, and then determine the measurement point that is most similar to the foundation pit settlement pattern of the measurement point to be predicted according to the analysis results as the key measurement point, and select the historical foundation pit settlement monitoring data that has a causal relationship with the historical foundation pit settlement monitoring data of the measurement point to be predicted from the historical foundation pit settlement monitoring data of the key measurement point as the key data; take the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output, train a preset neural network model to obtain the foundation pit settlement inference model; wherein, the foundation pit settlement inference model includes: a spatial information processing module and a representation sequence fusion module; the spatial information processing module is used to generate a corresponding representation sequence according to the time series of the key data; the representation sequence fusion module is used to fuse the representation sequence, and calculate the inference value of the foundation pit settlement data of the measurement point to be predicted according to the fused representation sequence.

[0040] Based on the above embodiments, another embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, the foundation pit settlement prediction method described in the above invention embodiment is implemented.

[0041] Based on the above embodiments, another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the foundation pit settlement prediction method described in the above invention embodiment.

[0042] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0043] The present invention provides a foundation pit settlement prediction method, which obtains foundation pit settlement monitoring data of a measurement point whose foundation pit settlement pattern is most similar to that of a measurement point to be predicted; inputs the foundation pit settlement monitoring data into a preset foundation pit settlement reasoning model, so that the foundation pit settlement reasoning model infers the foundation pit settlement data of the measurement point to be predicted based on the foundation pit settlement monitoring data, and outputs the foundation pit settlement data of the measurement point to be predicted.

[0044] The training of the foundation pit settlement inference model includes: obtaining the historical foundation pit settlement monitoring data of each measuring point; performing time autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis and spatiotemporal consistency analysis on the historical foundation pit settlement monitoring data of each measuring point, and then determining the measurement point that is most similar to the foundation pit settlement pattern of the measurement point to be predicted according to the analysis results as the key measurement point, and selecting the historical foundation pit settlement monitoring data that has a causal relationship with the historical foundation pit settlement monitoring data of the measurement point to be predicted from the historical foundation pit settlement monitoring data of the key measurement point as the key data; taking the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output, training a preset neural network model to obtain the foundation pit settlement inference model; wherein the foundation pit settlement inference model includes: a spatial information processing module and a representation sequence fusion module; the spatial information processing module is used to generate a corresponding representation sequence according to the time series of the key data; the representation sequence fusion module is used to fuse the representation sequence, and calculate the inference value of the foundation pit settlement data of the measurement point to be predicted according to the fused representation sequence.

[0045] When the foundation pit settlement inference model is trained, the present invention performs time autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis and spatiotemporal consistency analysis on the historical foundation pit settlement monitoring data of each measuring point, and can fully analyze and utilize the foundation pit settlement monitoring data, fully explore the internal mechanism and law of the foundation pit settlement data, thereby determining the measurement point that is most similar to the foundation pit settlement mode of the measurement point to be predicted, and the historical foundation pit settlement monitoring data that has a causal relationship with the historical foundation pit settlement monitoring data of the measurement point to be predicted. In addition, the foundation pit settlement inference model includes a spatial information processing module and a representation sequence fusion module, which generates a corresponding representation sequence according to the time series of key data, and can realize the comprehensive extraction of the spatiotemporal domain information of the foundation pit; then the representation sequence is fused, and then the inference value of the foundation pit settlement data of the measurement point to be predicted is calculated according to the fused representation sequence, so that the long-term global reasoning of the foundation pit settlement of the unknown settlement measurement point can be realized based on the monitoring data of the measurement point. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic flow chart of a method for predicting foundation pit settlement provided by one embodiment of the present invention;

[0047] Figure 2 It is a model training flow chart of the foundation pit settlement reasoning model of the present invention;

[0048] Figure 3 It is a schematic diagram of the model architecture of the foundation pit settlement reasoning model;

[0049] Figure 4 It is a schematic diagram of the hyperparameter tuning method of the foundation pit settlement inference model;

[0050] Figure 5 It is a structural schematic diagram of a foundation pit settlement prediction device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.

[0053] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is more than two, unless otherwise clearly and specifically defined.

