Arch dam overall deformation coordination monitoring method based on full convolutional neural network

Abstract

An arch dam overall deformation coordination monitoring method based on a full convolutional neural network comprises the following steps: S1, selecting a plurality of measuring points to construct an overall deformation measuring point monitoring network according to the arrangement condition of arch dam monitoring deformation measuring points; s2, collecting historical deformation monitoring data of the arch dam, and preparing overall deformation data samples of the arch dam at different moments in combination with the overall deformation monitoring network; s3, constructing a full convolutional neural network model for monitoring the overall deformation coordination of the arch dam; s4, performing unsupervised training on the full convolutional neural network by using the arch dam overall deformation data samples at different moments to obtain an arch dam overall deformation coordination monitoring model; and S5, for arch dam deformation at any moment, constructing overall deformation data according to the measuring points of the deformation monitoring network, and performing deformation coordination detection by using the trained full convolutional neural network to realize arch dam overall deformation coordination monitoring. According to the method, manual data annotation is not needed, and intelligent judgment on the overall deformation coordination of the arch dam can be realized based on historical operation data.

Description

Technical Field

[0001] The invention relates to a method for monitoring the overall deformation coordination of an arch dam based on a fully convolutional neural network. Background Art

[0002] Deformation monitoring is the most important means to understand the operating status of arch dams. Monitoring the overall deformation of arch dams is an important method to judge whether the dam is operating normally. Arch dam is a three-dimensional shell and a spatial hyperstatic structure. Under normal operation, its overall deformation is continuous, smooth, coordinated and symmetrical, that is, the overall deformation of the arch dam should be coordinated and unified. The coordination of overall deformation is an important indicator of the operating performance of arch dams. The discontinuity and incoordination of the overall deformation of arch dams indicate the abnormality of the operating performance of arch dams. Such abnormal phenomena can be used to monitor the safety of arch dam operation. At present, in engineering management practice, the overall deformation distribution map of arch dams is mainly drawn manually, and engineers manually judge whether the deformation is normal. The biggest disadvantage of this method is that the real-time judgment is poor, and the judgment cannot be separated from "people", so it is difficult to implement online monitoring. Another method for judging the coordination of the overall deformation of arch dams is to use computers for judgment, but the overall deformation of arch dams has many measurement points and complex patterns, and there is no good mathematical method to describe it so far. Summary of the invention

[0003] The purpose of the present invention is to provide a method for monitoring the overall deformation coordination of an arch dam based on a fully convolutional neural network, which does not require manual data annotation and can realize intelligent evaluation of the overall deformation coordination of the arch dam based on historical operation data.

[0004] In order to solve the above problems, the technical solution of the present invention is: A method for monitoring the overall deformation coordination of an arch dam based on a fully convolutional neural network comprises the following steps: S1: According to the arrangement of deformation monitoring points of the arch dam, multiple measurement points are selected to build an overall deformation monitoring point network; S2: Collect historical deformation monitoring data of the arch dam, combine it with the overall deformation monitoring network, and prepare data samples of the overall deformation of the arch dam at different times; S3: Construct a fully convolutional neural network model for monitoring the overall deformation coordination of arch dams; S4: The full convolutional neural network is trained unsupervisedly using the overall deformation data samples of the arch dam at different times to obtain the overall deformation coordination monitoring model of the arch dam; S5: For the deformation of the arch dam at any time, the overall deformation data is constructed according to the measuring points of the deformation monitoring network, and the deformation coordination detection is performed with the trained fully convolutional neural network to realize the overall deformation coordination monitoring of the arch dam.

[0005] The beneficial effects of the present invention are as follows: the present invention proposes an unsupervised method for monitoring the overall deformation coordination of an arch dam based on a fully convolutional neural network. The method does not require data labeling, does not require the specification of rules or logic, and can realize intelligent detection of deformation coordination according to the mutual relationship of the overall deformation, thereby providing an intelligent monitoring method for the safe operation of the arch dam. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The present invention will be further described below in conjunction with the accompanying drawings: Figure 1 It is an implementation flow chart of the present invention.

[0007] Figure 2 It is the measurement point diagram of the overall deformation of the arch dam.

