Method for realizing real-time acquisition of high-density vibration signals by parallel computing architecture

Abstract

Aiming at real-time acquisition of distributed optical fiber high-density vibration signal data, the invention provides a heterogeneous parallel computing architecture based on a central processing unit and a graphics processing unit, which is used for realizing real-time acquisition of the distributed optical fiber high-density vibration signal data, and belongs to a real-time acquisition method of the distributed optical fiber high-density vibration signal data. According to the heterogeneous parallel acquisition scheme, double-channel distributed optical fiber high-density vibration signal sensing data acquired by a high-speed data acquisition card is subjected to parallel acquisition by using a heterodyne acquisition algorithm, and the acquired data is subjected to waterfall plot real-time display, storage and the like. According to the invention, real-time acquisition of distributed optical fiber high-density vibration signal sensing data can be realized. The scheme has the advantages of simplicity in implementation, flexibility, easiness in use and the like, and is particularly suitable for related theoretical verification and research and development of a distributed optical fiber high-density vibration system.

Description

Technical Field

[0002] The present invention relates to the field of distributed optical fiber high-density vibration signal sensing, which is a method for real-time acquisition of high-density vibration signals using a heterogeneous parallel computing architecture. Background Art

[0004] Distributed fiber optic vibration sensing systems are widely used in geological exploration, rail transit, hydroacoustic detection, structural health monitoring and other fields due to their technical characteristics such as long distance, large scale and anti-electromagnetic interference. However, due to its distributed sensing characteristics and the increasing complexity of the acquisition algorithm, the acquisition of sensor data of distributed fiber optic vibration systems faces huge computing power challenges. The usual practice is to collect the sensor data of the distributed fiber optic vibration system and process it offline. However, for some specific application fields, such as the application of distributed fiber optic vibration systems in the field of hydroacoustic detection, it is necessary to detect and locate underwater targets in real time, and the offline acquisition solution cannot meet the actual application requirements. Therefore, the real-time acquisition technology of sensor data of distributed fiber optic vibration systems has become a research hotspot. In order to realize the real-time acquisition of high-density sensor data of distributed fiber optic vibration systems, the present invention proposes a method for real-time acquisition of high-density vibration signals using a heterogeneous parallel computing architecture. Summary of the invention

[0006] In order to realize the real-time acquisition of high-density sensing data of distributed optical fiber vibration system and solve the real-time acquisition problem of distributed optical fiber vibration system, the present invention proposes a method for realizing real-time acquisition of high-density vibration signals using a heterogeneous parallel computing architecture. The scheme has the technical characteristics of flexibility, ease of use, and simple implementation, and provides a new technical solution for the real-time acquisition of distributed optical fiber vibration sensing data.

[0007] The technical solution of the present invention:

[0008] The invention is characterized in that it comprises a narrow line width laser (1), an optical coupler 1 (2), an optical coupler 2 (6), an acousto-optic modulator 1 (3), an acousto-optic modulator 1 (4), a delay optical fiber (5), an erbium-doped optical fiber amplifier (7), an optical circulator (8), an optoelectronic converter (9), a high-speed data acquisition card (10), a heterogeneous parallel computing architecture (11), and a sensing optical fiber (12). The output laser of the ultra-narrow linewidth laser (1) is divided into two paths through the optical coupler 1 (2), and the frequencies are shifted to different frequencies through the acousto-optic modulator 1 (3) and the acousto-optic modulator 2 (4) respectively. The output light of the acousto-optic modulator 2 (4) passes through a delay optical fiber (5) and is combined with the output light of the acousto-optic modulator 1 (3) to generate a heterodyne double-pulse laser pair. After being amplified by the erbium-doped fiber amplifier (7), the laser pair enters the sensing optical fiber (12) through the port ① of the optical circulator (8). The photoelectric converter (9) converts the optical sensing signal of the sensing optical fiber (12) returned from the port ③ of the optical coupler (8) into an electrical signal, which is then collected by the high-speed data acquisition card (10) and sent to the central processing unit + graphics processing unit heterogeneous parallel computing architecture (11) for real-time acquisition.

[0009] When the sensing optical fiber (12) detects the sensing signal, the electrical signal converted by the high-speed data acquisition card (10) first enters the dynamic random access memory (11-4) of the central processing unit (11-10). Under the control of the central processing unit controller (11-1), the data is sent to the video memory (11-8) of the graphics processing unit via the data bus (11-9). At this time, the graphics processing unit (11-11) starts hundreds, thousands, or even tens of thousands of threads (11-11-1) according to parameter settings, and performs high-speed parallel processing on the incoming input. This process is not controlled by the central processing unit (11-10) and is independently completed by the graphics processing unit (11-11). Under the control of the graphics processor thread controller (11-5), each thread (11-11-1) of the graphics processor (11-11) processes data in the graphics processor video memory (11-8) in combination with the graphics processor arithmetic logic unit (11-6) and the graphics processor cache (11-7), and saves the processing results in the graphics processor video memory (11-8).