[0054] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0055] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0056] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0057] In the description of the embodiments of the present application, unless otherwise clearly specified and limited, technical terms such as "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0058] Embodiment 1

[0059] Please refer to Figure 1 , is a flow chart of a method for predicting foundation pit settlement provided by an embodiment of the present invention, comprising the following specific steps:

[0060] S1. Acquire foundation pit settlement monitoring data of a measuring point that has the most similar foundation pit settlement pattern to the measuring point to be predicted; wherein the foundation pit settlement monitoring data includes: the settlement conditions of the ground surface and buildings within a preset range of the foundation pit where the measuring point is located;

[0061] First, the foundation pit settlement monitoring data of the adjacent measuring point whose foundation pit settlement pattern is most similar to the measuring point to be measured is obtained; wherein the foundation pit settlement monitoring data includes: the settlement conditions of the surface and buildings within a preset range of the foundation pit where the measuring point is located.

[0062] S2, inputting the foundation pit settlement monitoring data into a preset foundation pit settlement reasoning model, so that the foundation pit settlement reasoning model can reason the foundation pit settlement data of the predicted measurement point according to the foundation pit settlement monitoring data, and output the foundation pit settlement data of the predicted measurement point;

[0063] Then the acquired foundation pit settlement monitoring data is input into the pre-trained foundation pit settlement inference model, and the foundation pit settlement inference model infers the foundation pit settlement data of the measurement point to be measured based on the foundation pit settlement monitoring data, and outputs the current foundation pit settlement data of the measurement point to be measured. Specifically, for m monitoring points There are monitoring data for n days of history When the kth measurement point P k When data is missing or abnormal, the foundation pit settlement monitoring data of other measuring points for the latest few days (including the n+1th day) are input into the foundation pit settlement inference model, and the output is the measuring point P obtained by model inference. k Foundation pit settlement data on day n+1.

[0064] The training of the foundation pit settlement inference model includes:

[0065] Obtain historical foundation pit settlement monitoring data at each measuring point;

[0066] Performing time autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis, and spatiotemporal consistency analysis on the historical foundation pit settlement monitoring data of each measuring point, and then determining the measuring point that is most similar to the foundation pit settlement pattern of the measuring point to be predicted according to the analysis results as the key measuring point, and selecting the historical foundation pit settlement monitoring data that has a causal relationship with the historical foundation pit settlement monitoring data of the measuring point to be predicted from the historical foundation pit settlement monitoring data of the key measuring point as the key data;

[0067] Taking the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output, a preset neural network model is trained to obtain the foundation pit settlement inference model;

[0068] Wherein, the foundation pit settlement reasoning model includes: a spatial information processing module and a representation sequence fusion module;

[0069] The spatial information processing module is used to generate a corresponding representation sequence according to the time series of the key data;

[0070] The representation sequence fusion module is used to fuse the representation sequences and calculate the inference value of the foundation pit settlement data of the measurement point to be predicted according to the fused representation sequence.

[0071] Specifically, in the existing foundation pit settlement reasoning, the deep learning algorithm model architecture used for foundation pit settlement reasoning lacks theoretical support, data analysis and utilization are insufficient, and the model interpretability is insufficient; most of them can only achieve short-term reasoning, and have high requirements for monitoring data. When data is missing or wrong, the error is large, but data anomalies are very common in actual projects; most of them focus on time series reasoning, and there is little mining of the spatial correlation of soil, and the global deduction of space is urgently needed. Based on this, the present invention mainly solves the following three problems:

[0072] (1) Establish a complete data analysis process for the temporal and spatial correlation of foundation pit settlement, which can fully explore and extract the settlement mechanism, pattern and law within the monitoring data;

[0073] (2) Establish a deep learning model for settlement inference using data from neighboring measurement points to complete and detect missing, erroneous, and other abnormal monitoring data;

[0074] (3) Use advanced interpretable machine learning models to explain the output results of deep learning models.

[0075] In view of the above three problems to be solved, the present invention proposes a foundation pit settlement prediction method, and makes a series of improvements to the deep learning algorithm. The specific improvements are as follows:

[0076] (1) Through temporal autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis, and spatiotemporal consistency analysis of historical foundation pit settlement monitoring data, the inherent mechanism and law of foundation pit settlement data can be fully explored to assist the architecture design of deep learning models and provide reference for the construction methods of similar foundation pit projects;

[0077] (2) The constructed deep learning model includes a spatial information processing module and a representation sequence fusion module, which can realize the comprehensive extraction of spatiotemporal information of foundation pits and realize long-term inference of settlement amounts of unknown settlement measurement points based on historical monitoring data of neighboring measurement points, with fast speed and high accuracy.

