A NEW SOFTWARE IMPLEMENTATION OF ULTRASONIC PHASED ARRAY INSPECTION AND 3D IMAGING APPLICATION BASED ON GPU PARALLEL COMPUTING
详细信息 查看官网全文
- 英文论文题名：A NEW SOFTWARE IMPLEMENTATION OF ULTRASONIC PHASED ARRAY INSPECTION AND 3D IMAGING APPLICATION BASED ON GPU PARALLEL COMPUTING
- 论文作者：Da-liang GAO ; Fang-fang SHI ; Bi-xing ZHANG
- 英文论文作者：Da-liang GAO ; Fang-fang SHI ; Bi-xing ZHANG ; State Key Laboratory of Acoustics ; Institute of Acoustics ; Chinese Academy of Sciences
- 年：2016
- 作者机构：State Key Laboratory of Acoustics, Institute of Acoustics,Chinese Academy of Sciences;
- 英文论文关键词：Ultrasonic phased array ; GPU parallel computation ; Open GL ; CUDA
- 会议召开时间：2016-10-21
- 会议录名称：2016年全国压电和声波理论及器件应用研讨会论文摘要集
- 英文会议录名称：Abstract Book of 2016 Symposium on Piezoelectricity, Acoustic Waves and Device Applications
- 语种：英文
- 分类号：TB559
- 学会代码：AGLU
- 会议名称：2016年全国压电和声波理论及器件应用研讨会
- 会议地点：中国陕西西安
- 主办单位：中国力学学会、中国声学学会、IEEE UFFC分会
- 学会名称：中国力学学会
- 页数：2
- 文件大小：10k
- 原文格式：D
- 会议级别：全国
摘要
Background, Motivation and Objective Ultrasonic phased array technology has been widely studied and used in the field of Non-Destructive Testing. With the development of ultrasonic phased array technology, the sampling frequency gets higher and the number of independent channels increased at the same time. As a result, huge amounts of data are collected by the data acquisition module. For example, the size of data uploaded by PCI(Peripheral Component Interconnect) devices can be as much as 133 MB per second. It is hard for the CPU to finish the data processing procedure and display the result in time. However, the data collected by ultrasonic phased array system is divided into different channels inherently; the data of each channel can be processed through same algorithms with limit dependence on the data of other channels. In recent years, GPU parallel computing is developing rapidly. The GPU parallel computing is ideal for solving this kind of data-parallel computing problem. Statement of Contribution/Methods CUDA is a GPU parallel computing architecture provided by the NVIDA Corporation which is supported by most of its GPU products. With the help of CUDA, we can implement and optimize the parallel algorithm of the phased array signal processing. Open GL(Open Graphics Library) is the premier environment for developing 2D and 3D graphics applications. With the supports of these two tools, we create a software system. CUDA programs execute multiple instruction streams on different cores in GPU to realize parallel computing of input signals which includes measuring peak value, FIR filtering and FFT processing. After the data parallel in GPU, Open GL is used for displaying the 2D and 3D result image on the screen with no need to copy the data back to the internal memory. Algorithms are optimized in order to fit different data processing requirement and specific hardware in the computer. Results The test is based on Geforce GT430 GPU and Intel i5-2800@2.8GHz CPU. We use 256 channels wave data stored in local memory to simulate the data uploaded by real inspection equipment. Each channel of data has 512 16-bits sample points with the pulse repetition frequency of 512 Hz. That means the input data rate is 128MB/s. In the experiment, the software detects the peak value, arrival time of the signal and calculates the energy, then realizes FFT algorithm to get frequency spectrum. After the data processing, the software receives a succession of data and fuses the data to form 2D and 3D images, then outputs the result images in time. Comparisons between software using GPU and CPU are made. The results show that the new software system which is based on CUDA and Open GL can accelerate the data processing procedure at the speedup ratio of 8 to 40. The maximum throughput rate is 256MB/s. Discussion and Conclusions With the cheap and widely used GPU like GT430, we speed up data processing and image displaying considerably, especially 3D image real-time displaying. As the GPU Computing architecture like CUDA is becoming more and more mature and the GPU hardware is developing better and better, it has great potential in the implementation of software on Ultrasonic Phased Array Inspection and Imaging Application.
