Analysis of Differential Pressure Fluctuation Signal Based on EMD and BP Neural Network

摘要

A new signal processing method based on empirical mode decomposition (EMD) and BP neural network was presented to analyze the differential pressure fluctuation signal of gas-liquid two-phase flow. A detail explanation of EMD and its algorithm was provided. Based on EMD, the Intrinsic Mode Functions (IMF) with different characteristic time scales were extracted from the signal respectively. In this work, the EMD method was used to decompose the nonlinear and non-stationary differential pressure fluctuation signal into a number of IMF components. And then, the local energy characteristic of each IMF component was obtained by using Teager energy operator. Meanwhile, these local energy characteristics were regarded as the inputs of BP neural network for training and flow pattern identification. The experimental results show that the proposed method based on EMD and BP neural network can reflect the local time-variant behaviors of signal and can implement the flow pattern identification successfully.