[0078] (3) Different from general deep learning methods, the algorithm proposed in the present invention can also explain the reasoning results through an interpretable machine learning model, thereby enhancing the reliability of the reasoning results and being applicable to large-scale practical engineering projects.

[0079] Please refer to Figure 2 , which is a model training flow chart of the foundation pit settlement reasoning model of the present invention, the above-mentioned improved content, i.e., the model training process of the foundation pit settlement reasoning model is specifically described below:

[0080] Preferably, the historical foundation pit settlement monitoring data of each measuring point are subjected to temporal autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis and spatiotemporal consistency analysis, including: performing temporal autocorrelation analysis on the historical foundation pit settlement monitoring data of each measuring point according to the autocorrelation function and the partial autocorrelation function, so as to obtain the autocorrelation between the historical foundation pit settlement monitoring data of each measuring point; performing spatial similarity analysis on the historical foundation pit settlement monitoring data of each measuring point according to dynamic time warping and clustering, so as to obtain the similarity of the foundation pit settlement patterns between different measuring points; performing spatiotemporal causality analysis on the historical foundation pit settlement monitoring data of each measuring point according to the Granger causality test method, so as to obtain the causal relationship between the variables in the historical foundation pit settlement monitoring data of each measuring point; performing spatiotemporal consistency analysis on the historical foundation pit settlement monitoring data of each measuring point according to the Spearman correlation coefficient matrix of the same measuring point in different periods, so as to determine whether the settlement patterns of the same measuring point in different periods are consistent.

[0081] (1) Step 1: Data analysis of foundation pit settlement monitoring data

[0082] Data analysis mainly includes four parts: temporal autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis, and spatiotemporal consistency analysis.

[0083] Time autocorrelation analysis: Study the autocorrelation of settlement data at each measuring point to clarify the impact of historical settlement on future development, verify the feasibility of time series prediction and the input window length of historical data. The correlation coefficient between a sequence and its lagged sequence is calculated through the autocorrelation function (ACF) and partial autocorrelation function (PACF), and the autocorrelation of the sequence is evaluated. The larger the correlation coefficient, the stronger the autocorrelation:

[0084] ACF(k)=corr(x t ,x t-k ) (1)

[0085]

[0086] Among them, x t and x t-k They represent the original time series and the shifted time series respectively. and represents the minimum mean square error (MSE) regression result x t and x t-k The predicted value of , corr represents the correlation coefficient.

[0087] ACF quantifies the linear relationship between a time series and its lagged series, while PACF reveals the relationship between a time series and a series with a specific lagged order.

[0088] Spatial similarity analysis part: A clustering algorithm is introduced to evaluate the similarity of settlement patterns between different measuring points. Measuring points that show similar settlement patterns can make greater contributions to settlement reasoning. As the foundation pit is excavated from one side to the other, the surface settlement of different measuring points shows a nonlinear lag relationship. Therefore, dynamic time warping (DTW) and clustering are used to compare the similarity between the two sequences. Among them, DTW minimizes the total distance by dynamically adjusting the corresponding points between the two sequences along the time axis, and then clusters the sequences of different measuring points based on the distance measured by DTW. The working principle of DTW is shown in Table 1 below:

[0089]

[0090]

[0091] Table 1 DTW working principle

[0092] Spatial-temporal causality analysis section: Historical monitoring data from other measurement points may be valuable when making inferences and can improve the accuracy and robustness of predictions. The Granger Causality Test is used to analyze causal relationships in time series data and evaluate whether one variable has a causal effect on another variable based on the predictive performance of the time series. Granger defines causality based on two principles: (1) Cause precedes effect. (2) Cause contains unique information about the future value of its impact. Given these two assumptions about causality, Granger proposed testing the following hypothesis to determine the causal effect of x on y:

[0093]

[0094] in, represents probability, A is an arbitrary non-empty set, and They represent the available information in the whole set at time t and the available information in the modified set excluding x. If the above assumptions are accepted, then x produces Granger causality on y.

