Model Based Optimisation of Ultrasonic Corrosion Measurement

Abstract

The detection and characterisation of corrosion in pipework is a significant challenge across a wide range of industries. A common inspection technique is the use of monolithic ultrasound probes to perform pulse echo measurements to characterise the remaining wall thickness. Although this is one of the most conceptually straightforward ultrasonic measurements, there are complicated interacting effects such as surface roughness, probe size, operating frequency and signal analysis method. The use of numerical modelling rather than experimental trials can accelerate the optimisation and qualification of the inspection. This paper presents results of applying an efficient sampling and interpolation algorithm that minimises the number of model evaluations required to build an accurate surrogate model of the inspection outcome. The application of sensitivity analysis to this model demonstrates that the inspection is insensitive to the dimensions of the probe and is highly sensitive to changes in the wall thickness and changes in the frequency properties of the probe, demonstrating the suitability of the inspection.