阅读说明：本技术 Sagnac分布光纤传感系统的离散化定位方法 (Discretization positioning method of Sagnac distributed optical fiber sensing system ) 是由 方捻 吕继东 王陆唐 于 2019-10-21 设计创作，主要内容包括：本发明公开了一种Sagnac分布光纤传感系统的离散化定位方法。以所需定位分辨率为间隔,将传感光纤等分为若干位置区间,并对每个位置区间进行编号；分别在各位置区间多次模拟传感光纤受到宽带信号的扰动,并采集相应的干涉信号；对干涉信号进行变换或分析,选取合适的特征变量及其数据长度；以干涉信号的特征数据为训练输入、位置区间编号为目标输出构建多分类模型；对新采集的待定位的某一位置区间受到扰动引起的干涉信号提取特征；把该干涉信号的特征数据输入已训练好的多分类模型,得到对应的位置区间编号,即实现了扰动信号的离散化定位。本发明简单有效,对噪声不敏感,定位分辨率灵活可调,可用于环形或直线式Sagnac分布光纤传感系统的扰动定位。(The invention discloses a discretization positioning method of a Sagnac distribution optical fiber sensing system. Equally dividing the sensing optical fiber into a plurality of position intervals by taking the required positioning resolution as an interval, and numbering each position interval; simulating the disturbance of the sensing optical fiber on the broadband signal for multiple times in each position interval respectively, and collecting corresponding interference signals; transforming or analyzing the interference signal, and selecting proper characteristic variables and data lengths thereof; constructing a multi-classification model by taking the characteristic data of the interference signal as training input and the position interval number as target output; extracting characteristics of interference signals caused by disturbance of a newly acquired certain position interval to be positioned; and inputting the characteristic data of the interference signal into the trained multi-classification model to obtain the corresponding position interval number, namely realizing the discretization positioning of the disturbance signal. The method is simple and effective, insensitive to noise, flexible and adjustable in positioning resolution, and can be used for disturbance positioning of an annular or linear Sagnac distribution optical fiber sensing system.)

1. A discretization positioning method of a Sagnac distributed optical fiber sensing system is characterized by comprising the following steps:

1) equally dividing the sensing optical fiber into a plurality of position intervals by taking the required positioning resolution as an interval, and numbering each position interval;

2) simulating the disturbance of the sensing optical fiber on the broadband signal for multiple times in each position interval respectively, and collecting corresponding interference signals;

3) transforming or analyzing the interference signal, and selecting proper characteristic variables and data length;

4) constructing a multi-classification model by taking the characteristic data of the interference signal as training input and the position interval number as target output;

5) extracting characteristics of interference signals caused by disturbance of a newly acquired certain position interval to be positioned;

6) and inputting the characteristic data of the interference signal into the trained multi-classification model to obtain the corresponding position interval number, namely realizing the discretization positioning of the disturbance signal.

2. The discretization positioning method of Sagnac distributed optical fiber sensing system according to claim 1, wherein the characteristic variable in step 3) is a frequency spectrum, or an intercepted interference signal, or a characteristic quantity after two frequency spectrum transformations, or a characteristic value obtained by principal component analysis, or a characteristic extracted by a convolutional neural network.

3. The discretization positioning method of Sagnac distributed fiber sensing system of claim 1, wherein the multi-classification model in step 4) is a support vector machine, a decision tree, an artificial neural network, a k-nearest neighbor, a Bayesian classifier, ensemble learning, or a clustering machine learning model.

Technical Field

The invention relates to a positioning method, in particular to a discretization positioning method of a Sagnac distributed optical fiber sensing system.

Background

The distributed optical fiber sensing system can be used for monitoring leakage of oil and gas pipelines, wherein the system based on the Sagnac interferometer has the advantages of strong anti-interference performance, low requirement on a light source and the like, and is a hotspot of current research. For the Sagnac distributed optical fiber sensing system, when the external disturbance is a broadband signal, the frequency spectrum of the phase change signal caused by the external disturbance has a periodic zero point, namely a zero frequency point. The position of the disturbance point can be calculated through the relation between the zero frequency point and the disturbance position. Due to the influence of factors such as random phase drift, thermal noise and the like in a sensing system, zero frequency and a large amount of noise coexist, and the reading of the zero frequency is influenced. The system structure or the noise reduction compensation algorithm can be improved by improving, for example, a Faraday rotator mirror is added in a linear Sagnac distributed optical fiber sensing system to compensate signal polarization fading, and random drift of a working point is eliminated and low-frequency component interference is reduced by phase generation carrier modulation; the method has the advantages that a more accurate phase change signal is obtained by improving a phase generation carrier demodulation algorithm, signal noise is further reduced through discrete wavelet transform, the descending trend of a frequency spectrum curve is eliminated through least square fitting, and positioning accuracy and the like are improved through secondary Fourier transform. However, these methods can only reduce the influence of system noise on the zero frequency point reading to a certain extent, and increase the complexity of the system and the algorithm.

