Reliability of Detection of Small Defects in Noisy Weldments by Advanced Signal Processing and Pattern Recognition Techniques

Abstract

Load bearing capacity of any structure for a given dimension can be increased by improving the reliability of detection of smaller defects in weldments. This is highly desirable in aeronautical and space industries where the overall weight of the structure is a major constraint as it decides the pay load capacity.

Maraging steels are widely used in space industries for fabrication of rocket motor casings. Conventional ultrasonic pulse echo technique is applied for defect detection in weldments. However, inspection by ultrasonic testing of top midsection of maraging steel weldments, which is acoustically noisy, poses problems for reliable detection of defects of size 3 mm long x 1 mm deep or less (a design requirement). This is because of small amplitude difference of echo signal (2dB) between the noise and defect signals. Reliability of detection of such defects is improved by adopting advanced signal processing and pattern recognition techniques, which in turn increases the load bearing capacity of the structures.

A developmental work was undertaken for reliable detection of small defects in maraging steel weldments. For this purpose, a fatigue crack of 3 mm x 1 mm size was created in top midsection of the maraging steel weldments representing expected weld defects formed during fabrication. This paper discusses the application of advanced techniques like, autocorrelation, cross power spectral analysis, demodulation and cluster analysis for detection of the simulated fatigue crack. By adopting these techniques 95% reliability has been achieved for the detection of the fatigue crack. The approach has been assessed for shop floor adaptability. The approach would yield considerable enhancement in the pay lead of space vehicle thus resulting in enhanced capability and effective utilization.