Machine learning-based online noise suppression method for ODVS engineering application system

Abstract

The application discloses an online noise suppression method of an ODVS engineering application system based on machine learning, and belongs to the technical field of calculation, estimation or counting. The method comprises the following steps: updating the closed-loop feedback of a noise pool through machine learning, updating the closed-loop feedback of online suppression of ODVS noise through machine learning, putting the noise collected by an ODVS sensing optical cable into the ODVS noise pool, extracting noise characteristics and adjusting noise parameters in the noise pool, determining an ODVS system noise model according to the adjusted noise parameters, feeding back the determined ODVS system noise model to the noise pool, adjusting ODVS noise reduction and de-noising algorithm parameters online according to the determined ODVS system noise model, enriching and perfecting the scene noise of the ODVS system through machine learning of the noise pool, improving the signal-to-noise ratio of ODVS system signal processing through online adjustment of the ODVS noise suppression algorithm, and improving the accuracy of the system.

Description

TECHNICAL FIELD

[0001] The present application relates to the noise pool closed loop improvement of the optical fiber distributed vibration sensing system (ODVS) and the method for adjusting the online noise suppression algorithm of the noise pool terminal machine, and specifically discloses an online noise suppression method of an ODVS engineering application system based on machine learning, belonging to the technical field of calculation, estimation or counting. BACKGROUND

[0002] In the engineering application of the optical fiber distributed vibration sensing system, the sensing optical cable of tens of kilometers is laid along the line, and various environmental and background noises are directly mixed in the effective disturbance signals sensed by the sensing optical cable along the line, resulting in a great decrease in the signal-to-noise ratio of the engineering application system and a decrease in the accuracy of the system in judging external interference. However, the research and manufacturing of the ODVS terminal machine are both carried out in a controllable environment. If the development of the terminal machine needs to consider various noises in the application scene, it is necessary to test the influence of the environmental and background noises in the actual conditions, and then find a method and way to suppress the system noise, which is time-consuming and laborious and has limited effect. Moreover, the on-site environment of the ODVS engineering application system is different, and the prior algorithm is difficult to replicate. Therefore, a method is needed to timely, preferably in real time, adjust the noise suppression measures of the terminal machine to achieve better engineering application effect. At present, there is no such method in the engineering application.

[0003] In order to design the optimal filter for different modes of environmental noise, the prior art proposes a noise scene recognition system and method, which determines the noise scene mode through offline and online comparison calculation. The noise scene recognition scheme is as follows: first, various typical environmental noises are collected offline, and the frequency spectrum of each environmental noise is obtained by using fast Fourier transform to form a typical environmental noise library; second, the current environmental noise signal is collected, and the frequency spectrum of the noise signal is obtained by using fast Fourier transform; and finally, the frequency spectrum of the current noise signal is compared and calculated with the frequency spectrum of the typical environmental noise in the noise library to determine the scene mode of the current noise. The noise scene recognition scheme mainly identifies the scene through online noise collection and offline judgment, and can only provide a reference basis for the offline design of the filter link, and cannot adjust the noise reduction measures in real time.

[0004] In order to reduce the false positive rate and the false negative rate of the distributed optical fiber sensing system in the detection scene, the prior art proposes a kind of optical fiber early warning method and device based on feature extraction model, the original data and the collected new data are divided into noise library and dynamic library according to whether they are identified as early warning signals, the dynamic library is used to store early warning signals, and the new data is added to the corresponding library in real time, the noise library and the dynamic library are updated accordingly, and the parameters are updated in real time;According to the updated parameters, the collected signals are calculated to determine the characteristic parameters and the warning threshold of the optical fiber safety state. The optical fiber early warning scheme mainly extracts the curve characteristic pulse, updates the characteristic parameters and the early warning parameters for judging the safety state of the optical fiber, and the extracted curve characteristic pulse cannot represent different types of noise, therefore, it cannot identify the types of noise and does not have the ability to learn the characteristics of new noise. SUMMARY

[0005] The present application aims to solve the technical problems of low efficiency and high cost of noise suppression technology in the prior art, and to achieve the purpose of real-time adjustment of noise suppression measures for ODVS terminal.

