Core Research – examples from 2014-18 programme
Incomplete array imaging (Imperial College)
Development of algorithms which will reduce the number of projections required for accurate image reconstruction for both X‐ray CT and ultrasonics while still enabling images of sufficient quality to be produced. Outputs include:
- Development of a fast hybrid Inverse‐Fourier algorithm for limited-view CT, which maximises the advantages of both inverse and Fourier based imaging methods.
- A regularisation scheme based on an existing binarisation technique.
- A study comparing the new algorithm with existing techniques, which showed better imaging results and significant speed improvements compared to other iterative algorithms.
Left (top)– standard reconstructions from limited data. Left (below) – the new FNSR algorithm developed on the project. Right – run times of typical algorithms, demonstrating how much faster FNSR is.
- Jones GA, Huthwaite P, NDT & E International, Vol 93, pp 98-109, 2018
Automatic methods for crack detection (University of Manchester)
The project aims to use automatic methods of image analysis developed in other fields to help identify cracks or abnormalities in manufactured components. The long term goal is increase the automation of imaging and interpretation, to improve reliability and reduce manpower.
- Understanding of machine learning techniques to identify abnormalities on images of components treated with dye penetrant.
- Improved methodology for modelling the normal appearance of components in a radiograph, enabling improved identification of abnormalities.
- Improved Random Forest based methods for detecting abnormalities in welds
- Novel Convolutional Neural Network approach for learning features useful for identifying abnormalities (which showed good performance when then combined with the Random Forest approach).
- Development of a procedure to automatically identify a particular location (weld region) for subsequent inspection;
- Dong X, Taylor CJ, Cootes TF, Proc International Conference on Pattern Recognition, 2018.
- Dong X, Taylor CJ, Cootes TF, Proc 2nd International Workshop on Compact and Efficient Feature Representation and Learning in Computer Vision, 2018.
Magnetic camera (University of Manchester)
Delivery of novel automation and objectivity to eddy current inspection (ET) and Magnetic Flux Leakage (MFL), by the development of a Magnetic Camera based on new Quantum Well Hall Effect (QWHE) sensors with enhanced sensitivity, frequency response and small physical size.
- QWHE sensor arrays were developed with sufficient sensitivity to detect small variations in induced transverse magnetic fields with increased lift-off distances.
- Prototype cameras were constructed and their capabilities demonstrated and quantified using applications provided by the industrial partners.
- A numerical simulation study of the interaction of AC magnetic fields with metallic objects, identifying the capabilities to measure crack length and depth.
- An experimental trial to compare the NDE performance of the QWHE sensors with current mainstream NDE techniques.
Colour QWHE image of Steel sample highlighting the indications of toe cracks, as well as weld boundaries, heat affected zones and other microstructures
- Liang, C-W., Balaban, E., Ahmad, E., Zhang, Z., Sexton, J., & Missous, M, Sensors and Actuators, A: Physical, Vol 265, pp 127-137, 2017
- Watson, J M., Liang, C-W., Sexton, J., Biruu, F., & Missous, M. 57th Annual British Conference on Non-Destructive Testing, Nottingham, United Kingdom, 2018.
THEME 2: CHARACTERISATION
Characterisation of defects (Universities of Bristol, Strathclyde and Iowa State)
Using database search techniques to extract quantitative characterisation information from NDT data, focusing on ultrasonic data and later exploring generalisation to other modalities.
- The approach is to form a large database from simulations and then, for measured data, search this database to find the closest match.
- A new way of looking at this problem, termed parametric-manifold mapping, has been developed which involves constructing a surface in principal component space that represents all possible defects of a given type. The characterisation challenge then becomes one of finding the closest approach of measured data to this surface.
- The shape of this surface reveals fundamental insights into the nature of the defect characterisation information, telling us for example, which defects are easy/hard to characterise.
- Crucially the method also enables probability maps of defect characterisation to be plotted.
- Velichko A, Bai L, Drinkwater BW, Proc. R. Soc. A Vol. 473, 2017.
- Bai L, Velichko A, Drinkwater BW, Journal of the Acoustical Society of America, Vol 143(1), pp 349–360, 2018.
- Safari, A., Zhang, J., Velichko, A., Drinkwater, BW, NDT & E International, Vol 94 pp 126–136, 2018.
Characterisation of the internal structure of scattering solids (University of Strathclyde)
This research was aimed at determining the interior microstructure of safety critical components using ultrasonic phased array data (without prior knowledge of the microstructure) and then using this information to improve defect detection and imaging.
- A Bayesian approach was developed for the reconstruction of the spatial material map of a heterogeneous material using Full Matrix Capture (FMC) data from an Ultrasonic Array; the method was developed to use the full waveform contained in the FMC data set.
- A semi-analytical model was developed to describe the wave propagation through the Voronoi tessellation.
- The objective function is based on the cross-correlation distance between the semi-analytical model FMC output and the experimentally measured FMC data.
- Case studies concluded that the method was able to reconstruct the material map to such a degree that the resulting image quality for a defect was commensurate with the image that was produced when using the known material map.
