Method, device and storage medium for repairing distorted data of structural monitoring sensor network

Abstract

The present invention discloses a method, device and storage medium for repairing distorted data of a structure monitoring sensor network, wherein the method comprises: obtaining the structural response data of each structure monitoring sensor, and inputting the distorted identification model respectively to obtain the corresponding identification result; for the structure monitoring sensor whose identification result is distorted, obtaining the structural response data of each structure monitoring sensor associated therewith, and performing normalization preprocessing on it, and inputting the distorted data reconstruction model to obtain the normalized correction value of the structural response data of the distorted structure monitoring sensor; performing denormalization processing on the normalized correction value of the structural response data to obtain the correction value of the structural response data of the distorted structure monitoring sensor. The method can realize the online intelligent identification and data repair of abnormal behavior of the structure monitoring sensor network, the data reconstruction method is reasonable, the generalizability is strong, the reconstruction performance is excellent, and it can be widely applied to the intelligent operation and maintenance of engineering structures.

Description

Technical Field

[0001] The present invention relates to the field of structural performance monitoring and evaluation, and in particular to a method, device and storage medium for repairing distorted data of a structural monitoring sensor network. Background Art

[0002] my country's large and complex engineering structures, such as bridges and buildings, are huge in size. Under the influence of multiple factors such as complex environments and adverse loads, they are prone to structural damage and resistance attenuation, and even cause catastrophic accidents in extreme cases. In order to ensure the safety of the structure, major engineering structures need to install health monitoring systems to monitor their safety during construction and service in real time. The sensor network is the core of structural health monitoring, and ensuring its healthy operation is crucial to the effective acquisition of service environment and structural response data. However, due to the influence of noise, human interference, harsh environment, etc., signal drift, outliers, missing and other problems often occur during the sensor acquisition process, and the distortion of structural response data will affect the accurate assessment of the structural state. Therefore, how to identify the type of distorted data and improve the reconstruction accuracy of distorted data is the main problem currently faced.

[0003] Traditional data reconstruction methods mainly include deletion, interpolation and maximum likelihood model prediction. However, these methods mainly expect to obtain effective statistical inference and pay less attention to the prediction accuracy of missing data. In recent years, artificial neural network methods have been widely used in data prediction, but this method is difficult to reflect the correlation between sample data, and there are still problems such as complex parameter adjustment and overfitting. The recursive network model integrates the monotonous neural network model, and the adjacent nodes of its hidden layer are interconnected. The output of the hidden layer is jointly determined by the input at the current moment and the output at the previous moment. The recursive network has a good learning ability for short time series information, but as the length of the time series increases, its memory of long-distance information will continue to weaken, and there are problems such as gradient disappearance or explosion, making it difficult to model large-scale distorted data. Summary of the invention

[0004] The present invention provides a method, device and storage medium for repairing distorted data of a structure monitoring sensor network, so as to solve the problem that the existing method for reconstructing distorted structural response data collected by the structure monitoring sensor network has poor effect.

[0005] In a first aspect, a method for repairing distorted data of a structural monitoring sensor network is provided, comprising:

[0006] Obtain the structural response data of each structural monitoring sensor, and input them into the distortion identification model respectively to obtain the corresponding identification results;

[0007] For the structural monitoring sensors whose identification results are distorted, the structural response data of each structural monitoring sensor associated with it is obtained, and normalized preprocessed, and input into the distorted data reconstruction model to obtain the normalized correction value of the structural response data of the distorted structural monitoring sensor; wherein the distorted data reconstruction model is obtained by training the BLSTM based on the historical normal structural response data of each structural monitoring sensor;

[0008] The normalized correction value of the structural response data is denormalized to obtain the structural response data correction value of the distorted structural monitoring sensor.

[0009] Furthermore, the distortion recognition model is obtained by the following method:

[0010] Obtain normal structural response data samples and distorted structural response data samples including various types of faults, set corresponding labels, and construct a training sample set;

[0011] The one-dimensional convolutional neural network is trained based on the training sample set to obtain a distortion recognition model.