[0006] The present application adopts the following technical solutions to achieve the above-mentioned purposes:

[0007] A kind of online noise suppression method for ODVS engineering application system based on machine learning, comprising the following steps:

[0008] Step 1, classify the noise of ODVS system;

[0009] Step 2, establish a scene prior noise model according to the noise model of each type of ODVS system, and establish a noise pool;

[0010] Step 3, set the noise suppression algorithm parameters of ODVS terminal according to the scene prior noise model;

[0011] Step 4, put the noise collected by ODVS system into the noise pool;

[0012] Step 5, extract the features of the noise collected by ODVS system, and adjust the noise parameters in the noise pool;

[0013] Step 6, determine the noise model of ODVS system according to the adjusted noise parameters in step 5;

[0014] Step 7, feed back the ODVS system noise model determined in step 6 to the noise pool;

[0015] Step 8, adjust the parameters of the noise suppression algorithm of ODVS system online according to the ODVS system noise model determined in step 6.

[0016] As a further optimization scheme of the online noise suppression method of the ODVS engineering application system based on machine learning, the method for classifying the noise of the ODVS system in step 1 is as follows: the noise of the ODVS engineering application system is classified into categories including but not limited to background noise, perceived cable link optical noise, and terminal noise.

[0017] As a further optimization scheme of the online noise suppression method of the ODVS engineering application system based on machine learning, the method for establishing a scene prior noise model according to the noise model of each type of ODVS system in step 2 is as follows: the noise parameters of the ODVS engineering application system are set according to the scene prior method, the noise model corresponding to the set noise parameters is superimposed, and a scene prior noise model corresponding to the set noise parameters is obtained.

[0018] As a further optimization scheme of the online noise suppression method of the ODVS engineering application system based on machine learning, the noise parameters of the generated scene prior noise model are set as the noise suppression algorithm parameters of the ODVS terminal in step 3.

[0019] As a further optimization scheme of the online noise suppression method of the ODVS engineering application system based on machine learning, the method for extracting the characteristics of the noise collected by the ODVS system in step 5 includes but is not limited to time segmentation, Fourier transform, wavelet window, and time and frequency domain analysis. The noise parameters in the noise pool are adjusted by comparing the noise parameters and the fitting data.

[0020] As a further optimization scheme of the online noise suppression method of the ODVS engineering application system based on machine learning, the specific method for determining the ODVS system noise model according to the adjusted noise parameters in step 5 is as follows: the noise model corresponding to the adjusted noise parameters in step 5 is superimposed to obtain the ODVS system noise model.

[0021] As a further optimization scheme of the online noise suppression method of the ODVS engineering application system based on machine learning, the specific method for adjusting the parameters of the noise suppression algorithm of the ODVS system online according to the ODVS system noise model determined in step 6 is as follows: the noise suppression algorithm parameters of the ODVS terminal are adjusted to be the noise parameters corresponding to the ODVS system noise model.

[0022] As a further optimization scheme of the online noise suppression method of the ODVS engineering application system based on machine learning, the expression for superimposing the noise model corresponding to the set noise parameters is as follows: a1*N1(x)+a2*N2(x)+……+ai*Ni(x), where a1, a2, …, ai are the first to the i-th type of noise parameters, and N1(x), N2(x), …, Ni(x) represent the models of the first, second, …, i-th type of noise.

[0023] An electronic device includes a memory and a processor, wherein the memory stores a computer program that runs on the processor, and the processor executes the steps of the above-described online noise reduction method for ODVS engineering application systems when running the computer program.

[0024] A computer-readable storage medium having a computer program stored thereon, wherein the computer program executes the steps of the above-described online noise reduction method for the ODVS engineering application system when it is run.

[0025] The present invention adopts the above-mentioned technical solution and has the following beneficial effects: The online noise suppression method for ODVS engineering application system proposed in this invention includes updating the closed-loop feedback of the noise pool through machine learning, suppressing the closed-loop feedback of ODVS noise online through machine learning, and continuously enriching and improving the scene noise of the ODVS system through machine learning, making the noise pool more versatile. By adjusting the ODVS noise suppression algorithm online, the signal-to-noise ratio of the ODVS system signal processing can be continuously improved, and the accuracy of the system can be improved. Compared with the existing noise suppression technology for ODVS engineering application system, this method can greatly improve the efficiency of on-site engineering debugging and reduce debugging costs. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall scheme of the online noise reduction method proposed in this invention.

