Method and device for coherent fading suppression of a phi-otdr system based on spatially resolved redundant data

Abstract

The application discloses a kind of Φ-OTDR system coherence fading inhibition method and device based on spatial resolution internal redundant data, specifically: first, simulate random one-dimensional continuous signal and noise waveform, and carry out channel replication expansion, add different random noise to the waveform of each channel, obtain multi-channel signal and noise data set;Then set the structure, parameter, activation function, loss function and optimizer of time sequence neural network model, data set is sent into model and is trained;Then judge whether the model error meets the requirements, if it does not meet the requirements, adjust the structure, parameter, activation function, loss function and optimizer of model, retrain the model;If it meets the requirements, the actual data detected by Φ-OTDR system is processed and packed, and is sent into the trained model for prediction, to obtain the signal characteristics distributed along the whole optical fiber.The application has low complexity, reduces the influence of coherence fading noise on Φ-OTDR system, and improves the transmission ability of the system.

Description

Technical Field

[0001] This invention relates to the fields of optical sensing technology and deep learning technology, and in particular to a method and apparatus for suppressing coherent fading in a Φ-OTDR system based on redundant data within spatial resolution. Background Technology

[0002] In recent years, the performance of distributed acoustic sensor systems (DAS) has been rapidly improving, enabling the reconstruction of complete vector fields of acoustic events at arbitrary locations in optical fibers, including information such as amplitude, frequency, and phase. Phase-sensitive optical time-domain reflectometry (Φ-OTDR) is an important area within DAS systems, achieving distributed sensing through the detection of backscattered Rayleigh light (RBS). Compared to traditional optical time-domain reflectometry systems, Φ-OTDR uses highly coherent lasers as the light source, resulting in higher sensitivity and responsivity. This technology has been widely applied in perimeter security inspection, structural health monitoring, and underwater sound source localization.

[0003] However, the sensing performance of Φ-OTDR is susceptible to coherent fading. Due to the spatially non-uniform distribution of the fiber's refractive index, the RBS generated by each scattering unit interferes with each other, resulting in random destructive interference points. At these points, the RBS signal strength may approach zero and be submerged in background noise, which is detrimental to the distributed sensing capabilities of Φ-OTDR. In recent years, researchers have proposed methods such as frequency division multiplexing, pulse modulation, and space division multiplexing to suppress coherent fading noise, effectively mitigating this issue. However, these methods are all implemented in hardware, increasing the system's cost and complexity.

[0004] With the improvement of computing power and the development of machine learning, deep learning technology based on temporal neural networks has been applied to various fields. The location of weak light intensity caused by coherent fading can be approximated as the location of weak signal-to-noise ratio (SNR), and temporal neural networks, trained on large datasets, have a significant ability to process low SNR signals. In Φ-OTDR systems, deep learning algorithms have been proven to effectively improve the SNR and unpack continuous phase information. Summary of the Invention

[0005] The purpose of this invention is to provide a method and apparatus for suppressing coherent fading in a Φ-OTDR system that has low system complexity, can reduce the adverse effects of coherent fading noise on the Φ-OTDR system, improve the system signal-to-noise ratio, and reduce the false alarm rate.

[0006] The technical solution to achieve the purpose of this invention is: a method for suppressing coherent fading in a Φ-OTDR system based on redundant data within spatial resolution, comprising the following steps:

[0007] Step 1: Based on the waveform characteristics of the demodulated signal of the Φ-OTDR system, simulate the random one-dimensional continuous signal and noise waveform;

[0008] Step 2: Expand the waveform generated in Step 1 by copying and adding channels, and add different random noise to the waveform of each channel to obtain a multi-channel signal and noise dataset;

[0009] Step 3: Set the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model, and input the multi-channel signal and noise dataset generated in Step 2 into the temporal neural network model for training;

[0010] Step 4: Determine whether the error meets the requirements through evaluation metrics. If it does, proceed to Step 5; otherwise, adjust the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model, and proceed to Step 3 to retrain the temporal neural network model.

[0011] Step 5: After processing and packaging the actual data detected by the Φ-OTDR system, the data is fed into the trained temporal neural network model for prediction to obtain optimized data. The optimized data is then subjected to spectrum analysis to obtain the signal characteristics distributed globally along the optical fiber.