[0008] Figure 3 This is an example of an arch dam overall deformation measurement point monitoring network.

[0009] Figure 4 This is an example of the overall deformation distribution of an arch dam. DETAILED DESCRIPTION

[0010] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0011] like Figure 1 As shown, a method for intelligently monitoring the overall deformation coordination of an arch dam comprises the following steps: S1: According to the arrangement of deformation monitoring points of the arch dam, multiple measurement points are selected to build an overall deformation monitoring point network; S2: Collect historical deformation monitoring data of the arch dam, combine it with the overall deformation monitoring network, and prepare data samples of the overall deformation of the arch dam at different times; S3: Construct a fully convolutional neural network model for monitoring the overall deformation coordination of arch dams; S4: The full convolutional neural network is trained unsupervisedly using the overall deformation data samples of the arch dam at different times to obtain the overall deformation coordination monitoring model of the arch dam; S5: For the deformation of the arch dam at any time, the overall deformation data is constructed according to the measuring points of the deformation monitoring network, and the deformation coordination is detected by the trained full convolutional neural network to realize the overall deformation coordination monitoring of the arch dam. The specific implementation process of step S1 is as follows: S1-1: Collect information on the arrangement of deformation monitoring points on the arch dam and analyze the arrangement of deformation monitoring points.

[0012] S1-2: Select multiple representative measuring points as representative points for monitoring the overall deformation of the arch dam, and form a monitoring point network for the overall deformation of the arch dam with the representative measuring points.

[0013] S1-3: The representative measuring points in S1-2 should satisfy the spatial relationship of approximately row and column distribution. Figure 2 This is an example of the selected arch dam overall deformation measurement point network.

[0014] S1-4: To meet the requirements of the convolution operator operation, the deformation measurement points should be distributed in the form of a two-dimensional matrix. Therefore, virtual measurement points are used to fill in the deformation monitoring network of the arch dam distributed in the form of a two-dimensional matrix, such as Figure 3 shown.

[0015] The specific implementation process of step S2 is as follows: S2-1: Collect historical deformation monitoring data of arch dams and select deformation data of representative measuring points as mentioned in S1-3.

[0016] S2-2: According to the monitoring frequency, the deformation array of all measuring points at a certain time t is taken as one data sample, denoted as u t ij , where i represents the row and j represents the column of the measurement point.

[0017] S2-3: To meet the operational requirements of the convolution operator, the values ​​of the array at the virtual point position in S2-2 are assigned 0.

[0018] S2-3: Figure 4 An example of the overall deformation data and distribution diagram at a certain moment.

[0019] The specific implementation process of step S3 is as follows: S3-1: Construct a fully convolutional neural network model. The fully convolutional neural network takes the deformation distribution of the arch dam at a certain moment as input and itself as the target output.

[0020] S3-2: The fully convolutional neural network described in S3-1 adopts the following form: 1) data input layer (taking the example in S2-3 as an example, the data dimension is 7×7); 2) convolution layer 1 (kernel size 5×5×1, number 16, step size 1) + Relu; 2) convolution layer 2 (kernel size 3×3×16, number 32, step size 1) + Relu; 3) maximum pooling layer 1 (kernel size 1×3×3, step size 2); 4) convolution layer 3 (kernel size 3×3×32, number 64, step size 1) + Relu; 5) convolution layer 4 (kernel size 3×3×64, number 128, step size 1) + Relu; 6) maximum pooling layer 2 (kernel size 1×2×2, step size 2); 7) transposed convolution layer 1 (kernel size 3×3×1, number 128, step size 2).

[0021] The specific implementation process of step S4 is as follows: S4-1: Standardize the deformation distribution data samples of the arch dam at different times: First, count the maximum and minimum values ​​of the deformation in all deformation samples, where the maximum value is recorded as u max , the minimum value is recorded as u min ; For any deformation value, normalize it and record the deformation v after normalization t ij =(u t ij -u min ) / (u max -u min ).

[0022] S4-2: Use the standardized deformation to train the fully convolutional neural network to obtain the optimized model.

[0023] S4-3: According to the training results, the standard deviation of the predicted results and the measured values ​​of all sample models is calculated, denoted as σ u0 ; Denormalize the standard deviation to get the absolute standard deviation σ u =σ u0 (u max -u min ).