[0027] Figure 2 This is a flowchart of the overall scheme of the online noise reduction method proposed in this invention.

[0028] Figure 3 This is a schematic diagram illustrating the classification and summarization of ODVS system noise based on a priori methods according to the present invention.

[0029] Figure 4 A flowchart for establishing the noise model for this invention.

[0030] Figure 5 This is a schematic diagram of the coherent interference ODVS system according to an embodiment of the present invention.

[0031] Figure 6 for Figure 5 The waveform of the noisy signal received by the coherent interference ODVS terminal is shown.

[0032] Figure 7 To Figure 5 The noise voltage waveform after feature extraction of the signal within the rectangle shown is displayed.

[0033] Figure 8 According to Figure 7 A GUI interface for adjusting noise parameters using extracted feature parameters.

[0034] Figures 9(a) and 9(b) are respectively based onFigure 7 Waveforms of the time-domain and frequency-domain noise models of the ODVS system after feature extraction.

[0035] Figures 10(a) and 10(b) show the noise-suppressed time-domain and frequency-domain signals obtained after adjusting the ODVS noise reduction algorithm online according to the noise model shown in Figure 9, respectively.

[0036] Figure 11 The waveform of the noise voltage before adjusting the software parameters of the ODVS noise suppression algorithm online. Detailed Implementation

[0037] The technical solution of the invention will now be described in detail with reference to the accompanying drawings.

[0038] like Figure 1 As shown, the overall framework of the online noise suppression method for ODVS engineering application systems based on machine learning proposed in this invention includes two closed-loop feedback mechanisms. The first closed-loop feedback mechanism updates the noise pool through machine learning, specifically including: adding noise collected by the ODVS sensing optical cable to the ODVS noise pool; extracting noise features and adjusting the noise parameters in the noise pool; determining the ODVS system noise model based on the adjusted noise parameters; and feeding the determined ODVS system noise model back to the noise pool. The second closed-loop feedback mechanism suppresses ODVS noise online through machine learning, specifically including: adding noise collected by the ODVS sensing optical cable to the ODVS noise pool; extracting noise features and adjusting the noise parameters in the noise pool; determining the ODVS system noise model based on the adjusted noise parameters; and adjusting the ODVS noise reduction algorithm parameters online based on the determined ODVS system noise model.

[0039] like Figure 2 As shown, the overall scheme of the online noise reduction method for ODVS engineering application system based on machine learning proposed in this invention includes the following 9 steps.

[0040] Step 1: Classify the noise of the ODVS system.

[0041] like Figure 3As shown, the noise of the ODVS system is classified and summarized, and the noise of the ODVS engineering application system is divided into three categories: background noise, sensing optical cable link optical noise, and terminal noise. The background noise is further divided into natural noise, human noise, and rainfall noise. The sensing optical cable link optical noise is further divided into interference fading noise, polarization fading noise, and fiber loss. The terminal noise is further divided into light source module noise, optical path module noise, and optoelectronic receiving module noise. The light source module noise is further divided into ultra-wideband laser intensity noise and narrowband laser phase noise. The optical path module noise is further divided into modulator noise and EDFA noise. The optoelectronic receiving module noise is further divided into photodetector noise and amplifier noise. Figure 3 Only one noise classification and summary method applicable to the present application is exemplarily illustrated. Those skilled in the art can also classify the noise more comprehensively and meticulously according to the actual engineering application requirements.

[0042] Step 2, establishing a scene prior noise model to form a parameter-adjustable noise pool

[0043] According to the scene prior method, the noise parameters of the engineering application system are set, and the scene prior noise model is generated. The noise parameters include but are not limited to noise decibels, gain coefficients, loss coefficients, and light source line widths. As shown, Figure 4 As shown, each scene prior noise model is generated by a noise program. Specifically, the values of the first to i-th noise parameters a1, a2, …, ai are substituted into the mathematical model a1*N1(x)+a2*N2(x)+…+ai*Ni(x) to generate a scene prior noise model. N1(x), N2(x), …, Ni(x) represent the models of the first, second, …, i-th noise. The scene prior noise waveform graph obtained after normalizing the output data of the generated noise model is stored in the noise library.

[0044] Step 3, setting the noise suppression algorithm parameters of the ODVS terminal

[0045] The noise parameters of the generated scene prior noise model are set as the noise suppression algorithm parameters of the ODVS terminal, and the preparation work for terminal noise reduction and suppression is completed.