[0012] Furthermore, the types of various faults include: offset, drift, accuracy degradation, gain, and three complete faults of constant, constant + noise, and noise;

[0013] The one-dimensional convolutional neural network includes three convolutional pooling layers and one fully connected layer, and uses the Softmax function to output classification labels.

[0014] Furthermore, the distorted data reconstruction model is obtained by the following method:

[0015] Obtain historical normal structural response data of each structural monitoring sensor;

[0016] Each structural monitoring sensor is used as a structural monitoring sensor to be predicted, and its historical normal structural response data and the historical normal structural response data of the structural monitoring sensor associated with it are extracted, and the data are normalized and preprocessed to obtain multiple training sample sets corresponding to each structural monitoring sensor;

[0017] Based on each training sample set, the historical normal structural response data of the associated structural monitoring sensor is used as input, and the historical normal structural response data of the structural monitoring sensor to be predicted is used as output. The BLSTM is trained, and the particle swarm optimization algorithm is used to obtain the optimal BLSTM parameters to obtain the distorted data reconstruction model.

[0018] Furthermore, the process of normalizing and preprocessing the data includes:

[0019] Let y represent the historical normal structural response data of the structural monitoring sensor to be predicted, and X represent the historical normal structural response data of the associated structural monitoring sensor, as shown below:

[0020]

[0021] In the formula, represents the measurement value of the nth associated structural monitoring sensor at time t; y t is the measured value of the structural monitoring sensor to be predicted at time t; M is the total number of associated structural monitoring sensors, and T represents the sequence length of the structural response data;

[0022] The data is normalized according to the following formula so that its value is in the range of [0,1]:

[0023]

[0024]

[0025] In the formula, and are respectively the maximum and minimum values ​​of all the measured values ​​of the nth associated structural monitoring sensor at all times; and They are respectively the maximum and minimum values ​​of the measured values ​​of the sensor monitoring the structure to be predicted at all times.

[0026] Furthermore, the memory cell structure of the BLSTM includes a forget gate, an input gate, an output gate, a candidate state, a storage unit, and a unit output, which are respectively represented as follows:

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033] In the formula, f t 、i t , o t , c t and h t They represent the forget gate, input gate, output gate, candidate state, storage unit and unit output at time t respectively; h t-1 is the unit output at time t-1; Wf , W i , W o and W c are the corresponding input-state weight matrices; U f , U i , U o and U c are the corresponding state-state weight matrices respectively; b f , b i , b o and b c are the corresponding bias vectors respectively; σ and tanh are nonlinear activation functions.

[0034] Furthermore, the parameters of the BLSTM include the following seven: data sequence length, batch size, number of neurons in the hidden layer, number of neurons in the fully connected layer, dropout ratio, learning rate, and number of iterations;

[0035] The process of using the particle swarm optimization algorithm to obtain the optimal BLSTM parameters includes:

[0036] Initialize the initial positions and velocities of multiple particles, each of which is composed of the above 7 parameters; set the maximum number of iterations, learning factor and fitness function;

[0037] Iteratively update the position and velocity of each particle, and calculate the fitness value after each iteration until the maximum number of iterations is reached or the fitness value is less than the preset value, and the optimal BLSTM parameters are obtained;

[0038] Among them, the fitness function is expressed as follows:

[0039]

[0040] In the formula, L represents the fitness value, T represents the sequence length of the structural response data, represents the predicted value of the structural response data of the monitoring sensor of the structure to be predicted at time t, y t Represents the true value of the structural response data of the structural monitoring sensor to be predicted at time t.

[0041] Furthermore, it also includes:

[0042] The same distorted data reconstruction model is used to correct and reconstruct the distorted structural response data of the structural monitoring sensor multiple times, and the mean and confidence interval of the structural response data correction values ​​of the distorted structural monitoring sensor are obtained. The probability distribution of the structural response data correction value of the distorted structural monitoring sensor is constructed according to the following formula: Quantify the inherent uncertainty in reconstruction models of distorted data;

[0043]

[0044] Where N represents normal distribution; represents the mean value of the structural response data correction value of the distorted structural monitoring sensor at time t; t is the standard deviation of the corrected value of the structural response data of the distorted structural monitoring sensor at time t.

