Tunnel construction collapse early warning method and system

Abstract

The invention relates to the technical field of collapse early warning, and particularly discloses a tunnel construction collapse early warning method and system, and the method achieves the real-time monitoring of the internal displacement of a tunnel through the deployment of distributed displacement sensors in the tunnel. And performing fine-grained time sequence analysis on the acquired position measurement data by adopting a time sequence analysis technology based on deep learning, capturing displacement mode characteristics among a plurality of measurement points of the tunnel at each time point, and further performing time sequence dynamic analysis on the displacement mode characteristics among the measurement points at each time point under a global view angle to obtain the displacement mode characteristics of the plurality of measurement points of the tunnel. Therefore, the global change trend of the displacement state in the tunnel is excavated, and intelligent early warning prompt of the tunnel collapse risk is realized. In this way, the accuracy and timeliness of tunnel construction collapse early warning can be effectively improved, and unnecessary project interruption and resource waste caused by false alarm or missing alarm are reduced.

Description

Technical Field

[0001] The present application relates to the technical field of landslide warning, and more specifically, to a tunnel construction landslide warning method and system. Background Art

[0002] With the acceleration of urbanization and the continuous expansion of transportation networks, tunnels, as important channels connecting different geographical areas, have become increasingly important in the modern transportation system. However, the safety risks faced during tunnel construction cannot be ignored, especially tunnel collapse accidents, which will not only seriously affect the progress of the project and cause huge economic losses, but may also threaten the lives of construction workers. Therefore, how to effectively predict and prevent tunnel collapse has become an important issue in tunnel construction safety management.

[0003] Traditional tunnel construction collapse warning methods mainly rely on geological surveys, empirical judgments and other means. Although these methods can reflect the stability of the tunnel to a certain extent, they have many limitations. For example, geological surveys can only provide a preliminary understanding of geological conditions before construction, and it is difficult to accurately predict changes in geological conditions during construction; empirical judgments are highly dependent on the personal experience and professional knowledge of engineers, which is highly subjective and lacks objectivity.

[0004] In recent years, with the rapid development of sensor technology and information technology, real-time monitoring of tunnel construction displacement has been achieved by deploying sensors in tunnels, providing a new technical means for tunnel collapse warning. However, the existing tunnel construction collapse warning system usually adopts a simple threshold comparison method, which sends out a warning signal when the monitored displacement exceeds the preset threshold. It lacks in-depth analysis of the time series change pattern of tunnel displacement and is prone to false alarms or missed alarms.

[0005] Therefore, an optimized tunnel construction collapse early warning method and system are expected. Summary of the invention

[0006] The present application provides a tunnel construction landslide warning method and system, which can effectively improve the accuracy and timeliness of tunnel construction landslide warning and reduce unnecessary project interruption and resource waste due to false alarms or missed alarms.

[0007] In a first aspect, a tunnel construction collapse early warning method is provided, comprising:

[0008] obtaining a data set of position measurement data collected by a distributed displacement sensor;

[0009] Performing fine-grained displacement pattern feature extraction on the data set of the position measurement data to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points;

[0010] Performing temporal dynamic propagation aggregation on the time queue of the semantic encoding feature vectors of the displacement pattern between the tunnel measurement points to obtain a temporal significant propagation aggregation feature vector of the displacement pattern;

[0011] Determining whether to generate a tunnel collapse warning prompt based on the displacement pattern time series significant propagation aggregated feature vector;

[0012] Among them, the time queue of the semantic coding feature vectors of the displacement pattern between the tunnel measurement points is subjected to temporal dynamic propagation aggregation, including: extracting the current semantic coding feature vector of the displacement pattern between the tunnel measurement points from the time queue of the semantic coding feature vectors of the displacement pattern between the tunnel measurement points; using the current semantic coding feature vector of the displacement pattern between the tunnel measurement points as a feature prompt gate, and performing feature modulation aggregation based on the temporal correlation strength on the time queue of the semantic coding feature vectors of the displacement pattern between the tunnel measurement points to obtain the temporal significant propagation aggregation feature vector of the displacement pattern.

[0013] The present application provides a tunnel construction collapse early warning method and system, which realizes real-time monitoring of the internal displacement of the tunnel by deploying distributed displacement sensors in the tunnel, and uses the time series analysis technology based on deep learning to perform fine-grained time series analysis on the collected position measurement data, so as to capture the displacement pattern characteristics between multiple measurement points in the tunnel at each time point, and then, through the time series dynamic analysis of the displacement pattern characteristics between the measurement points at each time point from a global perspective, the global change trend of the internal displacement state of the tunnel is excavated, so as to realize intelligent early warning prompts for tunnel collapse risks. In this way, the accuracy and timeliness of tunnel construction collapse early warning can be effectively improved, and unnecessary project interruptions and resource waste caused by false alarms or missed alarms can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.

[0015] Figure 1 This is a schematic flow chart of a tunnel construction landslide warning method according to an embodiment of the present application.

[0016] Figure 2 This is a data flow diagram of the tunnel construction landslide warning method according to an embodiment of the present application.

[0017] Figure 3 This is a schematic block diagram of a time queue of a semantically encoded feature vector of a displacement pattern between tunnel measurement points, in a tunnel construction collapse warning method according to an embodiment of the present application, in which fine-grained displacement pattern feature extraction is performed on a data set of position measurement data.

[0018] Figure 4 This is a schematic block diagram of a tunnel construction collapse warning method according to an embodiment of the present application, in which a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points is subjected to temporal dynamic propagation aggregation to obtain a temporal significant propagation aggregation feature vector of the displacement pattern.

