A low-resolution Brillouin frequency spectrum demodulation method based on artificial neural network

Abstract

The application relates to a low-resolution Brillouin frequency spectrum demodulation method based on an artificial neural network, which comprises the following steps: S1, in a model training stage, adjusting a time domain and a frequency domain sampling rate of a Brillouin optical time domain sensor to obtain a low-resolution Brillouin frequency spectrum and taking the Brillouin frequency spectrum as a sample of a data training set; and simultaneously changing an external physical quantity exertion mode of a to-be-measured optical fiber, such as a heating or strain position and a temperature or strain amplitude, and establishing quasi-continuous temperature or strain field distribution information under different parameter conditions and taking the temperature or strain field distribution information as a label of the data training set. The application breaks through the limitation of the time domain sampling rate on the spatial resolution, saves data storage space, reduces the time consumption of temperature or strain field calculation, and improves the demodulation precision of the physical quantity distribution information.

Description

Technical fields:

[0001] This invention belongs to the field of fiber optic sensing technology, specifically relating to a low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks. Background technology:

[0002] Optical fibers, with their low loss, corrosion resistance, and electromagnetic interference resistance, have immense practical value in fields such as communication and sensing. Distributed optical fiber sensing technology, in particular, utilizes optical effects to measure the distribution of various physical quantities along the fiber, such as temperature, strain, and vibration. Currently, distributed optical fiber sensing systems primarily utilize three scattering effects in the fiber to achieve distributed physical quantity measurement. Rayleigh scattering-based distributed optical fiber sensing systems are used for short-distance temperature measurement and fiber attenuation monitoring, while Raman scattering-based systems are widely used due to their high sensitivity to temperature measurement. Finally, Brillouin scattering-based distributed optical fiber sensing systems have attracted widespread attention from researchers due to their high sensitivity to a variety of physical quantities.

[0003] When incident light propagates in an optical fiber, the excited phonons and photons interact to produce scattered light with a frequency difference from the incident light. This scattered light includes Stokes light with a lower frequency shift and anti-Stokes light with a higher frequency shift. Its power spectrum is approximately Lorentz-shaped, and the offset of its center frequency from the incident light's center frequency is called the Brillouin frequency shift (BFS). When physical quantities outside the fiber change, the Brillouin frequency shift at each location also changes. By establishing a mapping relationship between the two, distributed measurement of physical quantities can be achieved. The mainstream methods are based on two time-domain signals: Brillouin Optical Time-Domain Analysis (BOTDA) and Brillouin Optical Time-Domain Reflectometry (BOTDR). Taking Brillouin optical time-domain reflectometry as an example, a set of Brillouin scattering spectra can be obtained by scanning the power of scattered light at different frequency differences. After extracting the Brillouin frequency shift of the Brillouin scattering spectrum at each fiber location and corresponding it with the physical quantity, a single physical quantity measurement is completed.

[0004] Brillouin Optical Time-Domain Sensors (BOTDS) possess exceptional sensing accuracy, enabling them to precisely locate temperature / strain events along optical fibers. This capability is crucial in numerous applications, such as manufacturing process monitoring and aircraft structural monitoring, where damaged modules must be accurately located and repaired to prevent safety incidents. For these systems, BOTDS requires extremely high precision and speed in measuring the edges, length, and amplitude of events.

[0005] Traditional Brillouin optical time-domain sensing systems require frequency scanning to construct a two-dimensional Brillouin frequency spectrum with high frequency resolution. This necessitates repeated measurements at different scattering frequencies, significantly increasing measurement time. Simultaneously, the high-precision demodulation requirements of physical quantities necessitate extremely high temporal sampling rates, which slows down data processing, compromises real-time performance, and increases hardware costs. Existing post-processing methods, such as deconvolution-based recovery techniques, aim to improve spatial resolution and avoid performance degradation caused by pulse width. However, these traditional methods, including those based on convolutional neural networks, cannot overcome the accuracy limitations imposed by the system's sampling rate. For example, when the transition region of a monitored event is 1 cm long and the distance between adjacent points in the sampled data is 50 cm, the highest spatial resolution achievable by traditional methods is limited to 50 cm. This limitation leads to a significant decrease in the spatial accuracy of physical quantity demodulation, caused by the inherent discreteness of the sampled data, and is difficult to overcome using existing traditional methods. Summary of the Invention:

[0006] The technical problem this invention aims to solve is to provide a low-resolution Brillouin frequency spectrum demodulation method based on an artificial neural network. This method obtains the Brillouin frequency spectrum through a lower frequency domain sampling rate, thereby avoiding the increased measurement time caused by extensive frequency sweeping. Simultaneously, it employs a lower time domain sampling rate to reduce the amount of data to be processed and alleviate storage pressure. Subsequently, an artificial neural network (ANN) is used to establish a mapping relationship between the discrete low-resolution Brillouin frequency spectrum and the quasi-continuous distribution field of temperature or strain. Finally, in the measurement stage, the downsampled sensor data is input, and the quasi-continuous distribution information of the measured physical quantity field is output.

[0007] The technical solution of this invention is to provide a low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks, comprising the following steps:

[0008] S1. During the model training phase, the time and frequency domain sampling rates of the Brillouin optical time-domain sensor are adjusted to obtain a low-resolution Brillouin frequency spectrum, which is then used as a sample in the data training set. Simultaneously, the application of external physical quantities to the optical fiber under test is changed, such as the heating or strain position and temperature or strain amplitude, to establish quasi-continuous temperature or strain field distribution information under different parameter conditions. This information is then used as a label for the data training set. These distribution information are associated with the corresponding low-resolution Brillouin frequency spectrum to construct a low-resolution Brillouin frequency spectrum-temperature or strain quasi-continuous distribution field sample dataset.

[0009] S2. Build an artificial neural network model and train the parameters of the artificial neural network model using a low-resolution Brillouin frequency spectrum-temperature or strain quasi-continuous distribution field sample dataset; by continuously adjusting the model parameters, it can accurately predict the corresponding temperature or strain distribution from the low-resolution Brillouin frequency spectrum.

[0010] S3. During the measurement phase, reduce the time and frequency domain sampling rates of the Brillouin optical time-domain sensor to obtain the Brillouin time-domain traces at each sensing position at different scattering frequencies of the fiber under test, and construct a low-resolution Brillouin frequency spectrum. The frequency domain sampling rate can be reduced by sampling at lower equal frequency intervals or unequal frequency intervals. The time domain sampling rate is achieved by sampling at lower equal time intervals.

[0011] S4. Input the low-resolution sensor data into the trained artificial neural network model to obtain a high-precision quasi-continuous temperature or strain field.

[0012] This invention employs only a low-sampling-rate Brillouin frequency spectrum in a Brillouin optical time-domain sensor. By utilizing a low-resolution Brillouin frequency spectrum demodulation method based on an artificial neural network, an artificial neural network model is built without introducing additional hardware or increasing the system sampling rate. Furthermore, by utilizing a sample dataset of "low-resolution Brillouin frequency spectrum - quasi-continuous temperature or strain distribution field," the limitation of time-domain sampling rate on spatial resolution is overcome, data storage space is saved, the computation time of temperature or strain fields is reduced, and the demodulation accuracy of physical quantity distribution information is improved.

[0013] Preferably, in step S1, the Brillouin optical time-domain sensor can be either a Brillouin optical time-domain analyzer or a Brillouin optical time-domain reflectometer.

[0014] Preferably, in step S1, the parameter conditions include the heating or strain position of the optical fiber under test, the temperature or strain amplitude, the linewidth of the Brillouin spectrum, the probe pulse width, the detector bandwidth, and the amplitude signal-to-noise ratio of the data.

[0015] Preferably, in step S1, the sample is low-resolution data corresponding to the Brillouin spectrum.

[0016] Preferably, in step S1, the label is the data corresponding to the quasi-continuous distribution field of temperature or strain.