[0045] In a second aspect, a device for repairing distorted data of a structural monitoring sensor network is provided, comprising:

[0046] The distortion data identification module is used to obtain the structural response data of each structural monitoring sensor and input them into the distortion identification model to obtain the corresponding identification results;

[0047] A distorted data reconstruction module is used to obtain the structural response data of each structural monitoring sensor associated with the structural monitoring sensor whose identification result is distorted, and perform normalization preprocessing on the data, and input it into the distorted data reconstruction model to obtain the normalized correction value of the structural response data of the distorted structural monitoring sensor; wherein the distorted data reconstruction model is obtained by training the BLSTM based on the historical normal structural response data of each structural monitoring sensor;

[0048] The anti-normalization module is used to perform anti-normalization processing on the normalized correction value of the structural response data to obtain the structural response data correction value of the distorted structural monitoring sensor.

[0049] In a third aspect, a computer-readable storage medium is provided, which stores a computer program, and when the computer program is executed by a processor, the method for repairing distorted data of a structural monitoring sensor network as described above is implemented.

[0050] Beneficial Effects

[0051] Ensuring the healthy operation of the structural monitoring sensor network and improving the prediction accuracy of distorted data are the prerequisites for structural safety risk assessment. The present invention proposes a method, device and storage medium for repairing distorted data of a structural monitoring sensor network, including two parts: intelligent identification of distorted data and data repair. Among them, intelligent identification of distorted data is based on a trained neural network model for identification; the correction and reconstruction of distorted data mainly adopts a bidirectional long short-term memory network model (BLSTM), uses a memory cell gate structure to select and memorize effective information, and combines a particle swarm (PSO) algorithm to optimize the parameters of the bidirectional long short-term memory network model, thereby mining the time-varying laws of structural response data from both forward and reverse directions, thereby realizing accurate correction and reconstruction of distorted data. Moreover, when correcting and reconstructing distorted data, the data collected by the structural monitoring sensor associated with the distorted structural monitoring sensor is selected as the input of the distorted data reconstruction model, thereby improving the prediction accuracy of the model. Based on the above framework, online intelligent identification and data repair of abnormal behavior of the structural monitoring sensor network can be realized. The data reconstruction method of the present invention is reasonable, has strong generalizability, and has excellent reconstruction performance, and can be widely used in intelligent operation and maintenance of engineering structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 It is a flow chart of a method for repairing distorted data of a structure monitoring sensor network provided by an embodiment of the present invention;

[0054] Figure 2 It is a structural monitoring sensor network Benchmark model provided by an embodiment of the present invention;

[0055] Figure 3 is a schematic diagram of spatiotemporal correlation analysis of a small number of structural monitoring sensors provided by an embodiment of the present invention;

[0056] Figure 4 Schematic diagram of the structure of a bidirectional long short-term memory network model (BLSTM) provided in an embodiment of the present invention;

[0057] Figure 5 is a schematic diagram of the memory cell structure provided by an embodiment of the present invention;

[0058] Figure 6 It is a flow chart of the BLSTM parameter particle swarm optimization algorithm provided by an embodiment of the present invention;

[0059] Figure 7It is a schematic diagram of prediction results of a distorted data reconstruction model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0061] Example 1

[0062] This embodiment provides a method for repairing distorted data of a structural monitoring sensor network, including:

[0063] S1: Obtain the structural response data of each structural monitoring sensor and input them into the distortion recognition model to obtain the corresponding recognition results. The distortion recognition model is obtained by the following method:

[0064] S11: obtaining normal structural response data samples and distorted structural response data samples including various types of faults, setting corresponding labels, and constructing a training sample set;

[0065] S12: Train the one-dimensional convolutional neural network based on the training sample set to obtain a distortion recognition model.