[0019] Figure 5 In the tunnel construction collapse warning method of an embodiment of the present application, the semantic encoding feature vector of the displacement pattern between the current tunnel measurement points is used as a feature prompt gate, and the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points is subjected to feature modulation aggregation based on the temporal correlation strength to obtain a schematic block diagram of the displacement pattern temporal significant propagation aggregation feature vector.

[0020] Figure 6 This is a schematic flow chart of the tunnel construction landslide warning system according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0022] In response to the above technical problems, the technical concept of this application is: to achieve real-time monitoring of the internal displacement of the tunnel by deploying distributed displacement sensors in the tunnel, and to use deep learning-based time series analysis technology to perform fine-grained time series analysis on the collected position measurement data, so as to capture the displacement pattern characteristics between multiple measurement points in the tunnel at each time point, and then, by performing a time series dynamic analysis of the displacement pattern characteristics between the measurement points at each time point from a global perspective, the global change trend of the internal displacement state of the tunnel can be excavated, thereby realizing intelligent early warning of the risk of tunnel collapse. In this way, the accuracy and timeliness of tunnel construction collapse warnings can be effectively improved, and unnecessary project interruptions and resource waste caused by false alarms or missed alarms can be reduced.

[0023] Figure 1 This is a schematic flow chart of a tunnel construction landslide warning method according to an embodiment of the present application. Figure 2 This is a data flow diagram of the tunnel construction collapse warning method according to the embodiment of the present application. Figure 1 and Figure 2As shown, the tunnel construction collapse warning method includes: S1, obtaining a data set of position measurement data collected by a distributed displacement sensor; S2, performing fine-grained displacement pattern feature extraction on the data set of position measurement data to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points; S3, performing temporal dynamic propagation aggregation on the time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points to obtain a temporal significant propagation aggregation feature vector of displacement patterns; S4, determining whether to generate a tunnel collapse warning prompt based on the temporal significant propagation aggregation feature vector of displacement patterns.

[0024] Exemplarily, in step S1, a data set of position measurement data collected by a distributed displacement sensor is obtained. It should be understood that, considering the limitations of the traditional threshold comparison method, the present application continuously collects the displacement data of the internal structure of the tunnel to achieve continuous monitoring of the displacement change of the tunnel, and then performs landslide warning based on the temporal change trend of the tunnel displacement, thereby effectively improving the accuracy and timeliness of the warning. In addition, in order to fully understand the changes in the internal structure of the tunnel, the present application adopts a distributed sensor layout. In an embodiment of the present application, inside the tunnel, a group of sensor groups are set at fixed intervals along its extension direction (i.e., the longitudinal direction of the tunnel), and each group of the sensor groups includes a number of inclination sensors, and the several inclination sensors of each group are located on the same cross section perpendicular to the extension direction of the tunnel, and the spacing between them is kept consistent. In this way, by covering multiple cross sections of the tunnel, it is possible to achieve all-round monitoring of the displacement of the internal structure of the tunnel, ensuring the comprehensiveness and accuracy of the displacement measurement data.

[0025] Exemplarily, in step S2, fine-grained displacement pattern feature extraction is performed on the data set of the position measurement data to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points. It should be understood that through fine-grained feature extraction, the slight displacement changes of the internal structure of the tunnel at different time points can be captured. These slight changes may be early signs that the tunnel is about to collapse, but they may not be easy to find in the original data. By organizing the displacement pattern features at different time points into a time queue, the global change trend of the displacement state inside the tunnel can be analyzed in a time series dynamic manner. This global perspective analysis helps to discover the long-term evolution law of the displacement pattern, thereby more accurately predicting the stability of the tunnel. Traditional threshold comparison methods are prone to false positives or false negatives because they usually only focus on the displacement data at a single time point. Through fine-grained feature extraction and time series analysis, the displacement pattern features of multiple time points can be comprehensively considered, so as to more reliably judge the collapse risk of the tunnel and reduce unnecessary engineering interruptions and resource waste caused by false positives or false negatives.

[0026] In one embodiment, Figure 3As shown, fine-grained displacement pattern feature extraction is performed on the data set of the position measurement data to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points, including: S21, data segmentation of the data set of the position measurement data according to timestamps to obtain a time series of displacement measurement data subsets; S22, respectively extracting displacement pattern features between measurement points of each displacement measurement data subset in the time series of the displacement measurement data subset to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points.

[0027] Exemplarily, in step S21, the data set of the position measurement data is segmented according to the timestamp to obtain the time series of the displacement measurement data subset. It should be understood that directly analyzing a large-scale time series data set will not only lead to low computational efficiency, but also easily cause the loss of detailed information in the process of data time series analysis. Therefore, the present application further segments the data set of the position measurement data according to the timestamp to decompose the large-scale time series data set into a plurality of displacement measurement data subsets in units of independent timestamps. Each of the displacement measurement data subsets contains the displacement data collected for all measurement points in the tunnel at the same time point. In this way, the complexity of data processing can be effectively reduced, which is beneficial to parallel computing and distributed processing, and helps to ensure that more detailed information is retained in the process of data time series analysis while improving data processing efficiency.