[0010] Further, after all data processing is completed, the central processing unit (11-10) and the graphics processing unit (11-11) re-establish communication, and under the control of the central processing unit (11-10), the data is sent back to the dynamic random access memory (11-4) of the central processing unit via the data bus (11-9) to wait for subsequent processing. In the process of data transmission between the central processing unit (11-10) and the graphics processing unit (11-11), the central processing unit controller (11-1) participates in it. In the subsequent processing process of obtaining data, under the control of the central processing unit controller (11-1), the data enters the central processing unit arithmetic logic unit (11-2) from the central processing unit dynamic random access memory (11-4) through the central processing unit cache (11-3) for processing.

[0011] During the entire acquisition process, the central processing unit (11-10) and the graphics processing unit (11-11) are only forced to synchronize and establish a connection during the data transmission process. During the acquisition process, the graphics processing unit (11-11) completes the parallel computing task independently and is not controlled by the central processing unit (11-10). This process is the so-called heterogeneous parallelization.

[0012] In summary, the present invention provides a method for real-time acquisition of high-density vibration signals using a heterogeneous parallel computing architecture. The heterogeneous parallel computing architecture acquisition solution based on a central processing unit + a graphics processing unit has the technical characteristics of simple implementation and flexible application. It can be applied not only to the field of distributed optical fiber vibration, but also to other fields with high computing power requirements, such as aerospace, intelligent driving, etc. Since the acquisition solution has a short compilation time and convenient and flexible programming implementation, it is particularly suitable for the improvement and verification of theories and systems in the scientific research stage, providing powerful computing power support for early theoretical scientific research innovation and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram of a method for real-time acquisition of high-density vibration signals based on a heterogeneous parallel computing architecture in an embodiment;

[0015] Figure 2 A schematic diagram of a structure for implementing real-time acquisition of high-density vibration signals for a distributed system based on a heterogeneous parallel computing architecture in an embodiment;

[0016] Figure 3 It is a schematic diagram of the principle of the parallel computing structure of a graphics processor for implementing real-time acquisition of a distributed optical fiber vibration system based on a heterogeneous parallel computing architecture in an embodiment. DETAILED DESCRIPTION

[0018] In order to more specifically explain the implementation modes, technical solutions and advantages of the present invention, a detailed description is given below in conjunction with the drawings of the implementation modes of the present invention. The real-time method described does not represent all implementation modes of the present invention.

[0019] like Figure 1 The figure shows a schematic diagram of the structure of the high-density vibration signal real-time acquisition technology based on the heterogeneous parallel computing architecture of the present invention. The acquisition scheme uses the powerful data parallel processing capability of the graphics processor to perform real-time acquisition operations on the distributed optical fiber vibration sensor data collected from the high-speed data acquisition card.

[0020] The specific implementation scheme is as follows: a narrow linewidth laser (1) emits a narrow linewidth laser beam, which is split by an optical coupler 1 (2) and modulated into two laser pulses of different frequencies by an acousto-optic modulator 1 (3) and an acousto-optic modulator 2 (4), respectively. The output light pulse of the acousto-optic modulator 2 (4) enters a delay optical fiber (5) and is combined with the output light pulse of the acousto-optic modulator 1 (3) to generate a heterodyne double-pulse laser pair. After power amplification by an erbium-doped fiber amplifier (7), the laser beam enters a sensing optical fiber (12) through port ① of an optical circulator (8). A photoelectric converter (9) detects the optical sensing signal returned by port ③ of the optical circulator (8), converts the optical sensing signal into an electrical signal, collects the signal through a high-speed data acquisition card (10), and sends the signal to a heterogeneous parallel computing architecture (11) for real-time acquisition.

[0021] Furthermore, the schematic diagram of the heterogeneous parallel computing architecture is as follows Figure 2 When the sensing signal detected by the sensing optical fiber (12) is converted into an electrical signal by the high-speed data acquisition card (10), the data first enters the dynamic random access memory (11-4) of the central processing unit under the control of the central processing unit controller (11-1), and then is sent to the display memory (11-8) of the graphics processing unit via the data bus (11-9).

[0022] Furthermore, the graphics processor (11-11) starts hundreds, thousands, or even tens of thousands of threads (11-11-1) according to parameter settings, such as Figure 3 As shown, the incoming input is processed in parallel at high speed, and the parallel processing process is not controlled by the central processing unit (11-10), but is independently completed by the graphics processing unit (11-11). Under the control of the graphics processing unit controller (11-5), each thread (11-11-1) of the graphics processing unit (11-11) processes the data in the graphics processing unit video memory (11-8) in combination with the graphics processing unit arithmetic logic unit (11-6) and the graphics processing unit cache (11-7), and saves the processing results in the graphics processing unit video memory (11-8).