[0012] A coherent fading suppression device for a Φ-OTDR system based on spatial resolution redundancy data is disclosed. This device is used to implement the aforementioned coherent fading suppression method for a Φ-OTDR system based on spatial resolution redundancy data. The device includes a simulation module, a channel replication and expansion module, a model training module, a judgment module, and a model prediction module, wherein:

[0013] The simulation module, combined with the waveform characteristics of the demodulated signal of the Φ-OTDR system, simulates random one-dimensional continuous signals and noise waveforms;

[0014] The channel copying and expansion module copies and expands the waveforms generated by the simulation module, and adds different random noise to the waveforms of each channel to obtain a multi-channel signal and noise dataset.

[0015] The model training module sets the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model, and feeds the multi-channel signal and noise dataset generated by the channel replication and expansion module into the temporal neural network model for training.

[0016] The judgment module uses evaluation metrics to determine whether the error meets the requirements. If it does, it proceeds to the model prediction module; otherwise, it adjusts the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model and proceeds to the model training module to retrain the temporal neural network model.

[0017] The model prediction module processes and packages the actual data detected by the Φ-OTDR system, then feeds it into the trained temporal neural network model for prediction to obtain optimized data. Finally, it performs spectral analysis on the optimized data to obtain the signal characteristics distributed globally along the optical fiber.

[0018] A mobile terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the coherent fading suppression method for the Φ-OTDR system based on redundant data within spatial resolution.

[0019] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the Φ-OTDR system coherent fading suppression method based on redundant data within spatial resolution.

[0020] Compared with the prior art, the present invention has the following significant advantages: (1) By comprehensively analyzing the redundant information data carried by the RBS of adjacent sampling points, the noise signal of individual channels is suppressed by using the multi-channel data synthesis method, and the temporal neural network is used as the core layer of the model, which improves the noise suppression capability and robustness and reduces the false alarm rate of the Φ-OTDR system; (2) The adverse effects of coherent fading noise of the Φ-OTDR system are reduced by the data end post-processing method, which reduces the hardware complexity and system cost of the system. Attached Figure Description

[0021] Figure 1 This is a schematic flowchart of the coherent fading suppression method for the Φ-OTDR system based on redundant data within spatial resolution, as proposed in this invention.

[0022] Figure 2 This is a schematic diagram of the temporal neural network model in this invention.

[0023] Figure 3 This is a comparison curve of system coherent fading noise before and after model processing of the Φ-OTDR system sensor data in this embodiment of the invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Typically, the spatial sampling rate of a Φ-OTDR system is much greater than its spatial resolution. This means that the RBSs of adjacent sampling points carry the same external vibration signals, thus providing redundant information. Therefore, these multi-channel data can be comprehensively analyzed to improve the ability to suppress coherent fading noise.

[0027] Combination Figure 1 The present invention discloses a method for suppressing coherent fading in a Φ-OTDR system based on redundant data within spatial resolution, comprising the following steps:

[0028] Step 1: Based on the waveform characteristics of the demodulated signal of the Φ-OTDR system, simulate the random one-dimensional continuous signal and noise waveform;

[0029] Step 2: Expand the waveform generated in Step 1 by copying and adding channels, and add different random noise to the waveform of each channel to obtain a multi-channel signal and noise dataset;

[0030] Step 3: Set the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model, and input the multi-channel signal and noise dataset generated in Step 2 into the temporal neural network model for training;

[0031] Step 4: Determine whether the error meets the requirements through evaluation metrics. If it does, proceed to Step 5; otherwise, adjust the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model, and proceed to Step 3 to retrain the temporal neural network model.

[0032] Step 5: After processing and packaging the actual data detected by the Φ-OTDR system, the data is fed into the trained temporal neural network model for prediction to obtain optimized data. The optimized data is then subjected to spectrum analysis to obtain the signal characteristics distributed globally along the optical fiber.

[0033] As a specific example, step 1, which combines the waveform characteristics of the demodulated signal from the Φ-OTDR system, simulates a random one-dimensional continuous signal and noise waveform, as follows:

[0034] For signal waveforms, several seed points are randomly generated, satisfying Gaussian or uniform distribution. The number of points is expanded by spline interpolation to form a continuous and smooth one-dimensional waveform curve. For noise waveforms, a random noise signal with Gaussian distribution is generated.

