旋压圆筒孔洞缺陷检测及特征信号盲提取研究

Research on hole defect detection and characteristic signalblind extraction of spinning cylinder

在固体火箭发动机金属壳体旋压成型过程中,针对旋压纹理所造成的检测回波信号混叠问题,提出了基于变分模态分解-快速独立成分分析(VMD-Fast ICA)和频域相关系数的旋压圆筒缺陷超声信号分离与提取方法。首先,利用频谱分析方法确定混叠信号的源信号个数,在此基础上采用VMD算法对实验信号进行分解和重构得到观测信号;其次,利用Fast ICA算法对观测信号进行盲源分离,通过计算独立分量与参考信号频域之间的相关系数,来判断各独立分量频域线性聚集性强弱,实现混叠信号的最优分解;最后,采用范数识别缺陷特征信号,利用小波阈值对其进行降噪后成像。实验结果表明,该方法能有效分离和提取出混叠信号中的缺陷回波信号,缺陷成像效果优于小波阈值方法和VMD-小波阈值方法,可实现对旋压圆筒孔洞缺陷的准确检测。

Aiming at the aliasing problem of detected echo signals caused by spinning texture during spinning process for solid rocket motor metal case,a method for separating and extracting ultrasonic signals of spinning cylinder defects based on Variational Mode Decomposition-Fast Independent Component Analysis(VMD-Fast ICA)and frequency domain correlation coefficient was proposed.Firstly,the number of source signals of aliasing signals was determined by spectral analysis method.On this basis,VMD algorithm was used to decompose and reconstruct the experimental signal to obtain the observation signal;secondly,blind source separation of observation signals was conducted by using Fast ICA algorithm.The correlation coefficient between each independent component and frequency domain of reference signal was calculated to judge linear aggregation intensity of each independent component in the frequency domain and realize the optimal decomposition of aliased signal;finally,the defect characteristic signal was identified by norms and denoised by wavelet threshold,then imaged.The experimental results show that the method can effectively separate and extract the defect echo signal in the aliasing signal,and the defect imaging effect is better than the wavelet threshold method and the VMD-wavelet threshold method,which can realize accurate detection for the hole defect of spinning cylinder.