Method, device and equipment for monitoring treatment effect of subsidence area, medium and program product
Through the Safeformer prediction model and safety level identification model, based on the next-time prediction of fiber monitoring data, the problem of insufficient reliability and accuracy of ground stability monitoring in subsidence areas in the prior art is solved, and more efficient monitoring of subsidence area governance effect is achieved.
Patent Information
- Application Number
- CN202411740808.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-29
- Publication Date
- 2025-05-06
AI Technical Summary
In the ground stability monitoring of subsidence areas, the prior art directly predicts security based on current data, and cannot effectively capture process change data, resulting in insufficient reliability and accuracy.
The Safeformer prediction model and security level identification model are used to obtain fiber monitoring sequence data in real time, encode it into an input matrix, feature extraction and security level prediction are performed, and the reliability of the prediction results is enhanced based on the predicted next-time data.
It improves the reliability and accuracy of monitoring of the governance effect of subsidence areas, and can more effectively capture the changing data during the governance process of subsidence areas, and provide more accurate safety level prediction.
Smart Images
Figure CN119939207A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of engineering safety monitoring, and in particular relates to a method, device, equipment, medium and program product for monitoring the effect of subsidence area treatment. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] The changes in the strata caused by groundwater extraction, underground mineral mining, natural gas extraction, and sediment compaction have caused cracks of varying widths and large-scale basin-like subsidence on the surface, resulting in surface movement and deformation, which seriously affects land development and utilization. In order to effectively repair the subsidence area, backfill technology is usually used to rebuild the stability and height of the ground. Before using the backfill area, it needs to be grouting reinforced. However, during the reinforcement process, the filling area may become unstable due to reasons such as uneven distribution of the grouting material in the underground voids. Therefore, it is necessary to monitor the mid-term and late-stage effects of the subsidence area governance throughout the life cycle.
[0004] Currently, for ground stability monitoring in subsidence areas, one method is to monitor deformation data in real time and determine whether there is a safety risk based on a set threshold. Another method is to use a deep learning model to predict whether there is a safety risk based on real-time monitoring of deformation and other data. However, since the safety issue in the subsidence area is a gradual accumulation process from quantitative change to qualitative change, directly predicting safety based on current data cannot capture effective process change data, and therefore has insufficient reliability and accuracy. Summary of the invention
[0005] To overcome the above-mentioned deficiencies of the prior art, the present invention provides a method, device, equipment, medium and program product for monitoring the effect of subsidence area treatment. Safety prediction is performed based on the predicted monitoring data at the next moment, thereby enhancing the reliability of the prediction result.
[0006] To achieve the above object, the first aspect of the present invention provides a method for monitoring the effect of subsidence area treatment, comprising the following steps:
[0007] Obtain optical fiber monitoring sequence data in real time and encode it into an input matrix;
[0008] Based on the Safeformer prediction model, the optical fiber monitoring sequence data for the next time period is predicted;
[0009] Extract features from the optical fiber monitoring sequence data for the next time period to obtain a feature value sequence;
[0010] According to the characteristic value sequence, based on the security level identification model, predicting the security level;
[0011] The training method of the Safeformer prediction model includes:
[0012] Obtaining optical fiber monitoring sequence sample data in real time and encoding it into multiple sets of input matrices and output matrices, each row of the input matrix represents monitoring data corresponding to a timestamp, and the input matrix is timestamped to obtain a corresponding output matrix;
[0013] The input matrix and the output matrix are used to obtain the Safeformer prediction model based on Transformer model training.
[0014] A second aspect of the present invention provides a device for monitoring the effect of subsidence area treatment, comprising the following steps:
[0015] A data acquisition module is configured to acquire optical fiber monitoring sequence data in real time and encode it into an input matrix;
[0016] A data prediction module is configured to predict the optical fiber monitoring sequence data for the next time period based on the Safeformer prediction model;
[0017] A feature extraction module is configured to extract features from the optical fiber monitoring sequence data in the next time period to obtain a feature value sequence;
[0018] A security prediction module is configured to predict the security level according to the feature value sequence and based on a security level recognition model;
[0019] The training method of the Safeformer prediction model includes:
[0020] Obtaining optical fiber monitoring sequence sample data in real time and encoding it into multiple sets of input matrices and output matrices, each row of the input matrix represents monitoring data corresponding to a timestamp, and the input matrix is timestamped to obtain a corresponding output matrix;
[0021] The input matrix and the output matrix are used to obtain the Safeformer prediction model based on Transformer model training.
