CN-121679480-B - DAS multipoint joint vibration source positioning method based on PINO-EPFES-TDOA

Abstract

The invention discloses a DAS multipoint joint vibration source positioning method based on PINO-EPFES-TDOA, which belongs to the field of signal processing and machine learning, and specifically comprises the steps of firstly inputting time slots Lower position The DAS original sensing signal is denoised, then the sensing DAS is utilized to obtain the position information of sampling points, the optimal propagation speed of the soil medium is calculated, the loss function of a PINO medium propagation model is calculated based on the propagation speed and the denoised signal, and the acoustic wave attenuation coefficient in the soil is obtained And in the boundary of the optimal characteristic interval, a TDOA sampling point time delay estimation model is established, a multi-point joint positioning error model is finally established, and a vibration source position is obtained by carrying out minimum solution. The invention provides a method with higher positioning precision for the vibration source positioning field.

Inventors

• WANG SONG
• LIANG ZIBIN
• HU YANZHU
• WANG YANHONG

Assignees

• 北京邮电大学

Dates

Publication Date

20260508

Application Date

20251212

Claims (4)

1. 1. A DAS multipoint joint vibration source positioning method based on PINO-EPFES-TDOA is characterized by comprising the following specific steps: Step one, inputting time slots Lower position Is subjected to de-noising to obtain de-noised signals : Step two, the first acquired by sensing DAS And (b) Position information of each sampling point, and calculating optimal propagation speed of soil medium : Step three, utilizing the denoising signal And optimum propagation velocity of soil medium Calculating PINO a loss function of a medium propagation model Obtaining the attenuation coefficient of sound wave in the soil ; Loss function The calculation formula is as follows: ; in the formula, Representing PINO the output of the final iteration of the neural network operator, Representing the attenuation coefficient of sound waves in the soil; Is a weight coefficient; optimal propagation velocity of soil medium Substitution loss function Auxiliary solution by minimizing loss function : ; Step four, utilizing the attenuation coefficient of sound wave in the soil Establishing a soil medium propagation model : ; In the formula, A soil medium propagation model; As a parameter of the amplitude value, ; For correction parameters, empirically defined; fifthly, calculating a soil medium propagation model by utilizing an energy main peak characteristic extraction strategy EPFES An optimal characteristic interval boundary of the energy main peak; The specific calculation process is as follows: First, a soil medium propagation model is calculated Energy main peak of (2) : ; In the formula, Sensing signals at different positions at the time t with maximum energy; Then, the energy main peak is determined Is defined in the following description: ; in the formula, Taking the position point of the maximum energy value as the position point Distance from front to back As interval boundaries, i.e. the optimal feature interval is ; Step six, establishing a TDOA sampling point time delay estimation model in the boundary of the optimal characteristic interval : ; In the formula, Representing coordinates of sampling points in which a coordinate system is established in the direction of optical fiber detection , , M represents indexes of different positions within the optimal characteristic interval; Expressed in terms of principal peak points Is the reference point position An estimate of the time delay between the two, Coordinates are Expressed as the dominant peak point The coordinates of the reference points correspond to the maximum energy positions; Is the vibration source position ; Step seven, utilizing a time delay estimation model Establishing a multipoint joint positioning error model, and carrying out minimization solving to obtain the vibration source position: ; in the formula, A multi-point joint positioning error model; ; solving for vibration source position by minimizing the model 。
2. 2. The method of claim 1, wherein the step one is specifically calculated as: ; in the formula, Representing a wavelet operator; Representing a threshold function; Representing time slots Lower position Is a DAS raw sense signal.
3. 3. The method of claim 1, wherein the step two is specifically calculated as: ; in the formula, Representing the propagation velocity of sound waves in the soil, Represent the first And (b) The distance between the sampling points; representing the experimentally measured first And (b) Time delay of the sampling points; , To sense all the sample point numbers of the DAS.
4. 4. The method of claim 1, wherein in step three, the time slots are targeted Lower position By PINO neural network operators Layer iteration to obtain output : In the formula, The number of layers of the operator of the neural network is represented, A linear projection is represented and is shown, Representing a fourier transform; representing a frequency domain convolution kernel; an initial value is set manually.