Background, Motivation and Objective Ultrasonic phased array technology has been widely studied and used in the field of Non-Destructive Testing. With the development of ultrasonic phased array technology, the sampling frequency gets higher and the number of independent channels increased at the same time. As a result, huge amounts of data are collected by the data acquisition module. For example, the size of data uploaded by PCI(Peripheral Component Interconnect) devices can be as much as 133 MB per second. It is hard for the CPU to finish the data processing procedure and display the result in time. However, the data collected by ultrasonic phased array system is divided into different channels inherently; the data of each channel can be processed through same algorithms with limit dependence on the data of other channels. In recent years, GPU parallel computing is developing rapidly. The GPU parallel computing is ideal for solving this kind of data-parallel computing problem. Statement of Contribution/Methods CUDA is a GPU parallel computing architecture provided by the NVIDA Corporation which is supported by most of its GPU products. With the help of CUDA, we can implement and optimize the parallel algorithm of the phased array signal processing. Open GL(Open Graphics Library) is the premier environment for developing 2D and 3D graphics applications. With the supports of these two tools, we create a software system. CUDA programs execute multiple instruction streams on different cores in GPU to realize parallel computing of input signals which includes measuring peak value, FIR filtering and FFT processing. After the data parallel in GPU, Open GL is used for displaying the 2D and 3D result image on the screen with no need to copy the data back to the internal memory. Algorithms are optimized in order to fit different data processing requirement and specific hardware in the computer. Results The test is based on Geforce GT430 GPU and Intel i5-2800@2.8GHz CPU. We use 256 channels wave data stored in local memory to simulate the data uploaded by real inspection equipment. Each channel of data has 512 16-bits sample points with the pulse repetition frequency of 512 Hz. That means the input data rate is 128MB/s. In the experiment, the software detects the peak value, arrival time of the signal and calculates the energy, then realizes FFT algorithm to get frequency spectrum. After the data processing, the software receives a succession of data and fuses the data to form 2D and 3D images, then outputs the result images in time. Comparisons between software using GPU and CPU are made. The results show that the new software system which is based on CUDA and Open GL can accelerate the data processing procedure at the speedup ratio of 8 to 40. The maximum throughput rate is 256MB/s. Discussion and Conclusions With the cheap and widely used GPU like GT430, we speed up data processing and image displaying considerably, especially 3D image real-time displaying. As the GPU Computing architecture like CUDA is becoming more and more mature and the GPU hardware is developing better and better, it has great potential in the implementation of software on Ultrasonic Phased Array Inspection and Imaging Application.
Background, Motivation and Objective Ultrasonic phased array technology has been widely studied and used in the field of Non-Destructive Testing. With the development of ultrasonic phased array technology, the sampling frequency gets higher and the number of independent channels increased at the same time. As a result, huge amounts of data are collected by the data acquisition module. For example, the size of data uploaded by PCI(Peripheral Component Interconnect) devices can be as much as 133 MB per second. It is hard for the CPU to finish the data processing procedure and display the result in time. However, the data collected by ultrasonic phased array system is divided into different channels inherently; the data of each channel can be processed through same algorithms with limit dependence on the data of other channels. In recent years, GPU parallel computing is developing rapidly. The GPU parallel computing is ideal for solving this kind of data-parallel computing problem. Statement of Contribution/Methods CUDA is a GPU parallel computing architecture provided by the NVIDA Corporation which is supported by most of its GPU products. With the help of CUDA, we can implement and optimize the parallel algorithm of the phased array signal processing. Open GL(Open Graphics Library) is the premier environment for developing 2D and 3D graphics applications. With the supports of these two tools, we create a software system. CUDA programs execute multiple instruction streams on different cores in GPU to realize parallel computing of input signals which includes measuring peak value, FIR filtering and FFT processing. After the data parallel in GPU, Open GL is used for displaying the 2D and 3D result image on the screen with no need to copy the data back to the internal memory. Algorithms are optimized in order to fit different data processing requirement and specific hardware in the computer. Results The test is based on Geforce GT430 GPU and Intel i5-2800@2.8GHz CPU. We use 256 channels wave data stored in local memory to simulate the data uploaded by real inspection equipment. Each channel of data has 512 16-bits sample points with the pulse repetition frequency of 512 Hz. That means the input data rate is 128MB/s. In the experiment, the software detects the peak value, arrival time of the signal and calculates the energy, then realizes FFT algorithm to get frequency spectrum. After the data processing, the software receives a succession of data and fuses the data to form 2D and 3D images, then outputs the result images in time. Comparisons between software using GPU and CPU are made. The results show that the new software system which is based on CUDA and Open GL can accelerate the data processing procedure at the speedup ratio of 8 to 40. The maximum throughput rate is 256MB/s. Discussion and Conclusions With the cheap and widely used GPU like GT430, we speed up data processing and image displaying considerably, especially 3D image real-time displaying. As the GPU Computing architecture like CUDA is becoming more and more mature and the GPU hardware is developing better and better, it has great potential in the implementation of software on Ultrasonic Phased Array Inspection and Imaging Application.
引文