[0028] Exemplarily, in step S22, the displacement pattern features between the measurement points of each displacement measurement data subset in the time series of the displacement measurement data subset are extracted respectively to obtain the time queue of the semantic encoding feature vector of the displacement pattern between the measurement points of the tunnel. It should be understood that in order to better reflect the spatial distribution characteristics of the internal structural displacement of the tunnel, the present application arranges each displacement measurement data subset according to the position distribution of the distributed displacement sensor, forms a time series of the displacement measurement distribution matrix, and intuitively displays the displacement changes of each monitoring point inside the tunnel in the form of a matrix, thereby realizing a visual display of the displacement state of the internal structure of the tunnel. It should be understood that, under normal circumstances, the displacement changes of each monitoring point inside the tunnel are not completely independent, but show a certain correlation, especially between adjacent monitoring points. Therefore, in order to effectively explore the mutual influence and correlation patterns between different monitoring points, the present application further uses a dilated convolutional neural network model to extract correlation features of each displacement measurement distribution matrix in the time series of the displacement measurement distribution matrix, and enhances the model's ability to capture spatial correlations in the displacement measurement distribution matrix through dilated convolution operations, so as to more accurately identify the spatial correlation pattern of the overall displacement state of the internal structure of the tunnel, and obtain the time queue of the semantically encoded feature vectors of the displacement pattern between the tunnel measurement points.

[0029] In one embodiment, the displacement pattern features between tunnel measurement points of each displacement measurement data subset in the time series of the displacement measurement data subset are extracted respectively to obtain a time queue of semantically encoded feature vectors of the displacement pattern between tunnel measurement points, including: constructing each displacement measurement data subset in the time series of the displacement measurement data subset into a time series of displacement measurement distribution matrices according to the position distribution of the distributed displacement sensors; and inputting each displacement measurement distribution matrix in the time series of the displacement measurement distribution matrix into a tunnel measurement point displacement pattern feature extractor based on a void convolutional neural network model to obtain a time queue of semantically encoded feature vectors of the displacement pattern between tunnel measurement points.

[0030] Exemplarily, in step S3, the time queue of the semantically encoded feature vectors of the displacement pattern between the measurement points of the tunnel is subjected to temporal dynamic propagation aggregation to obtain a temporal significant propagation aggregation feature vector of the displacement pattern. It should be understood that in order to capture the temporal variation trend of the overall displacement state of the internal structure of the tunnel, it is necessary to further perform a temporal analysis on the time queue of the semantically encoded feature vectors of the displacement pattern between the measurement points of the tunnel. In particular, considering that in the process of tunnel collapse risk analysis, the displacement pattern features at the current time point have a more important reference value, and the displacement information at the historical time points has a relatively low degree of influence. Therefore, the present application proposes a temporal analysis method based on an attention mechanism, which dynamically adjusts the temporal aggregation weight of the feature based on the correlation strength of the displacement pattern features of each historical time point relative to the current time point. In this way, more attention can be paid to the current displacement change trend, while the influence of historical displacement information can be reasonably considered to more accurately predict the collapse risk of the tunnel.

[0031] In one embodiment, Figure 4 As shown, the time queue of the semantic coding feature vectors of the displacement pattern between the tunnel measurement points is subjected to temporal dynamic propagation aggregation, including: S31, extracting the current semantic coding feature vector of the displacement pattern between the tunnel measurement points from the time queue of the semantic coding feature vectors of the displacement pattern between the tunnel measurement points; S32, using the current semantic coding feature vector of the displacement pattern between the tunnel measurement points as a feature prompt gate, and performing feature modulation aggregation based on the temporal correlation strength on the time queue of the semantic coding feature vectors of the displacement pattern between the tunnel measurement points to obtain the temporal significant propagation aggregation feature vector of the displacement pattern.

[0032] Exemplarily, in step S31, the semantic encoding feature vector of the displacement pattern between the current tunnel measurement points is extracted from the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points. That is, the semantic encoding feature vector of the displacement pattern between the current tunnel measurement points is extracted from the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points as the benchmark for feature propagation and update, and the focus is used to construct the perspective of the temporal propagation of the displacement pattern feature between the measurement points.

[0033] Exemplarily, in step S32, the semantically encoded feature vector of the displacement pattern between the current tunnel measurement points is used as a feature prompt gate, and the time queue of the semantically encoded feature vector of the displacement pattern between the tunnel measurement points is subjected to feature modulation aggregation based on the temporal correlation strength to obtain the temporal significant propagation aggregated feature vector of the displacement pattern. That is, by using the semantically encoded feature vector of the displacement pattern between the current tunnel measurement points as a feature prompt gate and subjecting the time queue to feature modulation aggregation based on the temporal correlation strength, the temporal variation trend of the displacement pattern inside the tunnel can be captured more accurately, the accuracy and timeliness of the tunnel collapse warning can be improved, false alarms and missed alarms can be reduced, and construction safety can be ensured.

[0034] In one embodiment, Figure 5 As shown, taking the current tunnel measurement point displacement pattern semantic coding feature vector as the feature prompt gate, the time queue of the tunnel measurement point displacement pattern semantic coding feature vector is subjected to feature modulation aggregation based on the temporal association strength to obtain the displacement pattern temporal significant propagation aggregated feature vector, including: S321, calculating the temporal association strength prompt value of each tunnel measurement point displacement pattern semantic coding feature vector in the time queue of the tunnel measurement point displacement pattern semantic coding feature vector relative to the current tunnel measurement point displacement pattern semantic coding feature vector to obtain the time queue of the tunnel measurement point displacement pattern temporal association strength prompt value; S322, based on the time queue of the tunnel measurement point displacement pattern temporal association strength prompt value, gated mask modulation is performed on the time queue of the tunnel measurement point displacement pattern semantic coding feature vector to obtain the time queue of the intensity modulated tunnel measurement point displacement pattern semantic coding feature vector; S323, forward temporal propagation aggregation is performed on the time queue of the intensity modulated tunnel measurement point displacement pattern semantic coding feature vector to obtain the displacement pattern temporal significant propagation aggregated feature vector.