[0022] A third aspect of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory stores computer instructions, and when the computer instructions are executed by the processor, the electronic device executes the method described.
[0023] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the program implements the method described when executed by a processor.
[0024] A fifth aspect of the present invention provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the cross-regional geographic entity data coordination processing method.
[0025] One or more of the above technical solutions are aimed at the application scenario of subsidence area management monitoring. First, the changes in the monitoring data at the next moment are predicted based on the transformer model, and then the safety level is evaluated based on the predicted data. Compared with the traditional prediction directly based on the current monitoring data, it is more reliable.
[0026] In addition, an improved model combining the ProbSparse Self-attention mechanism and the Time Self-attention was proposed for the Transformer model. The time decay factor was introduced when calculating the attention weight, so that the model can pay more attention to the features closer in time during prediction, thereby improving the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0028] Figure 1 Schematic diagram of an application scenario of a method for monitoring the effect of subsidence area treatment in an embodiment of the present invention;
[0029] Figure 2 It is an overall flow chart of the method for monitoring the effect of subsidence area treatment in an embodiment of the present invention;
[0030] Figure 3 A schematic diagram of input matrix elements in an embodiment of the present invention;
[0031] Figure 4 This is a schematic diagram of the probabilistic sparse self-attention mechanism in an embodiment of the present invention;
[0032] Figure 5 It is a structural block diagram of Safeformer in an embodiment of the present invention;
[0033] Figure 6 It is a structural diagram of PST Self-attention in an embodiment of the present invention;
[0034] Figure 7 It is a partial structural diagram of the Encoder of Safeformer in an embodiment of the present invention;
[0035] Figure 8FIG. 4 is another partial structural diagram of the Encoder of Safeformer in an embodiment of the present invention. DETAILED DESCRIPTION
[0036] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not intended to limit the scope of protection of the present application.
[0037] In the description of the embodiments of the present application, the term “including” and similar terms should be understood as open inclusion, that is, “including but not limited to.” The term “based on” should be understood as “based at least in part on.”
[0038] Traditional RNN and CNN models have the problem of gradient vanishing or gradient exploding when processing long sequence data, which limits their ability to model long-distance dependencies. Transformer uses a self-attention mechanism that can simultaneously consider the information of each position in the sequence, thereby better capturing long-distance dependencies and making it more effective in processing long sequence data. In addition, Transformer's parallel computing capability is superior to traditional loop structures. Traditional RNN models are difficult to parallelize due to sequential calculations, while Transformer, due to the characteristics of the self-attention mechanism, can calculate information at different positions in the sequence in parallel, greatly improving training speed and efficiency. In addition, Transformer introduces position encoding to retain the position information of sequence data, solving the problem of sequence data losing positional relationship in the model. The introduction of this position encoding enables the model to better understand and utilize the sequence structure of the input data.
[0039] Query and key are important concepts in Transformer when calculating attention scores. Query is used to measure the correlation or importance between each position (or token) in the input sequence and other positions. Key is used to match the query to determine which positions' input is important for the output of the current position. In the self-attention mechanism, both the query matrix (Q) and the key matrix (K) are obtained by linearly transforming the input matrix. Each row in the query matrix Q represents a query vector, and each row in the key matrix K represents a key vector. The query vector and the key vector are dot-producted. The dot product result is divided by a scaling factor (usually the square root of the dimension) and then passed through the softmax function to obtain the attention weight.
[0040] Principal Component Analysis (PCA) is a commonly used dimensionality reduction technique that aims to find the most important features in the data and transform the data into a more efficient and easier to process form. It helps reduce the data dimension and remove redundant information while retaining the maximum variance of the data.
[0041] Support Vector Machine (SVM) is a supervised learning algorithm used for classification and regression analysis. It can perform linear and nonlinear classification, and has the advantages of strong generalization ability and processing high-dimensional data. It is one of the widely used algorithms in the field of machine learning.
