Font Size: a A A

EMI Signal Drocessing Via Compressive Sensing

Posted on:2014-02-08Degree:MasterType:Thesis
Country:ChinaCandidate:H B TuFull Text:PDF
GTID:2248330395473825Subject:Structural engineering
Abstract/Summary:PDF Full Text Request
Electromechanical impedance (EMI) method was widely used in the field of civil engineering and aerospace, with the obtained signal very sensitive to the early local damage in the structure. Taking account of the difficulties encountered in data transmission and storage associated with the structural health monitoring (SHM), the novel technique called compressive sensing (CS) was employed in the process of data compression for the EMI signal in order to achieve efficient data transmission and storage.In order to make use of CS, we randomly observed the signal by Gauss random matrix and the sparsity level of the conductance signal is determined by Matching Pursuit. Thus, both the sparsity and the incoherence in the CS theory were meet. As an example, the one-dimensional damaged bar was considered to perform the EMI analysis. The root mean square deviation (RMSD) was extracted from the observation and compared with that from the original signal, arid the effects in data compression and the noise resistance ability of CS were discussed. It was found that the transmission bandwidth and storage space were reduced greatly; when the SNR (signal-to-noise rate) is greater than certain limit, the original signal can be reconstructed stably from the CS results, although noise exists. We establish the CS theory for EMI signal mathematically, and hence the EMI-SHM and the CS theory were successfully combined.A new method was developed to estimate the sparsity level based on the discrete cosine transform instead of Matching Pursuit. Since it is difficult to deduce analytically a mathematical relation between the measurements and the sparsity level, an empirical formula is fitted from the numerical simulation with the algorithm of compressive sampling matching pursuit. A one-dimension Euler-Bernoulli beam model is established for the analysis purpose, and it is shown that a small amount of the observed value almost contains most of the damage informationObserving the EMI signals by Gaussian random matrix for various degrees of damage in the numerical model, and extracting the principal components as the inputs of the BP neural network, we found that structural damage could be identified successfully after the right amount of training. After using the CS, the signal processing time and the storage space were reduced.The experimental study with single-sensor and multi-sensor was carried out. The sparsity level of the EMI signal was estimated and the number of measurements was calculated according to the empirical formula. The result indicates that the actual EMI signal can also obtain an effective compression.The research was financially supported by High Technology Research and Devel opment Program of China (863Program,2012AA050903), National Natural Scienc e Foundation of China (Grant No.51178413、51008272and50838008) and Nationa1Basic Research Program of China (973Program, Grant No.2009CB623200).
Keywords/Search Tags:Electromechanical impedance method (EMI), compressive sensing, resistance to noise, discrete cosine transform, compressed sampling matching pursuitalgorithm, BP neural network
PDF Full Text Request
Related items