AUTOMATED FLAW DETECTION SCHEME FOR CAST AUSTENITIC STAINLESS STEEL WELD SPECIMENS USING HILBERT-HUANG TRANSFORM OF ULTRASONIC PHASED ARRAY DATA

The objective of this work is to develop processing algorithms to detect and localize flaws using ultrasonic phased-array data. Data was collected on cast austenitic stainless steel (CASS) weld specimens on-loan from the U. S. nuclear power industry's Pressurized Water Reactor Owners Group (PWROG) traveling specimen set. Each specimen consists of a centrifugally cast stainless steel (CCSS) pipe section welded to a statically cast (SCSS) or wrought (WRSS) section. The paper presents a novel automated flaw detection and localization scheme using low frequency ultrasonic phased array inspection signals from the weld and heat affected zone of the base materials. The major steps of the overall scheme are preprocessing and region of interest (ROI) detection followed by the Hilbert-Huang transform (HHT) of A-scans in the detected ROIs. HHT offers time-frequency-energy distribution for each ROI. The accumulation of energy in a particular frequency band is used as a classification feature for the particular ROI.