[0035] Exemplarily, in step S321, the temporal correlation strength prompt value of each of the semantic coding feature vectors of the displacement pattern between tunnel measurement points in the temporal queue of the semantic coding feature vectors of the displacement pattern between tunnel measurement points relative to the semantic coding feature vectors of the displacement pattern between the current tunnel measurement points is calculated to obtain the temporal queue of the temporal correlation strength prompt value of the displacement pattern between tunnel measurement points. That is, by using the semantic coding feature vectors of the displacement pattern between the current tunnel measurement points as the feature prompt gate, by calculating the temporal correlation strength prompt value between the semantic coding feature vectors of the displacement pattern between other tunnel measurement points and the semantic coding feature vectors of the displacement pattern between the current tunnel measurement points, the relative importance of the displacement pattern features between the measurement points at each historical time point is evaluated, the contextual information that has a greater impact on the displacement pattern features between the current measurement points is identified, and a higher weight is given to the subsequent feature propagation. Specifically, the formula for calculating the temporal correlation strength prompt value of each of the semantic coding feature vectors of the displacement pattern between tunnel measurement points in the temporal queue of the semantic coding feature vectors of the displacement pattern between tunnel measurement points relative to the semantic coding feature vectors of the displacement pattern between the current tunnel measurement points to obtain the temporal queue of the temporal correlation strength prompt value of the displacement pattern between tunnel measurement points is as follows:

[0036] X={x1,x2,...,x i ,...,x p}

[0037]

[0038] Where X represents the time sequence of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points, x1, x2, x i and x p represent the first, second, i-th and current semantic coding feature vectors of the displacement pattern between tunnel measurement points in the time queue of the semantic coding feature vectors of the displacement pattern between tunnel measurement points respectively, p is the number of semantic coding feature vectors of the displacement pattern between tunnel measurement points, and denote the eigenvalue of the jth position of the semantic encoding feature vector of the displacement pattern between the i-th tunnel measurement points and the eigenvalue of the jth position of the semantic encoding feature vector of the displacement pattern between the current tunnel measurement points, exp(·) denotes a natural exponential function, L denotes the characteristic scale of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points, s i Represents the prompt value of the temporal correlation strength of the displacement pattern between the i-th tunnel measurement points.

[0039] Exemplarily, in step S322, based on the time queue of the temporal association strength prompt value of the displacement pattern between the tunnel measurement points, the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points is gated mask modulated to obtain the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points after intensity modulation. That is, in order to avoid the degradation of model performance due to a few extreme cases, the median is further extracted from the temporal association strength prompt value obtained above as the threshold switch, and the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points is subjected to feature intensity modulation, and the feature representation with weak correlation with the current displacement pattern between the tunnel measurement points is filtered out by using the gating mechanism, while paying more attention to the contextual information closely related to the characteristics of the displacement pattern between the current measurement points. Specifically, based on the time queue of the temporal association strength prompt value of the displacement pattern between the tunnel measurement points, the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points is gated mask modulated to obtain the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points after intensity modulation. The formula is as follows:

[0040] τ=median[{s1,s2,..,s i ..,s p}]

[0041]

[0042] Among them, x i The i-th tunnel measurement point displacement pattern semantic coding feature vector in the time queue representing the tunnel measurement point displacement pattern semantic coding feature vector, s1, s2, s i and p represents the temporal correlation strength prompt value of the displacement pattern between the first, second, i-th and p-th tunnel measurement points, respectively, where s p =1, median[·] means taking the median, τ means the temporal correlation strength gate threshold, mask(·) means the gate control unit, x i ′ represents the i-th semantically coded feature vector of the displacement pattern between tunnel measurement points after intensity modulation in the time queue of the semantically coded feature vector of the displacement pattern between tunnel measurement points after intensity modulation.

[0043] In one embodiment, based on the time queue of the temporal association strength prompt value of the displacement pattern between tunnel measurement points, the time queue of the semantic coding feature vector of the displacement pattern between tunnel measurement points is gated mask modulated to obtain the time queue of the semantic coding feature vector of the displacement pattern between tunnel measurement points after intensity modulation, including: extracting the median of the time queue of the temporal association strength prompt value of the displacement pattern between tunnel measurement points as the temporal association strength gate threshold; using the temporal association strength gate threshold as the switching threshold of the gating unit, inputting the time queue of the semantic coding feature vector of the displacement pattern between tunnel measurement points and the time queue of the temporal association strength prompt value of the displacement pattern between tunnel measurement points into the gating unit to obtain the time queue of the semantic coding feature vector of the displacement pattern between tunnel measurement points after intensity modulation.

[0044] In one embodiment, the temporal correlation strength gate threshold is used as the switching threshold of the gating unit, and the time queue of the semantic coding feature vector of the displacement pattern between tunnel measurement points and the time queue of the temporal correlation strength prompt value of the displacement pattern between tunnel measurement points are input into the gating unit to obtain the time queue of the semantic coding feature vector of the displacement pattern between tunnel measurement points after intensity modulation, including: if the temporal correlation strength prompt value of the displacement pattern between tunnel measurement points is less than the temporal correlation strength gate threshold, then the corresponding semantic coding feature vector of the displacement pattern between tunnel measurement points is set to 0 to obtain the semantic coding feature vector of the displacement pattern between tunnel measurement points after intensity modulation; if the temporal correlation strength prompt value of the displacement pattern between tunnel measurement points is greater than or equal to the temporal correlation strength gate threshold, then the temporal correlation strength prompt value is divided by the temporal correlation strength gate threshold as a weight coefficient and weighted modulated with the corresponding semantic coding feature vector of the displacement pattern between tunnel measurement points to obtain the semantic coding feature vector of the displacement pattern between tunnel measurement points after intensity modulation.