[0017] Preferably, in step S2, the artificial neural network model includes an input layer, a hidden layer, and an output layer. The input layer has nodes corresponding to low-resolution Brillouin frequency spectrum data, and the output layer has nodes corresponding to high-precision temperature or strain field distributions. The hidden layer can be any type of network layer, including but not limited to Long and Short-Term Memory (LSTM) layers, fully connected layers, and convolutional layers. During training, the sample dataset is divided into a training set and a test set. The model is trained using the training set and its performance is tested using the test set.

[0018] Compared with the prior art, the present invention has the following advantages:

[0019] This invention uses only a low-sampling-rate Brillouin frequency spectrum in the Brillouin optical time-domain sensor. By utilizing a low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks, it overcomes the limitation of time-domain sampling rate on spatial resolution without introducing additional hardware or increasing the system sampling rate. This saves data storage space, reduces the calculation time for temperature or strain fields, and improves the demodulation accuracy of physical quantity distribution information. Attached image description:

[0020] Figure 1 This is a schematic flowchart of the method of the present invention;

[0021] Figure 2 The Brillouin frequency spectrum obtained using a lower sampling rate;

[0022] Figure 3 This refers to the start position, length, and amplitude information of the event segment extracted in this embodiment of the invention. Detailed implementation method:

[0023] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0024] This invention provides a low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks, such as... Figure 1 As shown, the present invention performs low-resolution Brillouin frequency spectrum demodulation through the following steps:

[0025] 1) The sampling range of the Brillouin optical time-domain sensor was set to 0–200 MHz, with a large 20 MHz interval in the scanning frequency dimension, containing a total of 11 frequency scanning points; the time-domain sampling rate was set to 200 MHz, the local fiber length under test was set to 25 meters, containing a total of 51 sampling points, with a point-to-point interval of 0.5 meters; the temperature event start point positions were set at 0.01-meter intervals, ranging from 7.5 meters to 12.5 meters, for a total of 501 groups; the event lengths were set at 0.01-meter intervals, ranging from 0.5 meters to the maximum value that could be set for each data group. The obtained discrete low-resolution Brillouin frequency spectrum data was used as samples for the data training set. Temperature values ​​were set in 10℃ intervals, ranging from 20℃ to 80℃, for a total of 7 groups; detector pulse widths were set in 10ns intervals, ranging from 20ns to 50ns, for a total of 4 groups; amplitude signal-to-noise ratios were set in 5dBm intervals, ranging from 10dBm to 30dBm, for a total of 5 groups; and detector bandwidth values ​​were set in 10MHz intervals, ranging from 30MHz to 200MHz, for a total of 18 groups. The low-resolution Brillouin frequency spectrum under different parameter settings was correlated with the position and amplitude distribution of the temperature field to construct a training dataset.

[0026] 2) Divide the training dataset into training, testing, and validation sets in an 8:1:1 ratio, and build an artificial neural network model, including one input layer, one LSTM layer, two fully connected layers, and one output layer, with 50, 128, 256, 512, and 501 nodes in each layer, respectively. Train the neural network model using the backpropagation algorithm.

[0027] 3) Using a conventional Brillouin optical time-domain reflectometer, a Brillouin frequency spectrum with an average order of 4096 was obtained on a 1000-meter-long bend-insensitive (G.657) optical fiber. The incident pulse width was set to 20 ns, corresponding to a theoretical spatial resolution of 2 m. A photodetector was used to measure 11 time-domain traces at 20 MHz intervals within a 200 MHz sweep range at a time-domain sampling rate of 200 MSample / s, collecting 2001 sampling points at each measured frequency. A 1.2-meter-long section of the optical fiber was heated to 40°C. The Brillouin frequency spectrum obtained with low-resolution sampling is shown below. Figure 2 As shown.