[0042] Figure 1 The application environment of one or more embodiments of the present invention is shown, and the grouting drilling during the treatment process is used to lay optical cables, so as to avoid the problems of difficult construction, difficult installation and high cost caused by the special deployment of monitoring drilling after the treatment is completed. The collected data is demodulated by the optical fiber demodulator and transmitted to the terminal through the network, and the deep learning prediction model is used in the processor to predict the safety level. One or more embodiments of the present invention provide a method for monitoring the treatment effect of a subsidence area, which is applied to a processor. The present invention can predict the data at the next moment based on the current optical fiber monitoring data, and then predict the safety level based on the data at the next moment. Specifically, the method includes a training phase and a monitoring phase. The training phase includes the training of the Safeformer prediction model and the training of the safety level identification model.
[0043] The training method of the Safeformer prediction model includes:
[0044] (1) Obtaining optical fiber sample data; the optical fiber sample data includes stress, temperature, deformation, etc.
[0045] (2) The collected data is preprocessed and encoded into multiple sets of input matrices and output matrices, where each row of the input matrix represents monitoring data corresponding to a timestamp, and the input matrix is timestamped to obtain a corresponding output matrix.
[0046] The data of the distributed optical fiber is encoded into the input matrix X as follows;
[0047]
[0048] (3) Using the input matrix and the output matrix, the Safeformer prediction model is trained based on the Transformer model.
[0049] like Figure 5 As shown, the input matrix X en And the output matrix X deEach includes 10 timestamps, where the output matrix X de is based on the input matrix X de Based on this model, the monitoring data of 4 time stamps can be predicted backwards based on the input matrix.
[0050] The Transformer model introduces a probabilistic sparse self-attention mechanism to filter query-key pairs with high attention scores: only those query-key pairs that make a major contribution to the attention score are calculated, and pairs with lower scores are ignored, thereby reducing the computational complexity.
[0051] In the original Transformer structure, the calculation formula of the Attention mechanism is expressed as follows:
[0052]
[0053] Among them, Q-query, K-key, V-value are composed of input matrix X and weight matrix W Q , W K and W V The calculation is as shown in formula (2):
[0054]
[0055] The probability matrix calculated in Attention reflects the correlation between the upper and lower parameter sequences at different times in the input matrix X. In the study of Informer, it was found that the sparse self-attention score forms a long-tail distribution, that is, a few dot product pairs contribute to the main attention, and other dot product pairs can be ignored. The probabilistic sparse self-attention mechanism measures the importance of each query vector query through KL divergence, and the calculation formula is as follows:
[0056]
[0057] Remove the constant term and define the sparsity of the i-th query vector as:
[0058]
[0059] To alleviate the potential numerical stability issues of the Log-Sum-Exp (LES) operation, the approximate formula (3) is obtained by using the maximum mean measurement as follows:
[0060]
[0061] Among them, q i and k jThey represent the i-th row and j-th column in the query matrix Q and key matrix K respectively. In the long-tail distribution, the maximum value can be used to approximate LSE.
[0062] Calculate the attention weight sparse matrix Identify the main key-query pairs. The ProbSparse Self-attention calculation formula is as follows:
[0063]
[0064] On the basis of introducing the ProbSparse Self-attention mechanism, in the scenario of monitoring the effect of subsidence area governance, special events related to the effect of subsidence area governance have a stronger dependency on the monitoring data in the most recent time. In order to improve the prediction accuracy, some embodiments also introduce time self-attention. For the same time series, when calculating the attention weight for subsequences close in time, a higher weight is given, that is, the weight of the value of the adjacent parameter is amplified according to the distance in time, thereby strengthening the influence of the adjacent sequence on the result and improving the prediction accuracy. The model framework is as follows Figure 5 As shown in the figure, it includes an encoder and a decoder. The ProbSparse Self-attention mechanism and the Time Self-attention mechanism are denoted as the PSTSelf-attention mechanism in the figure, as follows:
[0065] Introducing the time distance decay factor T ij :
[0066]
[0067] Among them, σ is a constant parameter used to control the decay speed, and i and j correspond to different timestamps in the time series.