[0045] Exemplarily, in step S323, the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points after the intensity modulation is subjected to forward temporal propagation aggregation to obtain the temporal significant propagation aggregate feature vector of the displacement pattern. That is, the forward LSTM model is used to perform temporal association aggregation on the semantic encoding feature vector of the displacement pattern between the tunnel measurement points after the characteristic intensity modulation, so as to utilize the long-term memory ability of the LSTM model to capture the characteristic cumulative effect of the displacement pattern between the measurement points at each time point, and generate a temporal significant propagation aggregate feature vector of the displacement pattern, so as to reflect the global dynamic change of the displacement pattern between the measurement points. Specifically, the formula for performing forward temporal propagation aggregation on the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points after the intensity modulation to obtain the temporal significant propagation aggregate feature vector of the displacement pattern is as follows:

[0046] X′={x1′,x2′,...,x i′,...,x p ′}

[0047] x p+1 =LSTM(X′)

[0048] Among them, X′ represents the time sequence of the semantic encoding feature vector of the displacement pattern between tunnel measurement points after intensity modulation, x1′, x2′, x i ′ and x p ′ respectively represent the first, second, i-th and p-th semantic encoding feature vectors of the displacement pattern between tunnel measurement points after intensity modulation in the time queue of the semantic encoding feature vector of the displacement pattern between tunnel measurement points after intensity modulation, LSTM(·) represents the forward LSTM model, x p+1 The aggregated eigenvectors representing the significant propagation of the displacement pattern temporal series.

[0049] In one embodiment, the time queue of the semantically encoded feature vectors of the displacement pattern between the tunnel measurement points after intensity modulation is forward time-series propagated and aggregated to obtain the time-series significantly propagated aggregated feature vector of the displacement pattern, including: inputting the time queue of the semantically encoded feature vectors of the displacement pattern between the tunnel measurement points after intensity modulation into a feature dynamic propagation module based on a forward LSTM model to obtain the time-series significantly propagated aggregated feature vector of the displacement pattern.

[0050] Exemplarily, in step S4, based on the displacement pattern time series significant propagation aggregate feature vector, it is determined whether to generate a tunnel collapse warning prompt. In one embodiment, based on the displacement pattern time series significant propagation aggregate feature vector, it is determined whether to generate a tunnel collapse warning prompt, including: inputting the displacement pattern time series significant propagation aggregate feature vector into a collapse warning analysis module based on a classifier to obtain a warning analysis result, and the warning analysis result is used to indicate whether the tunnel collapse warning prompt is generated. It should be understood that the displacement pattern time series significant propagation aggregate feature vector has captured the global change trend of the displacement pattern inside the tunnel through fine-grained feature extraction and time series dynamic propagation aggregation. The displacement pattern time series significant propagation aggregate feature vector contains rich spatiotemporal information and can fully reflect the current state and historical changes of the tunnel. The collapse warning analysis module based on the classifier can use a machine learning algorithm to efficiently classify and predict the displacement pattern time series significant propagation aggregate feature vector. By training the model, it is possible to identify which feature combinations indicate a higher risk of collapse, thereby issuing a warning prompt in a timely manner to reduce false alarms and missed alarms. Inputting the displacement pattern time series significant propagation aggregate feature vector into the classifier can achieve automated decision-making. The classifier automatically determines whether an early warning is needed based on preset rules and trained models, reducing the uncertainty and subjectivity of human intervention and improving the reliability and consistency of the early warning.

[0051] In the technical solution of the present application, the semantic encoding feature vectors of displacement patterns between tunnel measurement points in the time queue of the semantic encoding feature vectors of displacement patterns between tunnel measurement points respectively represent the local neighborhood deep implicit correlation features between the displacement measurement values ​​of each tunnel measurement point at the same time point. Considering that the measurement performance of each displacement sensor in the distributed displacement sensor is different and the measurement performance of the same displacement sensor at different time points may also fluctuate, some interference components may be introduced when the time queue of the semantic encoding feature vectors of displacement patterns between tunnel measurement points is subjected to feature dynamic propagation encoding with the current node as the feature prompt gate, which makes the overall feature manifold representation of the displacement pattern time series significant propagation aggregation feature vector obtained by feature dynamic propagation aggregation in the high-dimensional feature space have fine-grained feature structure holes. The existence of fine-grained feature structure holes will not only make the semantic coverage of the features of the displacement pattern time series significant propagation aggregation feature vector corresponding to the class probability label insufficient, but also cause the outlier class regression inference mapping to miss the target, affecting the accuracy of the warning analysis result obtained by the input of the displacement pattern time series significant propagation aggregation feature vector into the landslide warning analysis module based on the classifier.

[0052] In the technical solution of the present application, the displacement pattern time series significant propagation aggregation feature vector is input into a landslide early warning analysis module based on a classifier to obtain an early warning analysis result, including:

[0053] Inputting the displacement pattern time series significant propagation aggregated feature vector into a pre-classifier based on a Softmax function to obtain a displacement pattern time series fine-grained class probability label vector;

[0054] Calculate the product between the displacement pattern temporal fine-grained class probability label vector and its transposed vector to obtain a displacement pattern temporal fine-grained full label domain modulation matrix;

[0055] Dividing the eigenvalues ​​of each position in the displacement pattern time series fine-grained full-label domain modulation matrix by the square root of the scale value of the displacement pattern time series significant propagation aggregated eigenvector to obtain the displacement pattern time series fine-grained full-label domain scale modulation matrix;

[0056] Inputting the displacement pattern temporal fine-grained full-label domain scale modulation matrix into a quasi-probability domain attention sparse module based on a multi-level gating function to obtain a sparse displacement pattern temporal fine-grained full-label domain scale modulation matrix;

[0057] Taking the displacement pattern time series significant propagation aggregate feature vector as a query feature vector, matrix multiplying the sparse displacement pattern time series fine-grained full-label domain scale modulation matrix with the displacement pattern time series significant propagation aggregate feature vector to obtain an optimized displacement pattern time series significant propagation aggregate feature vector;

[0058] The optimized displacement pattern time series significant propagation aggregate feature vector is input into the classifier-based landslide early warning analysis module to obtain the early warning analysis result.