[0095] Spatiotemporal consistency analysis: Ensuring the consistency of the sedimentation pattern between a set of measurement points in two different periods is crucial to verifying the effectiveness of the model trained on the early data in the later stage. By comparing the Spearman correlation coefficient matrix of the two periods, the consistency of the sedimentation mechanism in the two different stages can be evaluated. The Spearman correlation coefficient ranges from -1 (strong negative correlation) to 1 (strong positive correlation), and close to 0 indicates no correlation between variables x and y. For a sample of size n, the n original data x are sorted by i ,y i Convert to rank(x i ),rank(y i ). The Spearman correlation coefficient is calculated as follows:

[0096] r s =corr(rank(x),rank(y)) (4)

[0097] Preferably, the corresponding characterization sequence is generated according to the following formula:

[0098] r t = relu(W r ·s t +b r );

[0099] Among them, r t is the representation sequence corresponding to the time series of key data, W r is the weight matrix, b r is the bias matrix, relu is the nonlinear activation function, s t is the foundation pit settlement value of the measuring point at historical time t.

[0100] Preferably, the inference value of the foundation pit settlement data of the measurement point to be predicted is calculated by the following formula:

[0101] e=W e ·[r 1 ,…,r n ]+b e ;

[0102] Among them, e is the inference value of the foundation pit settlement data of the measurement point to be predicted, W e is the weight matrix; b e is the bias matrix; [r 1 ,…,r n ] means to merge and connect all representations.

[0103] (2) Step 2: Determine the model architecture and train it

[0104] According to the conclusion of data analysis, appropriate data and model architecture are selected for training. Temporal autocorrelation analysis can determine the predictability of data and determine the time window length of input data; spatial similarity analysis can determine the most similar measurement points to the measurement points to be predicted, thereby determining the spatial source of input data; spatiotemporal causality analysis can determine which measurement points have meaningful data when making predictions, thereby screening out data suitable as input features; spatiotemporal consistency analysis can evaluate the feasibility of intelligent reasoning of the data set and determine the generalization of the method in different construction stages.

[0105] Due to the limited samples available for model training in foundation excavation projects, traditional time series processing models such as LSTM and GRU often encounter overfitting problems. To solve this problem, the proposed model is divided into two main modules: spatial information processing module and representation sequence fusion module. These two modules run in sequence, the former processes spatial information and the latter manages sequence information, and finally obtains the estimated surface settlement. Please refer to Figure 3 , which is a schematic diagram of the model architecture of the foundation pit settlement reasoning model.

[0106] The purpose of the spatial information processing module is to The working principle of this module can be abstracted as extracting the common settlement patterns of different stages from the surface settlement time series. Generate a representation sequence in in, and The subscript t indicates a specific historical moment, where s t Auxiliary measuring point Sedimentation value (s 1 ,…,s m ). The same neural network processes the data s from each auxiliary measurement point in turn. t , thus obtaining the output r of the representation t In this module, the MLP neural network is used, and the calculation process can be expressed as:

[0107] r t = relu(W r ·s t +b r ) (5)

[0108] Among them, W r is the weight matrix; b r is the bias matrix; relu is the nonlinear activation function.

[0109] In order to obtain the characterization sequence After that, the next step involves fusing these representations. Given their sequential nature, one possible fusion approach is to use Recurrent Neural Networks (RNN) to process the representation sequence. However, since these representations already contain fine-grained spatial information, each of them is very important. To alleviate the problems of gradient vanishing and forgetting that are common in RNN-type networks, these representations are concatenated and then passed through an MLP, outputting an estimate of the sedimentation e:

[0110] e=W e ·[r 1 ,…,r n ]+b e (6)

[0111] Where W e is the weight matrix; b e is the bias matrix; [r 1 ,…,r n ] means merging and connecting all representations.

[0112] Preferably, the key data is used as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted is used as output to train a preset neural network model to obtain the foundation pit settlement reasoning model, including: obtaining preset hyperparameters; according to the hyperparameters, the key data is used as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted is used as output to train a preset neural network model, and during the training process, the hyperparameters are optimized according to an asynchronous continuous halving algorithm and a grid search method, and finally the foundation pit settlement reasoning model is obtained through training.

[0113] After determining the model architecture, appropriate hyperparameters need to be selected during the training process. Hyperparameters are manually configured settings used during model training, and deep learning models often show sensitivity to these hyperparameters. A well-performed hyperparameter search aims to identify parameter combinations that optimize the performance of a given task, thereby improving the accuracy and generalization of the model. This paper adopts a comprehensive method that combines the Asynchronous Successive Halving Algorithm (ASHA) and the Grid Search Method for hyperparameter optimization. ASHA determines the approximate search space, and the Grid Search Method identifies the optimal hyperparameter combination. Please refer to Figure 4 , which is a schematic diagram of the hyperparameter tuning method of the foundation pit settlement inference model.