[0023] Further, after all data processing is completed, the central processing unit (11-10) and the graphics processing unit (11-11) re-establish communication. Under the control of the central processing unit (11-10), the data is sent back to the dynamic random access memory (11-4) of the central processing unit via the data bus (11-9) to wait for subsequent processing. In the process of data transmission between the central processing unit (11-10) and the graphics processing unit (11-11), the central processing unit controller (11-1) participates in it. In the subsequent processing of acquiring data, under the control of the central processing unit controller (11-1), the data enters the central processing unit operation logic unit (11-2) through the central processing unit cache (11-3) for processing. In the entire acquisition process, the central processing unit (11-10) and the graphics processing unit (11-11) only perform forced synchronization and establish connection during the data transmission process. In the acquisition process, the graphics processing unit (11-11) completes the parallel computing task alone and is not controlled by the central processing unit (11-10). This process is the so-called heterogeneous parallel.

[0024] In summary, the present invention provides a method for obtaining high-density vibration signals using a heterogeneous parallel computing architecture of a central processing unit and a graphics processing unit, which can obtain high-density sensing data from a distributed optical fiber vibration system in real time. This technical solution has the technical characteristics of simple technical threshold and flexible application. It can be applied not only to the field of distributed optical fiber vibration sensing, but also to other scenarios with high computing power requirements, such as artificial intelligence, high-speed signal processing, etc. Since the acquisition solution has a short compilation time and convenient and flexible programming implementation, it is particularly suitable for the early theoretical innovation, algorithm improvement and verification stages of the scientific research stage, providing huge computing power support for early theoretical scientific research and has broad application prospects.

Claims

1. A method for real-time acquisition of high-density vibration signals using a parallel computing architecture, characterized in that: The invention comprises a narrow linewidth laser (1), an optical coupler 1 (2), an optical coupler 2 (6), an acousto-optic modulator 1 (3), an acousto-optic modulator 1 (4), a time-delay optical fiber (5), an erbium-doped optical fiber amplifier (7), an optical circulator (8), an optoelectronic converter (9), a high-speed data acquisition card (10), a heterogeneous parallel computing architecture (11), and a sensing optical fiber (12), wherein the output port of the narrow linewidth laser (1) is connected to the optical coupler 1 (2) and is divided into two paths, and the frequency is shifted to different frequencies through the acousto-optic modulator 1 (3) and the acousto-optic modulator 2 (4) respectively. The output port of the acousto-optic modulator 2 (4) is connected to a section of delay optical fiber (5), and its output is combined with the output of the acousto-optic modulator 1 (3) through an optical coupler 2 (6) to generate a heterodyne double-pulse laser pair, which is amplified by an erbium-doped fiber amplifier (7) and enters the sensing optical fiber (12) through port ① of an optical circulator (8). The photoelectric converter (9) converts the optical sensing signal detected by the sensing optical fiber (12) returned from port ③ of the optical coupler (8) into an electrical signal, which is then collected by a high-speed data acquisition card (10) and sent to a heterogeneous parallel computing architecture (11) for real-time acquisition.

2. The method for realizing real-time acquisition of sensor data of a distributed optical fiber vibration system based on a heterogeneous parallel computing architecture according to claim 1, characterized in that: The heterogeneous parallel computing architecture (11) is formed by combining a central processing unit (11-10) and a graphics processing unit (11-11). The central processing unit (11-10) internally includes: a central processing unit controller (11-1), a central processing unit operation logic unit (11-2), a central processing unit cache (11-3), and a central processing unit dynamic random access memory (11-4); the graphics processing unit (11-11) internally includes: a graphics processing unit thread controller (11-5), a graphics processing unit operation logic unit (11-6), a graphics processing unit cache (11-7), and a graphics processing unit video memory (11-8). Data transmission between the central processing unit (11-10) and the graphics processing unit (11-11) is connected via a data bus (11-9).

3. The method for realizing real-time acquisition of sensor data of a distributed optical fiber vibration system based on a heterogeneous parallel computing architecture according to claim 1, characterized in that: Under the control of the central processing unit (11-10), the graphics processing unit (11-11) assists the central processing unit (11-10) in processing data. In the heterogeneous parallel computing of the central processing unit and the graphics processing unit, the graphics processing unit (11-11) often plays the role of a coprocessor. When writing a graphics processing unit application, it is necessary to write the central processing unit and graphics processing unit codes. The graphics processing unit (11-11) needs to complete the computing task under the control of the central processing unit (11-10) and pass the computing result to the central processing unit (11-10). The communication between the central processing unit (11-10) and the graphics processing unit (11-11) is carried out through the data bus (11-9) for data transmission.

4. The heterogeneous parallel computing architecture according to claim 2, characterized in that: The high-speed data acquisition card (10) transmits the collected distributed optical fiber vibration sensing data to the dynamic random access memory (11-4) of the central processing unit, and under the control of the central processing unit controller (11-1), transmits the collected data to the video memory (11-8) of the graphics processing unit through the data bus (11-9). The graphics processing unit thread controller (11-5) allocates a thread (11-11-1), an operation logic unit (11-6) and a cache space (11-7) to each equivalent array data segment according to the parallel characteristics of the distributed optical fiber vibration data, and acquires and processes the distributed optical fiber vibration data. The acquisition result is stored in the video memory (11-8) under the control of the graphics processing unit controller, and finally transmitted back to the dynamic random access memory (11-4) of the central processing unit through the data bus (11-9) to wait for subsequent processing.

Images