In recent years, machine learning methods have unique advantages in signal processing, and a great deal of research has been carried out in intrusion signal identification based on machine learning methods such as a support vector machine and an artificial neural network. In the distributed optical fiber sensing positioning aspect, a learner approaches a disturbance position and an actual position determined by a wavelet de-noised frequency spectrum through a support vector machine algorithm to reduce a positioning error, but the method still needs to perform initial positioning by using a zero frequency method.

Disclosure of Invention

The invention aims to provide a discretization positioning method of a Sagnac distributed optical fiber sensing system aiming at the defects in the prior art and aiming at simplifying the system structure and the signal processing process. The positioning problem of the disturbance is converted into the classification problem of the interference signal, the processes of demodulation and denoising are omitted, the characteristics of the interference signal are extracted, a classification model is trained, and the position interval of the disturbance is directly obtained.

In order to achieve the purpose, the invention adopts the following technical scheme:

a discretization positioning method of a Sagnac distributed optical fiber sensing system comprises the following steps:

1) equally dividing the sensing optical fiber into a plurality of position intervals by taking the required positioning resolution as an interval, and numbering each position interval;

2) simulating the disturbance of the sensing optical fiber on the broadband signal for multiple times in each position interval respectively, and collecting corresponding interference signals;

3) and transforming or analyzing the interference signal, and selecting proper characteristic variables and data length.

4) Constructing a multi-classification model by taking the characteristic data of the interference signal as training input and the position interval number as target output;

5) extracting characteristics of interference signals caused by disturbance of a newly acquired certain position interval to be positioned;

6) and inputting the characteristic data of the interference signal into the trained multi-classification model to obtain the corresponding position interval number, namely realizing the discretization positioning of the disturbance signal.

The characteristic variables in the step 3) are frequency spectrums or intercepted interference signals, characteristic quantities after two times of frequency spectrum transformation, characteristic values obtained by principal component analysis, characteristics extracted by a convolutional neural network and the like.

The multi-classification model in the step 4) is a machine learning model such as a support vector machine, a decision tree, an artificial neural network, k-nearest neighbor, a Bayesian classifier, ensemble learning, clustering and the like.

The working principle and the characteristics of the invention are as follows:

regardless of the positioning method of the distributed optical fiber sensing system, the positioning result is generally limited in resolution due to various factors, and complete continuous positioning cannot be achieved. Therefore, the sensing optical fiber can be discretized according to the requirement of actual positioning resolution, namely, the sensing optical fiber is equally divided into a plurality of position intervals by taking the resolution as an interval, interference signals caused by disturbance in different position intervals are different, and the positioning problem of the different position intervals can be converted into the multi-classification problem of the interference signals caused by external disturbance.

The position sections of the sensing optical fibers are respectively numbered asP _i ，i=1,2…NThe label is output or classified as a target of the multi-classification model. And respectively carrying out broadband disturbance for multiple times in each position interval. And extracting the characteristics of the interference signals generated by the disturbance signals in each position interval. Inputting the characteristics into a multi-classification model, and training and testing the multi-classification model to obtain a disturbance position interval corresponding to the interference signalP _i Thereby achieving the purpose of positioning.

Compared with the prior art, the invention has the following obvious and prominent substantive characteristics and remarkable advantages:

the method is simple and effective, has good positioning real-time performance, is insensitive to noise, has flexible and adjustable positioning resolution, and can be used for disturbance positioning of an annular or linear Sagnac distribution optical fiber sensing system.

Drawings

Fig. 1 is a schematic structural diagram of a system according to embodiment 1.

FIG. 2 is a flow chart of the method of the present invention.

