Algorithm And Implementation Of Particle Filter Theory In Acoustic Emission Signals Analysis

Acoustic emission is an emerging dynamic nondestructive testing technique for diagnosing real-time damage in structure. The acoustic emission signal analysis methods majorly have two categories:the parameter analysis method and waveform analysis method, regardless of the methods used, noise removal is the most fundamental and critical of the work. Traditional time domain and frequency domain means are difficult to extract useful signal from the measured acoustic emission signals due to complexity and non-stationarity of acoustic emission signals, resulting in extremely constraint to acoustic emission signal analysis and processing. This dissertation came up with a real-time onboard acoustic emission signals processing method according to Rao-Blackwellised particle filter, which dealing with non-linear and non-Gaussian dynamic system state estimation, has no any restriction of system noise, enhancing signal to noise ratio and improving the speed of calculation, meanwhile, conducted profound and systematic study based on laboratory test data and existing research results.Firstly, seismic wave signal model was introduced in the state space model of acoustic emission signals, using Bayesian theory to deduce posterior probability density distribution of the unknown based on the latest observations, found particle filter have more estimated accuracy on dealing with non-linear and non-Gaussian signals through comparing advantage and disadvantage of Kalman Filter, Extended Kalman Filter, Unscented Kalman Filter and Particle Filter to process the synthetic signal;PCI-2 acoustic emission system was applied to collect three different kinds of acoustic emission signals of limestone, skarn and granite in SHPB testing system, enormously improved the speed of calculation by using Rao-Blackwellised particle filter to denoising acoustic emission signals with noise;provided a completed algorithm for Rao-Blackwellised particle filter, concluded that current state estimation just depended on a previous time point through real-time analysis of the measured acoustic emission signals on the language of Matlab, that is not matter to future data, verified that Rao-Blackwellised particle filter with real-time onboard processing function for dealing with acoustic emission signal;Finally, because of acoustic emission signals of limestone, skarn and granite possess characteristics of transient, complexity and non-stationarity, waveform of the signal is more clearer and the main frequency contents of the signal are not missing after implementing Rao-Blackwellised particle filter dealing with acoustic emission signals, then gain more internal features of the rock specimens through combining parameter analysis method and waveform analysis method to reverse derive physical characteristics of the rock specimens.The research work in this paper is on the forefront of science, using advanced mathematical calculation and signal processing methods for researching particle filter theory and its application in acoustic emission signals analysis, the work done in this paper has a high theoretical and application value, and provides a theoretical and technical support for the final solution to real-time noise removal for acoustic signals.