[0114] Preferably, after the foundation pit settlement reasoning model infers the foundation pit settlement data of the measurement point to be predicted based on the foundation pit settlement monitoring data and outputs the foundation pit settlement data of the measurement point to be predicted, it also includes: performing Shapley additive interpretation analysis on the foundation pit settlement data of the measurement point to be predicted output by the foundation pit settlement reasoning model according to a preset interpretable machine learning model, calculating the impact of changes in each input feature of the foundation pit settlement reasoning model on the inferred foundation pit settlement data of the measurement point to be predicted; and statistically obtaining the importance of each input feature based on the impact on the foundation pit settlement data of the measurement point to be predicted when each input feature changes.

[0115] Preferably, the foundation pit settlement data of the measurement point to be predicted output by the foundation pit settlement inference model is subjected to Shapley additive interpretation analysis according to the following method:

[0116]

[0117] Where g is the interpretable machine learning model, x′ is the simplified data input that can be mapped to the original input data point, m is the number of input features, and φ j is the attribute value of each input feature, φ 0 is the mean of all sample prediction values, {x 1 ,…,x p} represents the set of all input features, {x 1 ,…,x p}\{x j} means division by x j The possible set of all input features except x (S) is the prediction for feature subset S.

[0118] (3) Step 3: Use interpretable machine learning models to explain the inference results

[0119] In order to enhance the interpretability of the proposed model, Shapley additive explanations (SHAP) analysis is performed, which is a widely used technique in the field of interpretable machine learning. SHAP analysis originates from the concept of Shapley value in cooperative game theory. First, a reference value is established, which is usually the average regression value in the dataset. Then, for each feature, it calculates the impact on the model's inference value when the feature is changed in the dataset. The average of these impact values ​​represents the overall contribution of the feature to the regression result. SHAP analysis interprets Shapley values ​​as additive feature attributes and expresses the inference value as the sum of the attribute values ​​of each input feature:

[0120]

[0121] Where g is the explanation model; x′ is the simplified data input that can be mapped to the original input data points; m is the number of input features; φ j is the attribute value (Shapley value) of each feature; φ 0 is the mean of all sample prediction values; {x 1 ,…,x p} represents the set of all input features; {x 1 ,…,x p}\{x j} means division by x j The possible set of all input features except x (S) is the prediction for feature subset S.

[0122] The SHAP method is used to analyze the impact of all input features to illustrate the impact of each input feature on the final estimate. Generally speaking, most features have a negative impact on the results, that is, the smaller the eigenvalue, the greater the negative contribution. When neighboring points experience more obvious settlement, the settlement of the point to be inferred is usually greater. In addition, the SHAP method can also be used to perform importance analysis on input features to evaluate the most important part of all features from the data end. For example, the closer the data collected to the current moment is, the more important it is, and the more similar the settlement mechanism is to the point to be inferred, the more important the point is. Specifically, the SHAP method gives the impact of each feature on the result, and compares the importance of each type of feature by summing the impact of all data in each type of feature.

[0123] Embodiment 2

[0124] Please refer to Figure 5 , is a schematic diagram of the structure of a foundation pit settlement prediction device provided by an embodiment of the present invention, the device comprising: a data acquisition module, a data prediction module and a model training module;

[0125] The data acquisition module is used to acquire the foundation pit settlement monitoring data of the measurement point that has the most similar foundation pit settlement pattern to the measurement point to be predicted; wherein the foundation pit settlement monitoring data includes: the settlement conditions of the ground surface and buildings within a preset range of the foundation pit where the measurement point is located;

[0126] The data prediction module is used to input the foundation pit settlement monitoring data into a preset foundation pit settlement reasoning model, so that the foundation pit settlement reasoning model can reason the foundation pit settlement data of the measurement point to be predicted based on the foundation pit settlement monitoring data, and output the foundation pit settlement data of the measurement point to be predicted;

