异构双核构架下的脉冲涡流信号实时小波去噪

Real-time wavelet denoising for pulsed eddy current signal based on heterogeneous dual-core

针对工业现场的脉冲涡流(PEC)信号,设计了基于异构双核OMAP3530的实时嵌入式去噪的硬件与软件系统。分析了适用于实时去噪的小波基的选择,完成了基于DB4小波基的Penalty阈值法、B-M阈值法、默认阈值法的Matlab去噪仿真实验,并给出了均方根误差(RMSE)和信噪比(SNR)。结合OMAP3530硬件平台,讨论了基于DSPLINK的异构双核构架的数据共享方法,并提出一种基于固定步长滑动窗口的实时去噪算法。通过计算机仿真与实验结果表明:该系统具有信噪比高、实时性强、数据吞吐率大的优点,满足连铸现场自动检测要求。

In order to denoise the Pulsed Eddy Current （PEC） signal from industrial field, a real-time hardware and software system based on heterogeneous dual-core OMAP3530 was designed. First, how to select a wavelet basis was analyzed in the real-time process. The Matlab denoising model based on DB4 wavelet threshold method including Penalty method, B-M method and default method was established, and the Root of Mean Square En＇or （RMSE） and Signal-to-Noise Ratio （SNR） of the denoising model were given. With reference to the OMAP3530 hardware platform, the system software design of data sharing based on DSPLINK of heterogeneous dual-core was discussed. A real-time filtering algorithm based on fixing length step window was promoted to denoise the PEC signal. The simulation and experimental results indicate that the real-time denoising system has the features of high SNR, strong real-time and high throughput rate to satisfy the requirements of auto- testing on the casting production line.