Knowledge-Based Ultrasonic Inspection System Development Near Completion

摘要

The KBIS, which is presently in the final stages of validation testing, will improve an inspector's ability to analyze ultrasonic weld inspection data and assess discontinuities compared to the normal methods that evaluate signal amplitude, discontinuity length and the proximity between discontinuities. Initial results clearly demonstrate the ability of a neural network to classify crack, incomplete fusion, slag and porosity discontinuities. When fully validated, KBIS will become commercially available for use by shipbuilders to classify and document weld discontinuity type and size during manufacturing, overhaul and repair of steel ship and submarine structures. With proper validation, KBIS also can be applied to the inspection of critical welds in chemical and power generation equipment, buildings, bridges and heavy machinery.