[0046] Step 4, putting the noise collected by the ODVS system sensing optical cable into the noise pool

[0047] Step 5, extracting the features of the noise collected by the ODVS system sensing optical cable, and adjusting the noise parameters in the noise pool

[0048] First, feature extraction methods, including but not limited to time segmentation, Fourier transform, wavelet windowing, and time-domain frequency domain analysis, are used to obtain the noise parameters collected by the ODVS system sensing optical cable. The noise parameters collected by the ODVS system sensing optical cable include, but are not limited to, noise voltage. Next, the extracted noise parameters are compared with the noise parameters of the scene prior noise model in the noise pool. The noise parameters of the scene prior noise model are adjusted by fitting the data. The noise parameters of the noise pool can be adjusted through the classification interface of the noise pool.

[0049] Step 6: Determine the ODVS system noise model based on the noise parameters adjusted in Step 5.

[0050] Optionally, run Figure 4 The program shown generates a noise model of the ODVS system based on the noise parameters adjusted in step 5. The noise model of the ODVS system includes a time-domain model and a frequency-domain model.

[0051] Step 7: Feed back the ODVS system noise model determined in Step 6 to the noise pool.

[0052] Based on the scene noise model, the noise model established by real-time noise collection from the optical fiber sensing cable of the ODVS system is fed back to the noise pool, forming a closed-loop feedback for the machine learning part of the noise pool.

[0053] Step 8: Adjust the parameters of the ODVS system noise suppression algorithm online based on the ODVS system noise model determined in Step 6.

[0054] Based on the time-domain and frequency-domain characteristics of the ODVS system noise model determined in step 6, the noise suppression algorithm parameters of the ODVS terminal are adjusted online through the control software in the ODVS terminal backend to match the noise parameters corresponding to the ODVS system noise model. This forms a closed-loop feedback mechanism for ODVS noise suppression based on machine learning, thereby optimizing the noise reduction effect online.

[0055] Next, an example will be used to illustrate the specific application of the ODVS online noise reduction method proposed in this invention in engineering.

[0056] The laboratory has developed a coherent interferometric ODVS system. The system structure is as follows: Figure 5 As shown, the PIN-FET is the photodetector module, which, together with the amplifier (Amp), the data acquisition system (DAQ), and the PC, forms the main components of the terminal unit. After the laser emits a light pulse, it is amplified by the EDFA, and then the noise introduced by the EDFA is filtered by a filter to ensure stable output light power gain. After that, it enters the circulator through port 1 and the sensing fiber through port 2. In the sensing fiber, the backscattered Rayleigh signal is output through port 3 of the circulator, and then interferes with the local oscillator light at coupler C2, where it is collected by the photodetector.

[0057] Figure 6 Fig. 9 is a waveform diagram of a noise signal received by a coherent interference ODVS terminal machine, wherein the horizontal axis is distance, i.e. L in Fig. 8, and the vertical axis is signal intensity, and the selected region of the distance window corresponds to 950m-1050m. Figure 5 Figure 5 s , wherein the horizontal axis is distance, i.e. L in Fig. 8, and the vertical axis is signal intensity, and the selected region of the distance window corresponds to 950m-1050m.

[0058] Figure 7 Fig. 9 is a waveform diagram of a noise signal received by a coherent interference ODVS terminal machine, wherein the horizontal axis is distance, i.e. L in Fig. 8, and the vertical axis is signal intensity, and the selected region of the distance window corresponds to 950m-1050m. Figure 5

[0059] Figure 8 Fig. 9 is a waveform diagram of a noise signal received by a coherent interference ODVS terminal machine, wherein the horizontal axis is distance, i.e. L in Fig. 8, and the vertical axis is signal intensity, and the selected region of the distance window corresponds to 950m-1050m. Figure 7

[0060] Fig. 9 is a waveform diagram of a noise signal received by a coherent interference ODVS terminal machine, wherein the horizontal axis is distance, i.e. L in Fig. 8, and the vertical axis is signal intensity, and the selected region of the distance window corresponds to 950m-1050m. Figure 7

[0061] Fig. 9 is a waveform diagram of a noise signal received by a coherent interference ODVS terminal machine, wherein the horizontal axis is distance, i.e. L in Fig. 8, and the vertical axis is signal intensity, and the selected region of the distance window corresponds to 950m-1050m. Figure 11

[0062] In an embodiment of the present application, an electronic device is also provided, which comprises a memory and a processor, and the memory stores a computer program running on the processor, and the processor executes the steps of the ODVS engineering application system online noise suppression method as described above when running the computer program.