[0127] The model training module is used to obtain the historical foundation pit settlement monitoring data of each measuring point; perform time autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis and spatiotemporal consistency analysis on the historical foundation pit settlement monitoring data of each measuring point, and then determine the measurement point that is most similar to the foundation pit settlement pattern of the measurement point to be predicted according to the analysis results as the key measurement point, and select the historical foundation pit settlement monitoring data that has a causal relationship with the historical foundation pit settlement monitoring data of the measurement point to be predicted from the historical foundation pit settlement monitoring data of the key measurement point as the key data; take the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output, train a preset neural network model to obtain the foundation pit settlement inference model; wherein, the foundation pit settlement inference model includes: a spatial information processing module and a representation sequence fusion module; the spatial information processing module is used to generate a corresponding representation sequence according to the time series of the key data; the representation sequence fusion module is used to fuse the representation sequence, and calculate the inference value of the foundation pit settlement data of the measurement point to be predicted according to the fused representation sequence.

[0128] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0129] Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0130] Embodiment 3

[0131] Accordingly, an embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, the foundation pit settlement prediction method described in the above-mentioned embodiment of the invention is implemented.

[0132] The electronic device may be a computing device such as a desktop computer, a notebook, a palm computer, a cloud server, etc. The device may include, but is not limited to, a processor and a memory.

[0133] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the device, and various interfaces and lines are used to connect various parts of the entire device.

[0134] Embodiment 4

[0135] Accordingly, an embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the foundation pit settlement prediction method described in the above-mentioned embodiment of the invention.

[0136] The memory can be used to store the computer program, and the processor realizes various functions of the device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0137] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each method embodiment described above can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0138] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for predicting foundation pit settlement, characterized in that: include: Acquire foundation pit settlement monitoring data of a measurement point that has the most similar foundation pit settlement pattern to the measurement point to be predicted; wherein the foundation pit settlement monitoring data includes: the settlement conditions of the ground surface and buildings within a preset range of the foundation pit where the measurement point is located; Inputting the foundation pit settlement monitoring data into a preset foundation pit settlement reasoning model, so that the foundation pit settlement reasoning model can reason the foundation pit settlement data of the predicted measurement point according to the foundation pit settlement monitoring data, and output the foundation pit settlement data of the predicted measurement point; The training of the foundation pit settlement inference model includes: Obtain historical foundation pit settlement monitoring data at each measuring point; Performing time autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis, and spatiotemporal consistency analysis on the historical foundation pit settlement monitoring data of each measuring point, and then determining the measuring point that is most similar to the foundation pit settlement pattern of the measuring point to be predicted according to the analysis results as the key measuring point, and selecting the historical foundation pit settlement monitoring data that has a causal relationship with the historical foundation pit settlement monitoring data of the measuring point to be predicted from the historical foundation pit settlement monitoring data of the key measuring point as the key data; Taking the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output, a preset neural network model is trained to obtain the foundation pit settlement inference model; Wherein, the foundation pit settlement reasoning model includes: a spatial information processing module and a representation sequence fusion module; The spatial information processing module is used to generate a corresponding representation sequence according to the time series of the key data; The representation sequence fusion module is used to fuse the representation sequences and calculate the inference value of the foundation pit settlement data of the measurement point to be predicted according to the fused representation sequence.

2. The foundation pit settlement prediction method according to claim 1, characterized in that: The temporal autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis and spatiotemporal consistency analysis of the historical foundation pit settlement monitoring data of each measuring point include: According to the autocorrelation function and partial autocorrelation function, the temporal autocorrelation analysis of the historical foundation pit settlement monitoring data of each measuring point is carried out to obtain the autocorrelation between the historical foundation pit settlement monitoring data of each measuring point; Based on dynamic time warping and clustering, the spatial similarity analysis of the historical foundation pit settlement monitoring data of each measuring point is carried out to obtain the similarity of the foundation pit settlement patterns between different measuring points; According to the Granger causality test method, the spatiotemporal causality analysis of the historical foundation pit settlement monitoring data of each measuring point is carried out to obtain the causal relationship between the variables in the historical foundation pit settlement monitoring data of each measuring point; According to the Spearman correlation coefficient matrix of the same measuring point in different periods, the spatiotemporal consistency analysis of the historical foundation pit settlement monitoring data of each measuring point was carried out to determine whether the settlement patterns of the same measuring point in different periods were consistent.

3. The foundation pit settlement prediction method according to claim 1, characterized in that: According to the following formula, the corresponding representation sequence is generated: r t =relu(W r s t +b r ); Among them, r t is the representation sequence corresponding to the time series of key data, W r is the weight matrix, b r is the bias matrix, relu is the nonlinear activation function, s t is the foundation pit settlement value of the measuring point at historical time t.