Fig. 3 shows spectral features of interference signals extracted in the absence of noise.

Fig. 4 shows spectral characteristics of interference signals extracted in the case of noise.

Fig. 5 shows spectral features of a portion of training data and their corresponding class labels.

FIG. 6 shows parametersC、gAnd (4) the classification accuracy of the model verification set.

Detailed Description

The preferred embodiments of the present invention are described below with reference to the accompanying drawings:

monitoring of the pipeline leakage by utilizing an OptiSystemsoftware simulation annular Sagnac distribution optical fiber sensing system verifies the feasibility of the discretization positioning method.

As shown in fig. 1, the simulated ring-shaped Sagnac distributed optical fiber sensing system includes a continuous laser 1, a 2 × 2 bidirectional 3dB optical coupler 2, a sensing optical fiber 3, a sensing optical fiber 4, a delay optical fiber 5, a phase modulator 6, a photodetector 7, and a data acquisition and processing unit 8. Length of sensing optical fiber 3 and sensing optical fiber 4R ₁、R ₂The sum being equal to the length of the delay fibreR ₃. The effect of the pipe leakage perturbation signal on the light in the sensing fiber is simulated by the phase modulator 6. Since the bandwidth of the impact signal generated by the pipeline leakage is as high as 60kHz, the Sinc function with the bandwidth of 60kHz is used for simulating the pipeline leakage disturbance signal. By changingR ₁、R ₂Simulating different disturbance positions.

In this embodiment, the length of the delay fiberR ₃10km, the total length of the sensing fiber is also 10 km. And taking a sensing optical fiber within 8.5km from the coupler as an effective sensing area. The required positioning resolution is 100m, and the system sampling frequency is set to be 2 MHz.

The pipeline leak is located according to the flow shown in fig. 2. Firstly, discretizing the sensing optical fiber, dividing an effective sensing optical fiber area of 0-8.5 km into 86 position sections with the interval of 100m, wherein the number of the effective sensing optical fiber area is respectively numberedP ₁，P ₂，…，P ₈₆(ii) a And simulating pipeline leakage disturbance in each position interval respectively, and acquiring interference signals generated by the system to obtain interference signals corresponding to 86 position intervals. Adding different white Gaussian noises to the 86 noise-free interference signals through a data acquisition and processing unit to obtain a signal-to-noise ratio (SNR) 2580 noisy interference signals, 1 to 30, respectively, spaced 1 apart. Each position interval corresponds to 30 interference signals containing noise, and the position interval number is the category label of the interference signals. The sequence of 2580 noisy interference signals is scrambled, and 90% of the data is randomly selected as a training data set, and the remaining 10% of the data is selected as a test data set.

And performing fast Fourier transform on the interference signals of all classes to obtain frequency spectrums of the interference signals. And extracting the first 2048 points of the interference signal spectrum as characteristic data, and performing normalization processing. FIG. 3 shows the spectral characteristics of the interference signal in the noise-free case obtained at a 4km location under the effect of perturbation; figure 4 showsSNRThe spectral signature of the interference signal of = 10.

The frequency spectrum characteristic data is used as the training input of a classification model, and the position interval is numbered as the target output to constructA multi-classification model. Fig. 5 shows normalized spectral characteristics of the resulting interference signals at 1km, 2km, 3km, 4km, 5km and 6km from the optical coupler at the location of the leak, with corresponding class labels being 11, 21, 31, 41, 51 and 61, respectively. The method comprises the steps of using a support vector machine classification algorithm based on a radial basis kernel function, adopting a one-to-one multi-classification mode, using a three-fold cross validation method to evaluate the performance of a model, and obtaining an optimal parameter model through a simple grid parameter optimization method. FIG. 6 shows penalty parametersCNuclear parametersgAnd (4) the classification accuracy of the model verification set. Taking penalty parametersC=50, nuclear parametersg=0.001, at which time a verification set classification accuracy of up to 100% can be achieved.

In the testing process, the spectrum characteristics of the interference signal to be positioned are input into the trained multi-classification model of the support vector machine, and the number of the disturbance position interval corresponding to the interference signal is obtained, namely, the discretization positioning of the pipeline leakage disturbance is realized. The classification accuracy of the test set is up to 100%, which indicates that the interference signals corresponding to each position interval can be accurately classified, i.e. accurately positioned.