[0035] As a specific example, step 2 involves channel duplication and expansion of the waveform generated in step 1, and adding different random noise to the waveform of each channel to obtain a multi-channel signal and noise dataset, as detailed below:

[0036] Step 2.1: Use the signal and noise waveforms generated in Step 1 as the training dataset labels, and copy them into 5 copies to represent adjacent multi-channel waveforms;

[0037] Step 2.2: Add random Gaussian noise, salt-and-pepper noise, and drift noise to the waveform of each channel to obtain a multi-channel signal and noise dataset, so that the waveform of each channel exhibits different noise characteristics on the basis of the same signal characteristics.

[0038] As a specific example, the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model described in step 3 are as follows:

[0039] The temporal neural network LSTM is used as the core layer of the temporal neural network model. The LSTM has 5 input features, 256 hidden features, and 2 return layers. Convolutional layers (Conv) are used to optimize the model's prediction performance. The core size of the convolutional layer is 3×3, with a stride of 1 and padding of 1. After each convolutional layer, a normalized BN layer, a ReLU activation function layer, and a max pooling layer are added for processing. Finally, the data is expanded into a Flatten layer and a Linear layer to obtain the single-channel output result. The temporal neural network model is trained using Adam as the optimizer and MSE as the loss function.

[0040] As a specific example, step 5 involves processing and packaging the actual data detected by the Φ-OTDR system, then feeding it into a trained temporal neural network model for prediction to obtain optimized data. The optimized data is then subjected to spectral analysis to obtain the signal characteristics distributed globally along the optical fiber, as detailed below:

[0041] Step 5.1: Extract the actual phase signals of all n positions along the optical fiber detected by the Φ-OTDR system, and pack the data of every 5 adjacent positions into a group, for a total of n-4 groups of phase data;

[0042] Step 5.2: Feed the packaged data into the trained temporal neural network model for prediction to obtain optimized data;

[0043] Step 5.3: Perform spectral analysis on the optimized data to obtain the signal characteristics distributed globally along the optical fiber.

[0044] This invention also provides a coherent fading suppression device for a Φ-OTDR system based on spatial resolution redundancy data. This device is used to implement the aforementioned coherent fading suppression method for a Φ-OTDR system based on spatial resolution redundancy data. The device includes a simulation module, a channel replication and expansion module, a model training module, a judgment module, and a model prediction module, wherein:

[0045] The simulation module, combined with the waveform characteristics of the demodulated signal of the Φ-OTDR system, simulates random one-dimensional continuous signals and noise waveforms;

[0046] The channel copying and expansion module copies and expands the waveforms generated by the simulation module, and adds different random noise to the waveforms of each channel to obtain a multi-channel signal and noise dataset.

[0047] The model training module sets the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model, and feeds the multi-channel signal and noise dataset generated by the channel replication and expansion module into the temporal neural network model for training.

[0048] The judgment module uses evaluation metrics to determine whether the error meets the requirements. If it does, it proceeds to the model prediction module; otherwise, it adjusts the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model and proceeds to the model training module to retrain the temporal neural network model.

[0049] The model prediction module processes and packages the actual data detected by the Φ-OTDR system, then feeds it into the trained temporal neural network model for prediction to obtain optimized data. Finally, it performs spectral analysis on the optimized data to obtain the signal characteristics distributed globally along the optical fiber.

[0050] The present invention also provides a mobile terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the Φ-OTDR system coherent fading suppression method based on redundant data within spatial resolution.

[0051] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps in the Φ-OTDR system coherent fading suppression method based on redundant data within spatial resolution.

[0052] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0053] Example

[0054] Combination Figure 1 and Figure 2The coherent fading suppression method based on redundant data within spatial resolution provided in this embodiment has the following specific steps:

[0055] Step 1: Based on the waveform characteristics of the demodulated signal of the Φ-OTDR system, use simulation software to simulate the continuous signal and noise waveforms with a random one-dimensional data length of 256, as follows:

[0056] For the signal waveform, seed points are randomly generated within the range [1, 256]. The number and amplitude of the seed points are within the intervals (30, 40) and (0, 40), respectively, satisfying a Gaussian or uniform distribution. Spline interpolation is used to interpolate between the seed points, generating a continuous smooth waveform curve of length 256. For the noise waveform, a horizontal curve with an amplitude of Gaussian distribution between (0, 0.1) is generated. The signal and noise curves generated in the above steps are used as data labels for the training set.