[0068] Based on the time distance attenuation factor, the attention weight of formula (5) is adjusted. There are two ways to adjust the attention. The first way is shown in formula (8):
[0069]
[0070] The second method is shown in formula (9):
[0071]
[0072] L K is the length of the query vector, and d is the dimension of the input vector. In formula (8), Determine the query vector q i The key vector k with the greatest correlation j , then calculate T based on i and the determined j ij , get the i-th query vector q i Formula (9) introduces the time distance decay factor into the calculation process of the maximum correlation time. The above two calculation methods both increase the time distance decay factor to allow only the key to focus on the top-u queries that dominate, thereby improving the self-attention to the important parameters of the adjacent time steps.
[0073] The calculation formula for Time Self-attention is as follows:
[0074]
[0075] in, is a sparse matrix of the same size as q and contains only M t (q i , K)’s top-u queries.
[0076] The training method of the security level recognition model includes:
[0077] (1) Obtain optical fiber sample data and corresponding safety levels;
[0078] (2) performing preprocessing and feature extraction on the optical fiber sample data to obtain a feature value sequence; the preprocessing includes centering and standardization; and the feature extraction adopts a PCA algorithm;
[0079] Specifically, safety level labels are added to the eigenvalue sequence, and the safety level labels of the subsidence areas are defined as normal, dangerous level 3, dangerous level 2, and dangerous level 1, with the danger levels increasing in ascending order.
[0080] (3) According to the feature value sequence with security level labels, the security level recognition model is trained based on the SVM model.
[0081] The subsidence area treatment effect monitoring method specifically comprises the following steps:
[0082] Step 1: Obtain optical fiber monitoring sequence data in real time;
[0083] Step 2: preprocessing the optical fiber monitoring sequence data and encoding it into an input matrix;
[0084] Step 3: Based on the Safeformer prediction model, predict the optical fiber monitoring sequence data for the next time period;
[0085] Step 4: Extract features from the optical fiber monitoring sequence data in the next time period to obtain a feature value sequence;
[0086] Step 5: Predict the security level based on the security level identification model.
[0087] One or more embodiments of the present invention further provide a system for monitoring the effect of subsidence area treatment, comprising: a data acquisition module configured to acquire optical fiber monitoring sequence data in real time and encode it into an input matrix;
[0088] A data prediction module is configured to predict the optical fiber monitoring sequence data for the next time period based on the Safeformer prediction model;
[0089] A feature extraction module is configured to extract features from the optical fiber monitoring sequence data in the next time period to obtain a feature value sequence;
[0090] A security prediction module is configured to predict the security level according to the feature value sequence and based on a security level recognition model;
[0091] The training method of the Safeformer prediction model includes:
[0092] Obtaining optical fiber monitoring sequence sample data and encoding them into multiple groups of input matrices and output matrices, wherein each row of the input matrix represents monitoring data corresponding to a timestamp, and the input matrix is timestamped to obtain a corresponding output matrix;
[0093] The input matrix and the output matrix are used to obtain the Safeformer prediction model based on Transformer model training.
[0094] One or more embodiments of the present invention further provide an electronic device that can be used to implement the subsidence area treatment effect monitoring method in the above embodiments. The electronic device includes one or more processors, one or more memories coupled to the processors, and a communication module coupled to the processors.
[0095] The memory may include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, at least one of the following: read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, hard disk, compact disc (CD), digital video disc (DVD) or other magnetic storage and / or optical storage. Examples of volatile memories include, but are not limited to, at least one of the following: random access memory (RAM), or other volatile memories that will not last during the power outage. The computer program may be stored in the ROM. When the processor executes the computer program, the above-mentioned method for monitoring the effect of subsidence area governance is implemented.
[0096] In some embodiments, the program may be tangibly contained in a computer-readable medium, which may be included in a device (such as in a memory) or other storage device accessible by the device. The program may be loaded from the computer-readable medium to the RAM for execution. The computer-readable medium may include any type of tangible non-volatile memory, such as ROM, EPROM, flash memory, hard disk, and the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned method for monitoring the effect of subsidence area governance.
[0097] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a server or terminal, the process or function described in the embodiment of the present application is generated in whole or in part. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a server or terminal or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (such as a floppy disk, a hard disk and a tape, etc.), an optical medium (such as a digital video disk (digital video disk, DVD), etc.), or a semiconductor medium (such as a solid-state hard disk, etc.).
