Performance analysis of a wavelet packet transform applied to concrete ultrasonic detection signals

In the ultrasonic detection technology of concrete health conditions, it is difficult to identify damage features due to the complex wave synthesis of detection signals. We propose a method of applying wavelet packet transform (WPT) to decompose and reconstruct the ultrasonic detection signal from a C30 class concrete, thereby reducing noise and redundant information. We can obtain the main frequency node coefficients by wavelet packet decomposition, and reconstruct the denoised signal by the node coefficients. For analyzing the performance of this method, the extracted indices of signal-to-noise ratio, root mean square error, Pearson correlation coefficient, and smoothness are utilized and evaluated respectively. Ten simulated signals and ten actual detection signals are employed for evaluating the method. Simultaneously, the comparison with the calculation results of the empirical mode decomposition (EMD) method is created. It is shown from the experimental results that WPT has better denoising performance on the signals. Finally, an experiment of the processed signal quality assessment is constructed. The stochastic configuration network (SCN) is employed as the evaluation model for exploring the influence of the proposed denoising method on identification accuracy. The result shows that the proposed method of reconstructing the main frequency node coefficients by WPT retains useful information more effectively and improves the identification accuracy of the recognition model.