[0059] The optimization process of the displacement pattern time series significant propagation aggregate feature vector is expressed as:

[0060]

[0061]

[0062] Among them, v i represents the eigenvalue of the ith position of the displacement pattern time series significant propagation aggregated eigenvector, L1 represents the scale value of the displacement pattern time series significant propagation aggregated eigenvector, Represents the eigenvalue of the ith position of the fine-grained class probability label vector of the displacement pattern time series, V p represents the temporal fine-grained class probability label vector of the displacement pattern, represents matrix multiplication, T represents the transpose of the vector, M p represents the displacement pattern temporal fine-grained full-label domain scale modulation matrix, θ represents the predetermined threshold, represents the eigenvalue of the ith position of the fine-grained full-label domain scale modulation matrix of the displacement pattern temporal series, λ represents a predetermined hyperparameter, mask represents a mask function, represents the eigenvalue of the i-th position of the fine-grained full-label domain scale modulation matrix of the sparse displacement pattern temporal series, M p ' represents the fine-grained full-label domain scale modulation matrix of the sparse displacement pattern temporal sequence, and V' represents the optimized displacement pattern temporal sequence significant propagation aggregate feature vector.

[0063] Therefore, by adding a class probability domain-level optimizable disturbance to the original feature vector to strengthen the dependency entanglement between different variables in the original feature vector based on label domain modulation, the source domain feature vector can be optimized based on class probability query scale modulation. In this way, the significant inherent modal information in the source domain features can be more effectively retained, and the interference components in the feature distribution can be effectively masked, thereby improving the adversarial robustness of the manifold expression of the feature vector. In this way, the accuracy of the warning analysis result obtained by the landslide warning analysis module based on the classifier is improved.

[0064] In another embodiment, another possible implementation method of performing fine-grained displacement pattern feature extraction on the data set of the position measurement data to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points is as follows: first, preprocess the position measurement data, including data cleaning, removing outliers and noise, filling missing values, and normalizing the data to the same scale to ensure the integrity and consistency of the data. Next, feature extraction is performed using a method combining frequency domain analysis and time domain feature extraction. Frequency domain analysis converts time series data into frequency domain representation through Fourier transform (FFT) or wavelet transform (WT), extracts frequency spectrum and power spectrum density; time domain feature extraction calculates the mean, standard deviation, maximum value, minimum value, peak-to-peak value and root mean square (RMS) of the time series to reflect the statistical characteristics of the data. At the same time, the relative displacement and relative displacement change rate between different measurement points are extracted to reflect the deformation of the internal structure of the tunnel. Then, the frequency domain features, time domain features and spatial features are fused to form a multi-dimensional feature vector, and feature dimensionality reduction is performed using principal component analysis (PCA) or linear discriminant analysis (LDA) to reduce the dimension of the features and improve computational efficiency. Next, deep learning models such as autoencoders or variational autoencoders (VAE) are used to semantically encode the fused multi-dimensional feature vectors to generate semantically encoded feature vectors of the displacement pattern between tunnel measurement points. Finally, the semantically encoded feature vectors of each time point are arranged in chronological order to form a time queue of semantically encoded feature vectors of the displacement pattern between tunnel measurement points, retaining the time information for subsequent time series analysis and prediction. Through this process, fine-grained displacement pattern features can be extracted from the position measurement data, and a time queue of semantically encoded feature vectors of the displacement pattern between tunnel measurement points can be generated, providing a solid foundation for tunnel collapse warning.

[0065] In another embodiment, another possible implementation method of performing temporal dynamic propagation aggregation on the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points to obtain the temporal significant propagation aggregation feature vector of the displacement pattern is as follows: first, the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points is divided into multiple sliding windows, each window contains a certain number of time points, and the size of the sliding window can be adjusted according to actual needs and data characteristics, for example, a sliding window containing 10 time points is selected, and a new window is formed by moving forward one time point each time. In each sliding window, the statistical features of the feature vector are calculated, such as the mean, standard deviation, maximum value, minimum value, median, etc. These statistical features can capture the overall trend and fluctuation of the feature vector in the window. At the same time, the feature vector in each sliding window is subjected to time series analysis, such as autocorrelation analysis, partial autocorrelation analysis, ARI MA model, etc., to extract temporal dependencies and periodic patterns. Then, the aggregated feature vectors in each sliding window are arranged in chronological order to form a new feature vector sequence, and the new feature vector sequence is subjected to temporal propagation aggregation using a recurrent neural network (RNN) or a long short-term memory network (LSTM). RNN and LSTM can capture long-term dependencies and help preserve time information and dynamic changes in time series. In the recursive aggregation process, the attention mechanism is introduced to dynamically adjust the feature weights of different time points. By calculating the attention weight of the feature vector at each time point, the features of important time points are highlighted and irrelevant or noise features are suppressed. Finally, the feature vector sequence after recursive aggregation is pooled, such as maximum pooling or average pooling, to further compress the feature dimension and extract global features. Pooling operations help retain the most important feature information and reduce computational complexity. The pooled feature vector is fused with other global features (such as the overall displacement trend of the tunnel, geological conditions, etc.) to form the final displacement pattern time series significant propagation aggregation feature vector. Through this process, the global change trend of the displacement pattern between tunnel measurement points can be effectively extracted from the time queue, providing strong support for tunnel collapse warning.

