Friction Welding Ultrasonic Detection Of Wavelet Fractal Analysis

In this thesis, the major experimental object is ultrasonic signal detected from friction welding joints which is made from GH4169, a sort of common high temperature alloy which is used in aero engines. A method based on wavelet and fractal is put forward to denoise and identify the defects.Aiming at the characteristic of friction welding pseudo defect signals, such as low signal noise ratio (SNR), and being detected difficultly, this paper analyses characteristic of noise and proposes a wavelet denoising method to deal with the ultrasonic signals. The intention is to raise signal noise ratio and extract weak signals from the noisy signals. Firstly, on the basis of wavelet transform theory and denoising arithmetic, it was comparatively investigated to the difference of denoisingeffect of two denoising methods applied to typical signals—Heavysine and Bumps.These methods are wavelet soft thresholding denoising method (STD) and translation invariant denoising method (TID). The de-noising method based on translation invariant performs the cycle-spinning for the signal to be analyzed, and then one should shift the data of the de-noised signal in reverse. Do this for many times and average results so obtained. The simulation results show that this de-noising method can restrain jamming and greatly improve the denoising effect.To eliminate the noise hidden in detected signals, a denoising model is established and translation invariant denoising method is put to deal with the pretreatment signal. In the process of modifying the high frequency coefficients, the highest level low frequency coefficients is eliminated at the same time. The simulation results show that this de-noising method can restrain jamming, raise the signal noise ratio and greatly improve the de-noising effect. This work establishes the basement for following study.Theory of fractal is introduced to do quantitative and qualitative analysis of fractal welding ultrasonic testing signals. Firstly, wavelet transform is used to study the ultrasonic and test signals' self-similarity characteristic. The result shows that they are fractal and can be characterized quantitatively by fractal dimension. Then, non-scale region is determinated by gridding method and box dimension of various signals is calculated. In order to indicate the good statistical characteristic, the mean and the mean-root-square error of various signals are presented. At last, the relationship is founded between box dimension and signal classes so that the defects can be detected and classified exactly. Experiments indicate that the classified result is consonant with the testing exactly. This approach not only provides quantitativeguide lines but also decreases token parameters. It is simple, intuitionistic and useful in friction welding defects classes.