[0098] In addition, although each operation is described in a specific order, this should be understood as requiring such operation to be performed in the specific order shown or in a sequential order, or requiring that all illustrated operations should be performed to obtain desired results. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the application. Some features described in the context of a separate embodiment can also be implemented in a single implementation in combination. On the contrary, the various features described in the context of a single implementation can also be implemented in multiple implementations individually or in any suitable sub-combination mode.
[0099] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.
Claims
1. A method for monitoring the effect of subsidence area treatment, characterized in that: The following steps are involved: Obtain optical fiber monitoring sequence data in real time and encode it into an input matrix; Based on the Safeformer prediction model, the optical fiber monitoring sequence data for the next time period is predicted; Extract features from the optical fiber monitoring sequence data for the next time period to obtain a feature value sequence; According to the characteristic value sequence, based on the security level identification model, predicting the security level; The training method of the Safeformer prediction model includes: Obtaining optical fiber monitoring sequence sample data and encoding them into multiple groups of input matrices and output matrices, wherein each row of the input matrix represents monitoring data corresponding to a timestamp, and the input matrix is timestamped to obtain a corresponding output matrix; The input matrix and the output matrix are used to obtain the Safeformer prediction model based on the Transformer model training.
2. The method for monitoring the effect of subsidence area treatment according to claim 1, characterized in that: The Transformer model adopts a probabilistic sparse self-attention mechanism.
3. The method for monitoring the effect of subsidence area treatment according to claim 2, characterized in that: The attention weight calculation formula for the i-th query vector is: in, represents the sparsity of the i-th query vector, q i and k j represent the i-th row and j-th column in the query matrix Q and key matrix K respectively, i and j represent the position of the time step respectively.
4. The method for monitoring the effect of subsidence area treatment according to claim 1, characterized in that: The training method of the security level recognition model includes: Obtain fiber monitoring sequence sample data and corresponding safety levels; Extracting features from the optical fiber sample data to obtain a feature value sequence; According to the feature value sequence with security level labels, the security level recognition model is trained based on the SVM model.
5. A device for monitoring the effect of subsidence area treatment, characterized in that: The following steps are involved: A data acquisition module is configured to acquire optical fiber monitoring sequence data in real time and encode it into an input matrix; A data prediction module is configured to predict the optical fiber monitoring sequence data for the next time period based on the Safeformer prediction model; A feature extraction module is configured to extract features from the optical fiber monitoring sequence data in the next time period to obtain a feature value sequence; A security prediction module is configured to predict the security level according to the feature value sequence and based on a security level recognition model; The training method of the Safeformer prediction model includes: Obtaining optical fiber monitoring sequence sample data and encoding them into multiple groups of input matrices and output matrices, wherein each row of the input matrix represents monitoring data corresponding to a timestamp, and the input matrix is timestamped to obtain a corresponding output matrix; The input matrix and the output matrix are used to obtain the Safeformer prediction model based on the Transformer model training.
6. The device for monitoring the effect of subsidence area treatment according to any one of claim 5, characterized in that: The attention weight calculation formula for the i-th query vector is: in, represents the sparsity of the i-th query vector, q i and k j represent the i-th row and j-th column in the query matrix Q and key matrix K respectively, i and j represent the position of the time step respectively.
7. The device for monitoring the effect of subsidence area treatment according to any one of claim 6, characterized in that: The training method of the security level recognition model includes: Obtain fiber monitoring sequence sample data and corresponding safety levels; Extracting features from the optical fiber sample data to obtain a feature value sequence; According to the feature value sequence with security level labels, the security level recognition model is trained based on the SVM model.
8. An electronic device, characterized in that: The electronic device comprises a processor and a memory, wherein the memory stores computer instructions, and when the computer instructions are executed by the processor, the electronic device executes the method according to any one of claims 1 to 4.
9. A computer-readable storage medium having a computer program stored thereon, wherein the program implements the method according to any one of claims 1 to 4 when executed by a processor.
10. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, it implements the cross-regional geographic entity data coordination processing method described in any one of claims 1 to 4.