超声检测与特征分析在半固态铝合金产品中的应用

Application of Ultrasonic Testing and Feature Analysis in Semi-Solid Processed Aluminum Alloy

采用接触法和水浸聚焦法分别对平底孔、通孔、槽伤三类半固态用铝合金人工缺陷进行超声检测并提取特征参量进行分类识别。用频谱分析方法分析静态超声A扫信号,对缺陷信号段进行快速傅里叶变换,以偏度值、峰度值、主频率峰值、谐振频率间隔作为特征参量判断缺陷类型。槽伤的偏度值、峰度值最大,通孔次之,平底孔最小,随着缺陷埋深的增加频谱图趋于对称和正态分布的特征;主频率峰值通孔最大,平底孔次之,槽伤最小;谐振频率间隔平底孔最大,槽伤次之,通孔最小。分析动态超声C扫描信号随着缺陷埋深增加的衰减情况,根据图像边缘形状特征可判断出缺陷类型,并且测量值/实际值(ρ/ρ0)图表明,平底孔缺陷尺寸测量误差最小达到40μm,通孔最小误差达到30μm,槽伤误差最小达到50μm。人工缺陷检测结果为自然缺陷的类型识别提供了可靠的判断依据。

Using contact detection for the ultrasonic A-scan inspection and focusing method of water immersion for the ultrasonic C- scan inspection, ultrasonic testing followed by feature extraction was adopted to identify the three forms of artificial defects, i.e. fiat bottom-hole, through hole and notch, inside a semi-solid processed aluminum alloy. Spectrum analysis on ultrasonic signal of A-scan was performed applying fast Fourier transform （FFT） to the windowed signal indicating the defects, and then the defects were categorized according to parameters such as skewness, kurtosis, peak value in frequency domain and intervals between resonant frequencies. The skewness and kurtosis of notch had the maximal peak value, followed by through hole and the minimum for flat bottom hole; peak value in frequency domain of through hole was the maximal, followed by flat bottom hole and the minimum for notch ; the intervals between resonant frequencies of the flat bottom hole was the maximal, followed by notch and the minimum for through hole. The forms of defection could be identified according to the features of C-scan image. Also, by analysis of decay of C-scan signal with respect to the increasing burial depth of defects, the bias from the measured defect size to the actual defect size could be estimated. The results showed that the minimum measuring error between measured and actual values of flat bottom hole was 40 μm, while that of through hole and notch was 30 and 50 μm, respectively. The test results of artificial defects could provide diagnosis basis with detecting natural defects for both A-scan detecting and C-scan imagery.