[0028] 4) Using the artificial neural network trained in step 2), the quasi-continuous distribution of the temperature field can be obtained from the Brillouin frequency spectrum data obtained in step 3), and the demodulation result is as follows: Figure 3As shown, the low-resolution Brillouin frequency spectrum demodulation method of this invention achieves extremely high accuracy in identifying temperature event segments. With a pulse width of 20 ns, the measurement uncertainty for the starting point of the temperature event segment is 5.6 cm, the measurement uncertainty for the length of the temperature event segment is 4.1 cm, the measured length of the temperature event segment is 115.7 cm, and the temperature measurement error is 0.3 °C. This demonstrates that this invention can demodulate the quasi-continuous distribution information of temperature event segments with extremely high accuracy. Compared to the high-resolution Brillouin frequency spectrum obtained with a frequency sampling interval of 1 MHz and a time-domain sampling rate of 2 GHz, this invention reduces measurement time by 20 times, reduces storage space by 10%, and significantly reduces the hardware cost of the acquisition equipment.

[0029] This invention can obtain the Brillouin frequency spectrum at a lower sampling rate, thereby avoiding the increase in measurement time caused by a large number of frequency sweeps. It also obtains a high-precision quasi-continuous physical quantity field distribution through a low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks. Without introducing additional hardware equipment or increasing the system sampling rate, it breaks through the limitation of time domain sampling rate on spatial resolution, saves data storage space, reduces the calculation time of temperature or strain fields, and improves the demodulation accuracy of temperature or strain field distribution information.

[0030] The above description only illustrates preferred embodiments of the present invention and should not be construed as limiting the scope of the claims. Any equivalent procedural modifications made using this specification are included within the patent protection scope of this invention.

Claims

1. A low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks, characterized in that: Includes the following steps, S1. During the model training phase, the time and frequency domain sampling rates of the Brillouin optical time-domain sensor are adjusted to obtain a low-resolution two-dimensional Brillouin frequency spectrum, which is then used as a sample in the data training set. Simultaneously, the heating or strain position and temperature or strain amplitude of the fiber under test are changed to establish quasi-continuous temperature or strain field distribution information under different parameter conditions, which is then used as a label for the data training set. The distribution information of these measured physical quantities is then associated with the corresponding low-resolution Brillouin frequency spectrum to construct a low-resolution two-dimensional Brillouin frequency spectrum-temperature or strain quasi-continuous distribution field sample dataset. S2. Build an artificial neural network model and train the parameters of the artificial neural network model using a low-resolution Brillouin frequency spectrum-temperature or strain quasi-continuous distribution field sample dataset. S3. During the measurement phase, reduce the time and frequency sampling rates of the Brillouin optical time-domain sensor to obtain the Brillouin time-domain traces at each sensing position at different scattering frequencies of the fiber under test, and construct a low-resolution Brillouin frequency spectrum. S4. Input the low-resolution sensor data into the trained artificial neural network model to obtain a high-precision quasi-continuous temperature or strain field.

2. The low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks according to claim 1, characterized in that: In step S1, the Brillouin optical time-domain sensor is a Brillouin optical time-domain analyzer or a Brillouin optical time-domain reflectometer.

3. The low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks according to claim 1, characterized in that: In step S1, the parameter conditions include the heating or strain position of the optical fiber under test, the temperature or strain amplitude, the linewidth of the Brillouin spectrum, the probe pulse width, the detector bandwidth, and the amplitude signal-to-noise ratio of the data.

4. The low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks according to claim 1, characterized in that: In step S1, the sample is low-resolution data corresponding to the Brillouin spectrum.

5. The low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks according to claim 1, characterized in that: In step S1, the labels are data corresponding to the temperature or strain quasi-continuous distribution field.

6. The low-resolution Brillouin frequency spectrum demodulation method based on artificial neural networks according to claim 1, characterized in that: In step S2, the artificial neural network model includes an input layer, a hidden layer, and an output layer. The input layer has nodes corresponding to low-resolution Brillouin frequency spectrum data, and the output layer has nodes corresponding to high-precision temperature or strain field distributions. The hidden layer can be any type of network layer. During training, the sample dataset is divided into a training set and a test set. The model is trained using the training set and its performance is tested using the test set.

Images