Super resolution array imaging for the inspection of embedded rough 2D and 3D defects
Abstract
Ultrasonic Non-Destructive Evaluation (NDE) plays a crucial role in the accurate detection and characterisation of defects. There is a constant drive within the nuclear industry to improve upon the characterisation capabilities of current techniques in order to improve safety and reduce costs. Particular emphasis has been placed on the ability to characterise very small defects which could result in extended component lifespan and help to reduce the frequency of in-service inspections.
Significant research has been focussed on ultrasonic array imaging techniques which utilise full-matrix capture (FMC) data sets, such as the Total Focussing Method (TFM), to characterise defects. Super Resolution (SR) algorithms, also known as sampling methods, are a new family of imaging algorithms, which also use FMC data. These algorithms have been shown to demonstrate the capability to resolve scatterers separated by less than the diffraction limit when deployed in NDE inspection scenarios.
The majority of work to date on SR algorithms has focussed on idealised cases with defects with simplified geometries. Realistic defects, however, can be far from simple. One of the major challenges facing the characterisation of defects is that of roughness. All real defects exhibit some degree of roughness. It has been shown that the extent of roughness can significantly impact the scattered field from a defect and hinder accurate sizing using conventional techniques. The work presented here aims to investigate the capabilities of a range of SR imaging algorithms in characterising rough defects, whilst comparing their performance to conventional TFM. Ultrasonic array inspection has been simulated using both full 2D Finite Element (FE) analysis and 3D hybrid-FE modelling for a range of defect sizes, orientations and extent of roughness. A Monte Carlo approach has been applied for the 2D cases to allow statistically significant conclusions to be drawn. The effect of introducing a third dimension to the roughness on the capabilities of the imaging algorithms was then investigated and the findings compared to the 2D results.
Significant research has been focussed on ultrasonic array imaging techniques which utilise full-matrix capture (FMC) data sets, such as the Total Focussing Method (TFM), to characterise defects. Super Resolution (SR) algorithms, also known as sampling methods, are a new family of imaging algorithms, which also use FMC data. These algorithms have been shown to demonstrate the capability to resolve scatterers separated by less than the diffraction limit when deployed in NDE inspection scenarios.
The majority of work to date on SR algorithms has focussed on idealised cases with defects with simplified geometries. Realistic defects, however, can be far from simple. One of the major challenges facing the characterisation of defects is that of roughness. All real defects exhibit some degree of roughness. It has been shown that the extent of roughness can significantly impact the scattered field from a defect and hinder accurate sizing using conventional techniques. The work presented here aims to investigate the capabilities of a range of SR imaging algorithms in characterising rough defects, whilst comparing their performance to conventional TFM. Ultrasonic array inspection has been simulated using both full 2D Finite Element (FE) analysis and 3D hybrid-FE modelling for a range of defect sizes, orientations and extent of roughness. A Monte Carlo approach has been applied for the 2D cases to allow statistically significant conclusions to be drawn. The effect of introducing a third dimension to the roughness on the capabilities of the imaging algorithms was then investigated and the findings compared to the 2D results.