[0066] In this embodiment, the types of various faults include: offset, drift, precision degradation, gain, and three complete faults: constant, constant + noise, and noise. That is, the sample types in the constructed training sample set include normal data and the above-mentioned seven types of fault distortion samples. In this embodiment, the structural response data corresponding to each sample is a time series, which is a one-dimensional vector of 1×1024, where 1 is the number of channels, 1024 is the number of sampling points, and the number of each sample is set to 1000; of course, in other embodiments, the sample data length, the number of sampling points, and the number of samples can be adjusted according to actual conditions. The one-dimensional convolutional neural network includes three convolutional pooling layers and one fully connected layer. The convolutional layer is used to extract features, and the pooling layer is used to compress features. After repeating three times, it enters the fully connected layer for classification, and the Softmax function is used to output the classification label.

[0067] After the distortion recognition model is trained, it is only necessary to input the structural response data collected by each structural monitoring sensor into the distortion recognition model, and then input the classification label of the structural response data collected by each structural monitoring sensor to realize the identification of the distorted structural response data and the corresponding distorted structural monitoring sensor.

[0068] S2: For the structural monitoring sensor whose identification result is distorted, the structural response data of each structural monitoring sensor associated with it is obtained, and the structural response data is normalized and preprocessed, and input into the distorted data reconstruction model to obtain the normalized correction value of the structural response data of the distorted structural monitoring sensor.

[0069] Before performing this step, the distorted data reconstruction model needs to be trained in advance. During implementation, each structural monitoring sensor has a corresponding distorted data reconstruction model. The model uses the normal structural response data collected by the structural monitoring sensor associated with the structural monitoring sensor to be predicted as input to predict the normal data of the structural monitoring sensor to be predicted. First, through the spatiotemporal correlation analysis of the structural monitoring sensor, the correlation between each structural monitoring sensor and other structural monitoring sensors is determined, and then the structural monitoring sensor with stronger correlation is selected as the associated structural monitoring sensor.

[0070] like Figure 2 As shown in Figure 1, a Benchmark model of a structural monitoring sensor network is provided. The structural monitoring sensors are acceleration sensors, and 16 acceleration sensors are selected for analysis. Figure 3 As shown in the figure, it is a schematic diagram of the spatiotemporal correlation analysis between the 16 acceleration sensors. As shown in the figure, the correlation coefficients between No. 2 and No. 4, No. 6 and No. 8, No. 10 and No. 12, and No. 14 and No. 16 are all 1, indicating that they have strong correlation. By analyzing the correlation coefficients between all sensors, the k sensors with the strongest correlation are selected as the associated structural monitoring sensors.

[0071] After determining the associated structural monitoring sensors of each structural monitoring sensor, the corresponding distorted data reconstruction model can be constructed. The specific process is as follows:

[0072] S21: Acquire historical normal structural response data of each structural monitoring sensor.

[0073] S22: taking each structural monitoring sensor as a structural monitoring sensor to be predicted, extracting its historical normal structural response data and the historical normal structural response data of the structural monitoring sensor associated with it, and performing normalization preprocessing on the data to obtain multiple training sample sets corresponding to each structural monitoring sensor. The normalization preprocessing process includes:

[0074] Let y represent the historical normal structural response data of the structural monitoring sensor to be predicted, and X represent the historical normal structural response data of the associated structural monitoring sensor, as shown below:

[0075]

[0076] In the formula, represents the measurement value of the nth associated structural monitoring sensor at time t; y t is the measured value of the structural monitoring sensor to be predicted at time t; M is the total number of associated structural monitoring sensors, and T represents the sequence length of the structural response data;

[0077] The data is normalized according to the following formula so that its value is in the range of [0,1]:

[0078]

[0079]

[0080] In the formula, and are respectively the maximum and minimum values ​​of all the measured values ​​of the nth associated structural monitoring sensor at all times; and They are respectively the maximum and minimum values ​​of the measured values ​​of the sensor monitoring the structure to be predicted at all times.

[0081] S23: Divide each training sample set into a corresponding training set and test set respectively, and based on each training set and each test set, take the historical normal structural response data of the associated structural monitoring sensor as input and the historical normal structural response data of the structural monitoring sensor to be predicted as output, train the BLSTM, and use the particle swarm optimization algorithm to obtain the optimal BLSTM parameters to obtain the distorted data reconstruction model.