[0063] In an embodiment of the present application, a computer readable storage medium is also provided, which stores a computer program, and the computer program executes the steps of the ODVS engineering application system online noise suppression method as described above when running.

[0064] ​​​​​​The above specific embodiments and examples are specific supports for the technical idea of the present application, and cannot limit the protection scope of the present application. Any equivalent changes or equivalent modifications made according to the technical idea of the present application and based on the technical solutions of the present application still belong to the protection scope of the technical solutions of the present application.

Claims

1. A machine learning-based ODVS engineering application system online noise suppression method, characterized in that, The method comprises the following steps: Step 1: classifying the noise of the ODVS engineering application system: classifying the noise of the ODVS engineering application system into categories including background noise, sensing cable link optical noise, and terminal noise; Step 2: establishing a scene prior noise model of the ODVS system noise according to the various categories of ODVS engineering application system noise, and establishing a noise pool; Step 3: setting the ODVS terminal noise suppression algorithm parameters according to the scene prior noise model; Step 4: putting the noise sensed and collected by the ODVS system into the noise pool; Step 5: extracting the features of the noise sensed and collected by the ODVS system, comparing the extracted features with the noise parameters of the scene prior noise model in the noise pool, and adjusting the noise parameters in the noise pool; Step 6: determining the ODVS system noise model according to the noise parameters adjusted in step 5; Step 7: feeding back the ODVS system noise model determined in step 6 to the noise pool; Step 8: adjusting the parameters of the ODVS system noise suppression algorithm online according to the ODVS system noise model determined in step 6.

2. The online noise suppression method of the ODVS engineering application system based on machine learning according to claim 1, wherein, The method of step 2 for establishing a scene prior noise model of the ODVS system noise according to the various categories of ODVS engineering application system noise is: setting the parameters of the various categories of ODVS engineering application system noise according to the scene prior method, superimposing the noise models corresponding to the set noise parameters to obtain the scene prior noise model corresponding to the set noise parameters.

3. The online noise suppression method of the ODVS engineering application system based on machine learning according to claim 1, wherein, The noise parameters of the generated scene prior noise model are set as the ODVS terminal noise suppression algorithm parameters in step 3.

4. The online noise suppression method of the ODVS engineering application system based on machine learning according to claim 1, wherein, In step 5, the features of the noise sensed and collected by the ODVS system are extracted by using methods including but not limited to time segmentation, Fourier transform, wavelet window, time domain and frequency domain analysis, the extracted features are compared with the noise parameters of the scene prior noise model in the noise pool, and the noise parameters in the noise pool are adjusted through data fitting operations.

5. The online noise suppression method of the ODVS engineering application system based on machine learning according to claim 1, wherein, The specific method of step 6 for determining the ODVS system noise model according to the noise parameters adjusted in step 5 is: superimposing the noise models corresponding to the noise parameters adjusted in step 5 to obtain the ODVS system noise model.

6. The online noise suppression method of an ODVS engineering application system based on machine learning according to claim 1, wherein, The specific method of step 8 for adjusting the parameters of the ODVS system noise suppression algorithm online according to the ODVS system noise model determined in step 6 is: online tuning the ODVS terminal noise suppression algorithm parameters to the noise parameters corresponding to the ODVS system noise model.

7. The online noise suppression method of the ODVS engineering application system based on machine learning according to claim 2, wherein, The superposition expression of the noise models corresponding to the set noise parameters is: a1*N1(x)+a2*N2(x)+……+ai*Ni(x), where a1, a2, …, ai are the 1st to i th noise parameters, and N1(x), N2(x), …, Ni(x) represent the models of the 1st, 2nd, …, i th noise.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program which runs on the processor, and the processor executes the steps of the ODVS engineering application system online noise suppression method of claim 1 when running the computer program.

9. A computer readable storage medium having stored thereon a computer program which, when executed, performs the steps of the ODVS engineering application system online noise suppression method of claim 1.

Images