Description

DAS multipoint joint vibration source positioning method based on PINO-EPFES-TDOA Technical Field The invention belongs to the field of signal processing and machine learning, and particularly relates to a DAS (distributed acoustic sensing) multipoint combined vibration source positioning method based on PINO-EPFES-TDOA (physical constraint neural network-energy main peak feature extraction strategy-arrival time difference ,Physics-Informed Neural Network-Energy Peak Feature Extraction Strategy-Time Difference of Arrival）). Background At present, aiming at the DAS vibration source positioning problem, the method is mainly realized by calculating an arrival angle (AOA) and Received Signal Strength (RSS), and the general processing flow is to firstly fit signals, establish a propagation model in a medium and then solve according to an energy propagation estimation algorithm to realize positioning. For the medium propagation modeling problem of DAS signals, common methods include a ray tracing method based on uniform medium assumption, a simplified modeling method based on an empirical green function and the like, but the former has lower positioning precision, the latter has poorer simplified modeling method, and in addition, due to strong non-uniformity and anisotropy of underground medium, conditions such as bending of a propagation path, abnormal speed structure and the like can occur, so that the positioning precision of a vibration source is influenced. The traditional method based on beam forming has good spatial resolution capability for vibration source positioning, but has the problems of limited angular resolution, high array calibration requirement, high calculation complexity and the like, and particularly has the possibility of angle blurring, positioning accuracy reduction and the like for some short-distance vibration sources. Regarding the application of DAS vibration source positioning, the development in many fields is relatively mature. For example, better vibration source positioning effects are achieved in aspects of threat outside pipelines, perimeter security positioning and the like by using CNN, KNN and various neural network algorithms. Because of the higher and higher safety requirements of people and the wide application of the distributed optical fiber sensing technology in the fields of aerospace, geological monitoring, industrial safety and the like, the DAS vibration source positioning precision and accuracy are required to be relatively high. Therefore, to accurately realize the vibration source positioning in real time and meet the requirement of vibration source positioning, an efficient and accurate DAS vibration source positioning method must be established, so that the positioning error is effectively reduced, the positioning precision of DAS vibration source positioning is improved, real-time and accurate vibration source positioning results are provided for a plurality of application fields of distributed optical fiber sensing technology, and the problem can be found in time by workers conveniently, and decisions can be made in advance. Disclosure of Invention The invention provides a method for positioning a DAS multipoint combined vibration source based on PINO-EPFES-TDOA, which realizes high-precision and real-time positioning of the vibration source and solves the problems of low positioning precision and poor fitting degree of the vibration source in a complex heterogeneous medium environment. The DAS multipoint joint vibration source positioning method based on PINO-EPFES-TDOA comprises the following specific steps: Step one, inputting time slots Lower positionIs subjected to de-noising to obtain de-noised signals: In the formula,Representing a wavelet operator; Representing a threshold function; Representing time slots Lower positionIs a DAS raw sense signal. Step two, the first acquired by sensing DASAnd (b)Position information of each sampling point, and calculating optimal propagation speed of soil medium: In the formula,Representing the propagation velocity of sound waves in the soil,Represent the firstAnd (b)The distance between the sampling points; representing the experimentally measured first And (b)Time delay of the sampling points;, To sense all the sample point numbers of the DAS. Step three, utilizing the denoising signalAnd optimum propagation velocity of soil mediumCalculating PINO a loss function of a medium propagation modelObtaining the attenuation coefficient of sound wave in the soil; The specific process is as follows: First, for a time slot Lower positionBy PINO neural network operatorsLayer iteration to obtain output: In the formula,The number of layers of the operator of the neural network is represented,A linear projection is represented and is shown,Representing a fourier transform; representing a frequency domain convolution kernel; Manually setting an initial value; Then, the PINO neural network operator is finally iterated to output