[0082] Among them, the structure of the bidirectional long short-term memory network model (BLSTM) is as follows Figure 4 As shown in the figure, BLSTM adds reverse propagation to the forward propagation of the LSTM structure, that is, it transmits the state that exists in the future to the current moment, and the forward and reverse memory units jointly determine the output at the current moment. Figure 5 As shown in the figure, the memory cell structure of BLSTM includes a forget gate, an input gate, an output gate, a candidate state, a storage unit, and a unit output, which are respectively represented as follows:

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089] In the formula, f t 、i t , o t , c t and h t They represent the forget gate, input gate, output gate, candidate state, storage unit and unit output at time t respectively; h t-1 is the unit output at time t-1; W f , W i , W o and W c are the corresponding input-state (i.e. current input to current state) weight matrices; U f , U i , U o and U c are the corresponding state-state (i.e., previous state to current state) weight matrices; b f 、b i 、b o and b c are the corresponding bias vectors respectively; σ and tanh are nonlinear activation functions, specifically, σ is the logistic sigmoid nonlinear activation function.

[0090] The parameters of the BLSTM include the following seven: data sequence length, batch size, number of neurons in the hidden layer, number of neurons in the fully connected layer, dropout ratio, learning rate, and number of iterations. The particle swarm optimization algorithm is used to optimize the parameters of the BLSTM to obtain the optimal BLSTM parameters. The optimization process specifically includes:

[0091] Initialize the initial positions and velocities of multiple particles, each of which is composed of the above 7 parameters; set the maximum number of iterations, learning factor and fitness function;

[0092] Iteratively update the position and velocity of each particle, and calculate the fitness value after each iteration until the maximum number of iterations is reached or the fitness value is less than the preset value, and the optimal BLSTM parameters are obtained;

[0093] Among them, the fitness function is determined by the average mean square error of the training set and the test set, which is expressed as follows:

[0094]

[0095] In the formula, L represents the fitness value, T represents the sequence length of the structural response data, represents the predicted value of the structural response data of the monitoring sensor of the structure to be predicted at time t, y t Represents the true value of the structural response data of the structural monitoring sensor to be predicted at time t.

[0096] like Figure 6 As shown in FIG. 1 , it is a flowchart of the particle swarm optimization algorithm for optimizing parameters. For each particle in the particle swarm, they change dynamically as the optimization search proceeds. All particles in the particle swarm will approach the optimal position as the iteration progresses, and the maximum number of iterations is met or the fitness value is less than the preset value (the global optimal position meets the minimum limit), then the optimal hyperparameter combination of BLSTM is obtained.

[0097] S3: Denormalize the normalized correction value of the structural response data to obtain the distorted structural response data correction value of the structural monitoring sensor. Figure 7 It is a schematic diagram of prediction results of a distorted data reconstruction model provided by an embodiment of the present invention.

[0098] Preferably, it also includes:

[0099] S4: Integrate multiple distorted data reconstruction models (PSO-BLSTM models) to obtain the predicted probability distribution of the reconstructed data to quantify the inherent uncertainty of the distorted data reconstruction model. That is, use the same distorted data reconstruction model to correct and reconstruct the distorted structural response data of the structural monitoring sensor multiple times, and obtain the mean and confidence interval of the structural response data correction values ​​of the distorted structural monitoring sensor multiple times, and construct the probability distribution of the structural response data correction value of the distorted structural monitoring sensor according to the following formula: Quantify the inherent uncertainty in reconstruction models of distorted data;

[0100]

[0101] Where N represents normal distribution; represents the mean value of the structural response data correction value of the distorted structural monitoring sensor at time t; t is the standard deviation of the corrected value of the structural response data of the distorted structural monitoring sensor at time t.