[0057] Step 2: Expand the waveform generated in Step 1 by copying and adding channels, and add different random noise to the waveform of each channel to obtain a multi-channel signal and noise dataset with 5 channels, as follows:

[0058] Step 2.1: Use the signal and noise waveforms generated in Step 1 as the training dataset labels, and copy them into 5 copies to represent adjacent multi-channel waveforms;

[0059] Step 2.2: Add random Gaussian noise, salt-and-pepper noise, and drift noise to the waveform of each channel, so that the waveform of each channel exhibits different noise characteristics on the basis of the same signal characteristics.

[0060] Since the temporal neural network model uses ReLU as the activation function, the absolute value of the minimum value is added to the simulation data of each channel to make the overall data greater than 0.

[0061] Step 3: Set the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model. Input the multi-channel signal and noise dataset generated in Step 2 into the model for training. Figure 2 As shown, the details are as follows:

[0062] The Temporal Sequence Neural Network (LSTM) is used as the core layer of the model. The LSTM has 5 input channels, 256 hidden channels, and 2 return layers. The input data dimension is 5×256×1. After LSTM processing, the number of channels is expanded to 256, and the output data dimension is 256×256×1. Convolutional layers (Conv) are used to optimize the model's prediction performance. The core size of the convolutional layer is 3×3, with a stride of 1 and padding of 1. After each convolutional layer, a normalized Batch Normalization (BN) layer, a ReLU activation function layer, and a MaxPool layer are added for processing. The model undergoes Conv, BN, ReLU, and MaxPool processing three times, and the data dimension is halved after each processing, resulting in an output data dimension of 32×32×1. Finally, the model passes through a Flatten layer and a Linear layer to obtain the output result, with an output data dimension of 1×256×1. The model training uses Adam as the optimizer and MSE as the loss function. The training process consists of 300 stages, with a batch size of 256 and a learning rate of 0.0001.

[0063] Step 4: Determine whether the error meets the requirements through evaluation metrics. If it does, proceed to Step 5; otherwise, adjust the model's structure, parameters, activation function, loss function, and optimizer, and proceed to Step 3 to retrain the model.

[0064] Step 5: After processing and packaging the actual data detected by the Φ-OTDR system, it is fed into the trained model for prediction to obtain optimized data. The optimized data is then subjected to spectral analysis to obtain the signal characteristics distributed globally along the optical fiber, as follows:

[0065] Step 5.1: Extract the actual phase signals of all n positions along the optical fiber detected by the Φ-OTDR system, and pack the data of every 5 adjacent positions into a group, for a total of n-4 groups of phase data;

[0066] Step 5.2: Feed the packaged data into the trained network model for prediction to obtain optimized data;

[0067] Step 5.3: Perform spectral analysis on the optimized data to obtain the signal characteristics distributed globally along the optical fiber.

[0068] Figure 3 This is a comparison curve of the system coherent fading noise before and after model processing of the Φ-OTDR system sensor data.

[0069] Depend on Figure 3It can be seen that the coherent fading suppression method based on redundant data within spatial resolution provided by this invention reduces the adverse effects of coherent fading noise on the Φ-OTDR system by using redundant information of adjacent channels within the integrated spatial resolution and temporal neural network technology without increasing the hardware complexity and cost of the system. This improves the noise suppression capability and robustness, and reduces the false alarm rate of the Φ-OTDR system.