[0066] In summary, the tunnel construction collapse warning method according to the embodiment of the present application is explained, which realizes real-time monitoring of the internal displacement of the tunnel by deploying distributed displacement sensors in the tunnel, and uses the time series analysis technology based on deep learning to perform fine-grained time series analysis on the collected position measurement data, capturing the displacement pattern characteristics between multiple measurement points in the tunnel at each time point, and then, through the time series dynamic analysis of the displacement pattern characteristics between the measurement points at each time point from a global perspective, the global change trend of the internal displacement state of the tunnel is excavated, thereby realizing intelligent early warning prompts for tunnel collapse risks. In this way, the accuracy and timeliness of tunnel construction collapse warnings can be effectively improved, and unnecessary project interruptions and resource waste caused by false alarms or missed alarms can be reduced.

[0067] Figure 6 FIG. 1 is a schematic flow chart of a tunnel construction collapse warning system according to an embodiment of the present application. Figure 5 As shown, the tunnel construction collapse warning system 100 includes: a position measurement data acquisition module 110, which is used to obtain a data set of position measurement data collected by a distributed displacement sensor; a fine-grained displacement pattern feature extraction module 120, which is used to perform fine-grained displacement pattern feature extraction on the data set of position measurement data to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points; a displacement pattern temporal dynamic propagation aggregation module 130, which is used to perform temporal dynamic propagation aggregation on the time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points to obtain a displacement pattern temporal significant propagation aggregation feature vector; a tunnel collapse warning prompt module 140, which is used to determine whether to generate a tunnel collapse warning prompt based on the displacement pattern temporal significant propagation aggregation feature vector.

[0068] In one embodiment, the displacement pattern temporal dynamic propagation aggregation module is used to: extract the current displacement pattern semantic coding feature vector between tunnel measurement points from the time queue of the displacement pattern semantic coding feature vector between tunnel measurement points; use the current displacement pattern semantic coding feature vector between tunnel measurement points as a feature prompt gate, and perform feature modulation aggregation based on the temporal correlation strength on the time queue of the displacement pattern semantic coding feature vector between tunnel measurement points to obtain the displacement pattern temporal significant propagation aggregation feature vector.

[0069] In one embodiment, the fine-grained displacement pattern feature extraction module is used to: segment the data set of the position measurement data according to timestamps to obtain a time series of displacement measurement data subsets; and respectively extract the displacement pattern features between measurement points of each displacement measurement data subset in the time series of the displacement measurement data subset to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points.

[0070] Here, those skilled in the art will appreciate that the specific operations of the various modules and units in the above-mentioned tunnel construction collapse warning system have been described in detail above. Figures 1 to 5 The invention has been described in detail in the description of the tunnel construction collapse early warning method, and therefore, its repeated description will be omitted.

[0071] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0072] It should be understood that the specific examples in this article are only intended to help those skilled in the art better understand the embodiments of the present application, rather than to limit the scope of the embodiments of the present application.

[0073] It should also be understood that in the various embodiments of the present application, the size of the serial number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0074] It should also be understood that the various implementation methods described in this specification can be implemented individually or in combination, and the embodiments of the present application are not limited to this.

[0075] Unless otherwise stated, all technical and scientific terms used in the embodiments of the present application are the same as the meanings generally understood by those skilled in the art of the technical field of the present application. The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the scope of the present application. The term "and / or" used in the present application includes any and all combinations of one or more related listed items. The singular forms of "a kind of", "above" and "the" used in the embodiments of the present application and the appended claims are also intended to include majority forms, unless the context clearly indicates other meanings. In addition, the terms "first", "second" etc. are only used for description purposes and cannot be understood as indicating or suggesting relative importance.

[0076] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0077] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0078] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0079] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0080] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A tunnel construction collapse early warning method, characterized in that: include: obtaining a data set of position measurement data collected by a distributed displacement sensor; Performing fine-grained displacement pattern feature extraction on the data set of the position measurement data to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points; Performing temporal dynamic propagation aggregation on the time queue of the semantic encoding feature vectors of the displacement pattern between the tunnel measurement points to obtain a temporal significant propagation aggregation feature vector of the displacement pattern; Determining whether to generate a tunnel collapse warning prompt based on the displacement pattern time series significant propagation aggregated feature vector; Among them, the time queue of the semantic coding feature vectors of the displacement pattern between the tunnel measurement points is subjected to temporal dynamic propagation aggregation, including: extracting the current semantic coding feature vector of the displacement pattern between the tunnel measurement points from the time queue of the semantic coding feature vectors of the displacement pattern between the tunnel measurement points; using the current semantic coding feature vector of the displacement pattern between the tunnel measurement points as a feature prompt gate, and performing feature modulation aggregation based on the temporal correlation strength on the time queue of the semantic coding feature vectors of the displacement pattern between the tunnel measurement points to obtain the temporal significant propagation aggregation feature vector of the displacement pattern.

2. The tunnel construction collapse early warning method according to claim 1, characterized in that: Performing fine-grained displacement pattern feature extraction on the data set of the position measurement data to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points, including: Slicing the data set of the position measurement data according to the timestamps to obtain a time series of the displacement measurement data subsets; The displacement pattern features between the measurement points of each displacement measurement data subset in the time series of the displacement measurement data subset are extracted respectively to obtain the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points.

3. The tunnel construction collapse early warning method according to claim 2, characterized in that: Extracting the displacement pattern features between tunnel measurement points of each displacement measurement data subset in the time series of the displacement measurement data subset respectively to obtain the time queue of the semantic encoding feature vector of the displacement pattern between tunnel measurement points, including: constructing each displacement measurement data subset in the time series of the displacement measurement data subset into a time series of a displacement measurement distribution matrix according to the position distribution of the distributed displacement sensors; Each displacement measurement distribution matrix in the time series of the displacement measurement distribution matrix is ​​input into the tunnel measurement point displacement pattern feature extractor based on the void convolutional neural network model to obtain the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points.