[0102] Example 2

[0103] This embodiment provides a device for repairing distorted data of a structure monitoring sensor network, including:

[0104] The distortion data identification module is used to obtain the structural response data of each structural monitoring sensor and input them into the distortion identification model to obtain the corresponding identification results;

[0105] A distorted data reconstruction module is used to obtain the structural response data of each structural monitoring sensor associated with the structural monitoring sensor whose identification result is distorted, and perform normalization preprocessing on the data, and input it into the distorted data reconstruction model to obtain the normalized correction value of the structural response data of the distorted structural monitoring sensor; wherein the distorted data reconstruction model is obtained by training the BLSTM based on the historical normal structural response data of each structural monitoring sensor;

[0106] The anti-normalization module is used to perform anti-normalization processing on the normalized correction value of the structural response data to obtain the structural response data correction value of the distorted structural monitoring sensor.

[0107] Example 3

[0108] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the method for repairing distorted data of a structure monitoring sensor network as described above is implemented.

[0109] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0110] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0111] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0113] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0114] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0115] The present invention is different from the prior art route in that:

[0116] The present invention discloses a method, device and storage medium for repairing distorted data of a structural monitoring sensor network, including two parts: intelligent diagnosis of distorted data and data repair. Among them, the intelligent diagnosis of distorted data is based on a one-dimensional convolutional neural network model, and the convolution layer is used to perform feature learning on normal and multi-type distorted data, the pooling layer is used to compress the features, and the fully connected layer is used to connect all features, and the multi-type distorted data classification is realized through the output label; the data repair mainly adopts a bidirectional long short-term memory network model (BLSTM), and the memory cell gate structure is used to select and memorize effective information, and the particle swarm (PSO) algorithm is combined to optimize the parameters of the deep learning model, and then the time-varying law of sensor data is mined from the forward and reverse directions; multiple PSO-BLSTM models are integrated to obtain the probability distribution of their predicted output, so as to quantify the uncertainty of data reconstruction. Based on the above framework, intelligent detection and rapid repair of a small amount of sensor distorted data can be realized. The prediction method of the present invention is reasonable, has strong generalizability, and has excellent prediction performance, and can be widely used in intelligent operation and maintenance of engineering structures.

[0117] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for repairing distorted data of a structural monitoring sensor network, characterized in that: include: Obtain the structural response data of each structural monitoring sensor, and input them into the distortion identification model respectively to obtain the corresponding identification results; For the structural monitoring sensors whose identification results are distorted, the structural response data of each structural monitoring sensor associated with it is obtained, and normalized preprocessed, and input into the distorted data reconstruction model to obtain the normalized correction value of the structural response data of the distorted structural monitoring sensor; wherein the distorted data reconstruction model is obtained by training the BLSTM based on the historical normal structural response data of each structural monitoring sensor; Denormalizing the normalized correction value of the structural response data to obtain the structural response data correction value of the distorted structural monitoring sensor; The distortion identification model is obtained by the following method: Obtain normal structural response data samples and distorted structural response data samples including various types of faults, set corresponding labels, and construct a training sample set; The one-dimensional convolutional neural network is trained based on the training sample set to obtain a distortion recognition model; The distorted data reconstruction model is obtained by the following method: Obtain historical normal structural response data of each structural monitoring sensor; Each structural monitoring sensor is used as a structural monitoring sensor to be predicted, and its historical normal structural response data and the historical normal structural response data of the structural monitoring sensor associated with it are extracted, and the data are normalized and preprocessed to obtain multiple training sample sets corresponding to each structural monitoring sensor; Based on each training sample set, the historical normal structural response data of the associated structural monitoring sensor is used as input, and the historical normal structural response data of the structural monitoring sensor to be predicted is used as output. The BLSTM is trained, and the particle swarm optimization algorithm is used to obtain the optimal BLSTM parameters to obtain the distorted data reconstruction model; wherein BLSTM is a bidirectional long short-term memory network model.

2. The method for repairing distorted data of a structural monitoring sensor network according to claim 1, characterized in that: The types of various faults include: offset, drift, accuracy degradation, gain, and three complete faults: constant, constant + noise, and noise; The one-dimensional convolutional neural network includes three convolutional pooling layers and one fully connected layer, and uses the Softmax function to output classification labels.