[0024] The specific implementation process of step S5 is as follows: S5-1: For the deformation of the arch dam at any time, the deformation distribution is obtained according to the steps of S1 and S2, and the data is standardized according to the method in S4.

[0025] S5-2: Use the fully convolutional neural network trained in S4 to analyze the deformation of the arch dam at any time and obtain the predicted value of the overall deformation of the arch dam.

[0026] S5-3: Analyze the stability of the arch dam deformation measurement, and use Δu1 to represent the measurement volatility; analyze the arch dam measurement accuracy requirements under the corresponding monitoring method, and use Δu2 to represent the measurement accuracy.

[0027] S5-4: Δu=max(3σ u , Δu1, Δu2) is the evaluation criterion for the overall deformation coordination of the arch dam.

[0028] S5-5: The difference between the arch dam deformation predicted by the fully convolutional neural network and the measured deformation is examined. If the difference is greater than Δu, the measuring point is identified as an incoordinated deformation point.

[0029] The present invention uses the historical monitoring data of the overall deformation of the arch dam to construct an overall deformation coordination model of the arch dam through an unsupervised fully convolutional neural network, so as to achieve accurate and rapid judgment of the deformation coordination of the arch dam. The method automatically models the characteristics of the data distribution, abstracts the overall deformation characteristics of the arch dam into the fully convolutional neural network model, and does not require artificial rules or methods, and has recognition performance similar to that of human engineers. The results of the present invention can be directly used for the coordination monitoring of the deformation of the arch dam, and can improve the accuracy and comprehensiveness of the safety monitoring of the operation of the arch dam.

[0030] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for monitoring the overall deformation coordination of an arch dam based on a fully convolutional neural network, characterized in that: The steps include: S1: According to the arrangement of deformation monitoring points of the arch dam, multiple measurement points are selected to build an overall deformation monitoring point network; S2: Collect historical deformation monitoring data of the arch dam, combine it with the overall deformation monitoring network, and prepare data samples of the overall deformation of the arch dam at different times; S3: Construct a fully convolutional neural network model for monitoring the overall deformation coordination of arch dams; S4: The full convolutional neural network is trained unsupervisedly using the overall deformation data samples of the arch dam at different times to obtain the overall deformation coordination monitoring model of the arch dam; S5: For the deformation of the arch dam at any time, the overall deformation data is constructed according to the measuring points of the deformation monitoring network, and the deformation coordination detection is performed with the trained fully convolutional neural network to realize the overall deformation coordination monitoring of the arch dam.

2. According to claim 1, a method for monitoring the overall deformation coordination of an arch dam based on a fully convolutional neural network is characterized in that: In step S1, a plurality of measuring points are distributed according to a two-dimensional matrix, and for measuring points in the points of the two-dimensional matrix, virtual points are added to the missing points.

3. The method for monitoring the overall deformation coordination of an arch dam based on a fully convolutional neural network according to claim 1 is characterized in that: In step S2, the deformation array of all measuring points at a certain time t is taken as a data sample and is set to u t ij , where i represents the row and j represents the column of the measuring point.

4. The method for monitoring the overall deformation coordination of an arch dam based on a fully convolutional neural network according to claim 1 is characterized in that: In step S3, the fully convolutional neural network takes the deformation distribution of the arch dam at a certain moment as input and outputs itself as the target.

5. The method for monitoring the overall deformation coordination of an arch dam based on a fully convolutional neural network according to claim 1, characterized in that: In step S4, the method for training the data sample includes the following steps: S4-1: Standardize the deformation distribution data samples of the arch dam at different times: Count the maximum and minimum values ​​of deformation in all deformation samples, where the maximum value is recorded as u max , the minimum value is recorded as u min ; For any deformation value, normalize it and record the deformation v after normalization t ij =(u t ij -u min ) / (u max -u min ); S4-2: Use the standardized deformation to train the fully convolutional neural network to obtain the optimized model; S4-3: According to the training results, the standard deviation of the predicted results and the measured values ​​of all sample models is calculated, denoted as σ u0 ; Denormalize the standard deviation to get the absolute standard deviation σ u =σ u0 (u max -u min ).

Images