针对航空涡轮盘难检区域缺陷的超声相控阵检测技术，分析了检测信号及图像特征，提出了检测信号的小波包软阈值降噪方法，并对小波阈值进行改进，最后基于降噪后的信号进行了相控阵图像重构。结果表明：缺陷特征信号几乎分布于检测信号的整个频域范围，单纯的频域滤波必然导致缺陷信号的损失；基于改进的小波包阈值，并且有针对性地选择部分分解信号进行滤波降噪，可在较好保留有用信号的同时滤除噪声信号，有效地提高了检测信号信噪比及检测图像分辨率，对于提高涡轮盘的超声相控阵检测能力具有重要作用。

Ultrasonic phased array technology for testing deep, small and bad location defect is studied, and the characteristic signals and imagery related to defects are analyzed.In order to enhance signal to noise ratio and imagery resolution, the wavelet packet soft threshold denoising method with improved threshold is proposed. The results indicate that defect's signals are widely distributed in whole frequency domain, so that simple filtering in frequency domain on testing signal will lead to losses of defect's component inevitably. Wavelet packet denosing technique can be used to remove noise on the premise of keeping defect's component, through optimizing denosing threshold and choosing decomposed signals for composing, and wavelet denoise method proposed in the research is liable to enhance the signal and imagery quality of ultrasonic phased array testing on aero turbine disc.