4. The tunnel construction collapse early warning method according to claim 3 is characterized in that: Taking the semantic encoding feature vector of the displacement pattern between the current tunnel measurement points as a feature prompt gate, performing feature modulation aggregation based on the temporal association strength on the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points to obtain the temporal significant propagation aggregation feature vector of the displacement pattern, including: Calculate the temporal association strength prompt value of each of the semantic coding feature vectors of the displacement pattern between tunnel measurement points in the temporal queue of the semantic coding feature vectors of the displacement pattern between tunnel measurement points relative to the semantic coding feature vector of the displacement pattern between the current tunnel measurement points to obtain the temporal queue of the temporal association strength prompt values ​​of the displacement pattern between tunnel measurement points; Based on the time queue of the temporal association strength prompt value of the displacement pattern between the tunnel measurement points, gated mask modulation is performed on the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points to obtain the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points after intensity modulation; The time queue of the semantically encoded feature vector of the displacement pattern between the tunnel measurement points after the intensity modulation is forward-sequentially propagated and aggregated to obtain the displacement pattern time-series significant propagation aggregated feature vector.

5. The tunnel construction collapse early warning method according to claim 4, characterized in that: Based on the time queue of the temporal association strength prompt value of the displacement pattern between the tunnel measurement points, gated mask modulation is performed on the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points to obtain the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points after intensity modulation, including: Extracting the median of the time queue of the temporal association strength prompt value of the displacement pattern between the tunnel measurement points as the temporal association strength gate threshold; The temporal association strength gate threshold is used as the switching threshold of the gating unit, and the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points and the time queue of the temporal association strength prompt value of the displacement pattern between the tunnel measurement points are input into the gating unit to obtain the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points after intensity modulation.

6. The tunnel construction collapse early warning method according to claim 5, characterized in that: The temporal correlation strength gate threshold is used as the switch threshold of the gate control unit, and the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points and the time queue of the temporal correlation strength prompt value of the displacement pattern between the tunnel measurement points are input into the gate control unit to obtain the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points after the intensity modulation, including: If the temporal correlation strength prompt value of the displacement pattern between the tunnel measurement points is less than the temporal correlation strength gate threshold, the corresponding semantic coding feature vector of the displacement pattern between the tunnel measurement points is set to 0 to obtain the semantic coding feature vector of the displacement pattern between the tunnel measurement points after intensity modulation; If the prompt value of the temporal correlation strength of the displacement pattern between the tunnel measurement points is greater than or equal to the temporal correlation strength gate threshold, then it is divided by the temporal correlation strength gate threshold and used as a weight coefficient to perform weighted modulation with the corresponding semantic coding feature vector of the displacement pattern between the tunnel measurement points to obtain the semantic coding feature vector of the displacement pattern between the tunnel measurement points after intensity modulation.

7. The tunnel construction collapse early warning method according to claim 6, characterized in that: The forward time series propagation aggregation is performed on the time queue of the semantic coding feature vector of the displacement pattern between the tunnel measurement points after the intensity modulation to obtain the time series significant propagation aggregation feature vector of the displacement pattern, including: The time queue of the semantically encoded feature vector of the displacement pattern between the tunnel measurement points after the intensity modulation is input into the feature dynamic propagation module based on the forward LSTM model to obtain the temporal significant propagation aggregation feature vector of the displacement pattern.

8. The tunnel construction collapse early warning method according to claim 7, characterized in that: Determining whether to generate a tunnel collapse warning prompt based on the displacement pattern time series significant propagation aggregation feature vector includes: The displacement pattern time series significant propagation aggregate feature vector is input into a classifier-based collapse warning analysis module to obtain a warning analysis result, and the warning analysis result is used to indicate whether the tunnel collapse warning prompt is generated.

9. A tunnel construction collapse warning system, characterized in that: include: A position measurement data acquisition module, used to acquire a data set of position measurement data collected by a distributed displacement sensor; A fine-grained displacement pattern feature extraction module, used to perform fine-grained displacement pattern feature extraction on the data set of the position measurement data to obtain a time queue of semantically encoded feature vectors of displacement patterns between tunnel measurement points; A displacement pattern temporal dynamic propagation aggregation module is used to perform temporal dynamic propagation aggregation on the time queue of the displacement pattern semantic encoding feature vectors between the tunnel measurement points to obtain a displacement pattern temporal significant propagation aggregation feature vector; A tunnel collapse warning prompt module is used to determine whether to generate a tunnel collapse warning prompt based on the displacement mode time series significant propagation aggregation feature vector; Among them, the displacement pattern temporal dynamic propagation aggregation module is used to: extract the current displacement pattern semantic coding feature vector between tunnel measurement points from the time queue of the displacement pattern semantic coding feature vector between the tunnel measurement points; use the current displacement pattern semantic coding feature vector between the tunnel measurement points as the feature prompt gate, and perform feature modulation aggregation based on the temporal correlation strength on the time queue of the displacement pattern semantic coding feature vector between the tunnel measurement points to obtain the displacement pattern temporal significant propagation aggregation feature vector.

10. The tunnel construction collapse warning system according to claim 9, characterized in that: The fine-grained displacement pattern feature extraction module is used to: Slicing the data set of the position measurement data according to the timestamps to obtain a time series of the displacement measurement data subsets; The displacement pattern features between the measurement points of each displacement measurement data subset in the time series of the displacement measurement data subset are extracted respectively to obtain the time queue of the semantic encoding feature vector of the displacement pattern between the tunnel measurement points.

Images