4. The foundation pit settlement prediction method according to claim 1, characterized in that: The inference value of the foundation pit settlement data of the measurement point to be predicted is calculated by the following formula: e=W e ·[r1,…,r n ]+b e ; Among them, e is the inference value of the foundation pit settlement data of the measurement point to be predicted, W e is the weight matrix; b e is the bias matrix; [r1,…,r n ] means to merge and connect all representations.

5. The method for predicting foundation pit settlement according to claim 1, characterized in that: The method uses the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output to train a preset neural network model to obtain the foundation pit settlement inference model, including: Get the preset hyperparameters; According to the hyperparameters, a preset neural network model is trained with the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output. During the training process, the hyperparameters are optimized according to the asynchronous continuous halving algorithm and the grid search method, and finally the foundation pit settlement inference model is obtained through training.

6. The foundation pit settlement prediction method according to claim 1, characterized in that: After the foundation pit settlement inference model infers the foundation pit settlement data of the predicted measurement point according to the foundation pit settlement monitoring data and outputs the foundation pit settlement data of the predicted measurement point, it also includes: According to the preset interpretable machine learning model, Shapley additive interpretation analysis is performed on the foundation pit settlement data of the measurement point to be predicted output by the foundation pit settlement reasoning model, and the influence of the input features of the foundation pit settlement reasoning model on the foundation pit settlement data of the measurement point to be predicted is calculated; According to the influence of each input feature on the inferred foundation pit settlement data of the measurement point to be predicted when the input feature changes, the importance of each input feature is statistically obtained.

7. The method for predicting foundation pit settlement according to claim 6, characterized in that: The Shapley additive interpretation analysis is performed on the foundation pit settlement data of the predicted measurement points output by the foundation pit settlement inference model in the following way: Where g is the interpretable machine learning model, x′ is the simplified data input that can be mapped to the original input data point, m is the number of input features, and φ j is the attribute value of each input feature, φ0 is the mean of all sample prediction values, {x1,…,x p } represents the set of all input features, {x1,…,x p }\{x j } means division by x j The possible set of all input features except x (S) is the prediction for feature subset S.

8. A foundation pit settlement prediction device, characterized in that: include: Data acquisition module, data prediction module and model training module; The data acquisition module is used to acquire the foundation pit settlement monitoring data of the measurement point that has the most similar foundation pit settlement pattern to the measurement point to be predicted; wherein the foundation pit settlement monitoring data includes: the settlement conditions of the ground surface and buildings within a preset range of the foundation pit where the measurement point is located; The data prediction module is used to input the foundation pit settlement monitoring data into a preset foundation pit settlement reasoning model, so that the foundation pit settlement reasoning model can reason the foundation pit settlement data of the measurement point to be predicted based on the foundation pit settlement monitoring data, and output the foundation pit settlement data of the measurement point to be predicted; The model training module is used to obtain the historical foundation pit settlement monitoring data of each measuring point; perform time autocorrelation analysis, spatial similarity analysis, spatiotemporal causality analysis and spatiotemporal consistency analysis on the historical foundation pit settlement monitoring data of each measuring point, and then determine the measurement point that is most similar to the foundation pit settlement pattern of the measurement point to be predicted according to the analysis results as the key measurement point, and select the historical foundation pit settlement monitoring data that has a causal relationship with the historical foundation pit settlement monitoring data of the measurement point to be predicted from the historical foundation pit settlement monitoring data of the key measurement point as the key data; take the key data as input and the historical foundation pit settlement data corresponding to the measurement point to be predicted as output, train a preset neural network model to obtain the foundation pit settlement inference model; wherein, the foundation pit settlement inference model includes: a spatial information processing module and a representation sequence fusion module; the spatial information processing module is used to generate a corresponding representation sequence according to the time series of the key data; the representation sequence fusion module is used to fuse the representation sequence, and calculate the inference value of the foundation pit settlement data of the measurement point to be predicted according to the fused representation sequence.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for predicting foundation pit settlement as claimed in any one of claims 1 to 7 is implemented.

10. A storage medium, characterized in that: The storage medium includes a stored computer program, wherein when the computer program is executed, the device where the storage medium is located is controlled to execute the foundation pit settlement prediction method according to any one of claims 1 to 7.

Images