3. The method for repairing distorted data of a structural monitoring sensor network according to claim 1, characterized in that: The process of normalizing and preprocessing the data includes: Let y represent the historical normal structural response data of the structural monitoring sensor to be predicted, and X represent the historical normal structural response data of the associated structural monitoring sensor, as shown below: In the formula, represents the measurement value of the nth associated structural monitoring sensor at time t; y t represents the measured value of the structural monitoring sensor to be predicted at time t; M represents the total number of associated structural monitoring sensors, and T represents the sequence length of the structural response data; The data is normalized according to the following formula so that its value is in the range of [0,1]: In the formula, and are respectively the maximum and minimum values ​​of all the measured values ​​of the nth associated structural monitoring sensor at all times; and They are respectively the maximum and minimum values ​​of the measured values ​​of the sensor monitoring the structure to be predicted at all times.

4. The method for repairing distorted data of a structural monitoring sensor network according to claim 1, characterized in that: The memory cell structure of the BLSTM includes a forget gate, an input gate, an output gate, a candidate state, a storage unit, and a unit output, which are respectively represented as follows: In the formula, f t 、i t , o t , c t and h t They represent the forget gate, input gate, output gate, candidate state, storage unit and unit output at time t respectively; h t-1 is the unit output at time t-1; W f , W i , W o and W c are the corresponding input-state weight matrices; U f , U i , U o and U c are the corresponding state-state weight matrices respectively; b f 、b i 、b o and b c are the corresponding bias vectors respectively; σ and tanh are nonlinear activation functions.

5. The method for repairing distorted data of a structural monitoring sensor network according to claim 1, characterized in that: The parameters of the BLSTM include the following seven: data sequence length, batch size, number of neurons in the hidden layer, number of neurons in the fully connected layer, dropout ratio, learning rate, and number of iterations; The process of using the particle swarm optimization algorithm to obtain the optimal BLSTM parameters includes: Initialize the initial positions and velocities of multiple particles, each of which is composed of the above 7 parameters; set the maximum number of iterations, learning factor and fitness function; Iteratively update the position and velocity of each particle, and calculate the fitness value after each iteration until the maximum number of iterations is reached or the fitness value is less than the preset value, and the optimal BLSTM parameters are obtained; Among them, the fitness function is expressed as follows: In the formula, L represents the fitness value, T represents the sequence length of the structural response data, represents the predicted value of the structural response data of the monitoring sensor of the structure to be predicted at time t, y t Represents the true value of the structural response data of the structural monitoring sensor to be predicted at time t.

6. The method for repairing distorted data of a structural monitoring sensor network according to claim 1, characterized in that: Also includes: The same distorted data reconstruction model is used to correct and reconstruct the distorted structural response data of the structural monitoring sensor multiple times, and the mean and confidence interval of the structural response data correction values ​​of the distorted structural monitoring sensor are obtained. The probability distribution of the structural response data correction value of the distorted structural monitoring sensor is constructed according to the following formula: Quantify the inherent uncertainty in reconstruction models of distorted data; Where N represents normal distribution; represents the mean value of the structural response data correction value of the distorted structural monitoring sensor at time t; t is the standard deviation of the corrected value of the structural response data of the distorted structural monitoring sensor at time t.

7. A device for repairing distorted data of a structural monitoring sensor network, characterized in that: The method for repairing distorted data of a structural monitoring sensor network according to claim 1 comprises: The distortion data identification module is used to obtain the structural response data of each structural monitoring sensor and input them into the distortion identification model to obtain the corresponding identification results; A distorted data reconstruction module is used to obtain the structural response data of each structural monitoring sensor associated with the structural monitoring sensor whose identification result is distorted, and perform normalization preprocessing on the data, and input it into the distorted data reconstruction model to obtain the normalized correction value of the structural response data of the distorted structural monitoring sensor; wherein the distorted data reconstruction model is obtained by training the BLSTM based on the historical normal structural response data of each structural monitoring sensor; The anti-normalization module is used to perform anti-normalization processing on the normalized correction value of the structural response data to obtain the structural response data correction value of the distorted structural monitoring sensor.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for repairing distorted data of a structural monitoring sensor network as described in any one of claims 1 to 6 is implemented.

Images