[0070] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for suppressing coherent fading in a Φ-OTDR system based on redundant data within spatial resolution, characterized in that, Includes the following steps: Step 1: Based on the waveform characteristics of the demodulated signal of the Φ-OTDR system, simulate the random one-dimensional continuous signal and noise waveform; Step 2: Expand the waveform generated in Step 1 by copying and adding channels, and add different random noise to the waveform of each channel to obtain a multi-channel signal and noise dataset; Step 3: Set the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model, and input the multi-channel signal and noise dataset generated in Step 2 into the temporal neural network model for training; The structure, parameters, activation function, loss function, and optimizer of the specified temporal neural network model are as follows: The temporal neural network LSTM is used as the core layer of the model. The LSTM has 5 input features, 256 hidden features, and 2 return layers. Convolutional layers (Conv) are used to optimize the model's prediction performance. The core size of the convolutional layer is 3×3, with a stride of 1 and padding of 1. After each convolutional layer, a normalized BN layer, a ReLU activation function layer, and a max pooling layer are added for processing. Finally, a flatten layer and a linear layer are used to obtain the single-channel output. The temporal neural network model is trained using Adam as the optimizer and MSE as the loss function. Step 4: Determine whether the error meets the requirements through evaluation metrics. If it does, proceed to Step 5; otherwise, adjust the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model, and proceed to Step 3 to retrain the temporal neural network model. Step 5: After processing and packaging the actual data detected by the Φ-OTDR system, it is fed into the trained temporal neural network model for prediction to obtain optimized data. The optimized data is then subjected to spectral analysis to obtain the signal characteristics distributed globally along the optical fiber, as detailed below: Step 5.1: Extract the actual phase signals at all n locations along the optical fiber detected by the Φ-OTDR system, and pack the data of every 5 adjacent locations into a group, for a total of n-4 groups of phase data; Step 5.2: Feed the packaged data into the trained temporal neural network model for prediction to obtain optimized data; Step 5.3: Perform spectral analysis on the optimized data to obtain the signal characteristics distributed globally along the optical fiber.

2. The method for suppressing coherent fading in a Φ-OTDR system based on redundant data within spatial resolution according to claim 1, characterized in that, Step 1 involves simulating random one-dimensional continuous signals and noise waveforms by combining the waveform characteristics of the demodulated signal from the Φ-OTDR system, as detailed below: For signal waveforms, several seed points are randomly generated, satisfying Gaussian or uniform distribution. The number of points is expanded by spline interpolation to form a continuous and smooth one-dimensional waveform curve. For noise waveforms, a random noise signal with Gaussian distribution is generated.

3. The method for suppressing coherent fading in a Φ-OTDR system based on redundant data within spatial resolution according to claim 1, characterized in that, Step 2 involves channel replication and expansion of the waveform generated in Step 1, and adding different random noise to the waveform of each channel to obtain a multi-channel signal and noise dataset, as detailed below: Step 2.1: Use the signal and noise waveforms generated in Step 1 as the training dataset labels, and copy them into 5 copies to represent adjacent multi-channel waveforms; Step 2.2: Add random Gaussian noise, salt-and-pepper noise, and drift noise to the waveform of each channel to obtain a multi-channel signal and noise dataset, so that the waveform of each channel exhibits different noise characteristics on the basis of the same signal characteristics.

4. A coherent fading suppression device for a Φ-OTDR system based on redundant data within spatial resolution, characterized in that, This device is used to implement the coherent fading suppression method for Φ-OTDR systems based on redundant data within spatial resolution as described in any one of claims 1 to 3. The device includes a simulation module, a channel replication and expansion module, a model training module, a judgment module, and a model prediction module, wherein: The simulation module, combined with the waveform characteristics of the demodulated signal of the Φ-OTDR system, simulates random one-dimensional continuous signals and noise waveforms; The channel copying and expansion module copies and expands the waveforms generated by the simulation module, and adds different random noise to the waveforms of each channel to obtain a multi-channel signal and noise dataset. The model training module sets the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model, and feeds the multi-channel signal and noise dataset generated by the channel replication and expansion module into the temporal neural network model for training. The judgment module uses evaluation metrics to determine whether the error meets the requirements. If it does, it proceeds to the model prediction module; otherwise, it adjusts the structure, parameters, activation function, loss function, and optimizer of the temporal neural network model and proceeds to the model training module to retrain the temporal neural network model. The model prediction module processes and packages the actual data detected by the Φ-OTDR system, then feeds it into the trained temporal neural network model for prediction to obtain optimized data. Finally, it performs spectral analysis on the optimized data to obtain the signal characteristics distributed globally along the optical fiber.

5. A mobile terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the coherent fading suppression method for the Φ-OTDR system based on redundant data within spatial resolution as described in any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the Φ-OTDR system coherent fading suppression method based on redundant data within spatial resolution as described in any one of claims 1 to 3.

Images