- Tant, K. M. M., Galetti, E., Mulholland, A. J., Curtis, A., & Gachagan, A. (2018). Inverse Problems, Vol 34(9), 2018
Future transducer technologies (University of Strathclyde)
This project is exploring the use of emerging piezoelectric materials to achieve improved transduction performance and to cope with challenging environments, and is also investigating more radical design concepts inspired by natural phenomena such as insect hearing. Achievements include:
- An ultrasonic array device using a Cantor Set (CS) fractal geometry has been explored using both finite element (FE) modelling and experimentation and has shown improvements in both image resolution and signal strength compared to conventional arrays.
- A further advancement of the fractal geometry has been proposed which comprises orthogonal CS fractal geometries, known as the Cantor Tartan (CT). This is predicted to provide further improvements in array performance.
- A 32-element, 5MHz linear array, incorporating PMN-PT 1-3 piezo-polymer composite, has been designed and optimised using FE and is now being fabricated for performance evaluation.
- A pipe organ-inspired backplate which can replace the conventional transducer backplate to increase the bandwidth of an air-coupled micro-machined transducer without sensitivity loss has been designed, built and evaluated.
- Walker AJ, Mulholland AJ, IMA Journal of Applied Mathematics, Vol 82 (6) pp1135–1150, 2017.
THEME 3: NEW TECHNIQUES
Multi-wave imaging for damage precursor detection (University of Bristol)
Development of new inspection approaches that exploit mixing between different wave modes (a pump and probe wave), specifically ultrasound with another wave e.g. ultrasound, thermal or magnetic. The programme has focused on nonlinear ultrasonics and provided significant advances in the field:
- Development of the nonlinear diffuse energy imaging technique which has enabled the detection and sizing of closed cracks with unprecedented accuracy while requiring the use of just a single ultrasonic array.
- A family of new nonlinear imaging techniques have been developed which utilises the more conventional coherent component of the field. While these techniques are not as sensitive as diffuse energy imaging, they are necessary for applications where a statistical diffuse state cannot be achieved.
- It has been shown how suppression of linear scattering features can be improved through compensation for instrumentation nonlinearity.
- Development of a nonlinear subharmonic array imaging technique, including an analysis of all possible mode pairs in application to fatigue crack imaging.
- Potter JN, Croxford AJ, IEEE Transactions on ultrasonics, ferroelectrics and frequency control, Vol 65 (5) pp 870-880, 2018
- Cheng J, Potter JN, Croxford AJ, Drinkwater BW, Smart Mater. Struct., Vol 26 055006, 2017
- Cheng J, Potter JN, Drinkwater BW, Smart Mater. Struct., Vol. 27 065002, 2018
Near infra-red techniques for NDE (University of Warwick)
Development of new approaches for inspection of materials and coatings using imaging and spectroscopy in the near-infrared and mid-infrared wavelength range. Outputs include:
- Demonstration of direct imaging in both the NIR and mid-IR wavelength bands as a function of source wavelength, together with demonstration of spectroscopy systems in both wavelength ranges.
- Imaging and monitoring of corrosion beneath paint layers
- Imaging the internal structure of non-metallic objects – e.g. honeycomb composites and thick glass fibre composite.
- Combination with ultrasound for the detection of corrosion under insulation (CUI)
- NIR spectroscopy for imaging of water ingress
- Tomographic imaging of polymer utility pipes
Near infrared Images of an epoxy/foam lightweight composite containing a delamination. This used a scanned laser diode/photodiode combination at a wavelength of 1.064 μm (see Senni et al., 2019).
- Laureti S, Sfarra S, Malekmohammadia H, Burrascano P, Hutchins DA, Senn L, Silipigni G, Maldague XPV, Ricci M, NDT & E International, Vol 98, pp 147-154, 2018.
- Silipigni G, Burrascano P, Hutchins DA, Laureti S, Petrucci R, Senni L, Torre L, Ricci M, NDT & E International Vol 87, pp 100-110, 2017.
- Senni L, Laureti S, Rizwan MK, Burrascano P, Hutchins DA, Davis LAJ, Ricci M, NDT & E International, Vol 102, pp 281-286, 2019.
NDE for additive manufacturing (University of Nottingham)
This project has been focused on selective laser melting (SLM), an AM process technology that has been extensively developed in the last decade and now available commercially, and was also focused on an NDE method (Spatially Resolved Acoustic Spectroscopy – SRAS) which has particular advantages for this application.
- A demonstration SRAS setup was developed and integrated into an SLM build chamber to prove the feasibility of using the technique in an industrial context..
- Models were developed to quantify the spatial measurement capability, temporal inspection ability, and the associated costs of different measurement approaches.
- Research into piecewise rework of surface defects based on SRAS inspections.
- Measuring contaminants using SRAS scans.
- Influencing and measuring the microstructure of builds using seed crystals.
- Investigating other AM technologies where SRAS could be used to measure build parameters.
(left) A micrograph of a typical Ti64 SLM printed surface (Ra ~6μm).
(right) A velocity map produced using SRAS to show the material texture contrast on the as-deposited rough surface.
- Hirsch M, Patel R, Li W, Guan G, Leach R K, Sharples S D, Clare A T, Additive Manufacturing Vol 13, pp 135-142, 2017.
- Patel R, Hirsch M, Dryburgh P, Pieris D, Achamfuo-Yeboah S, Smith R, Light R, Sharples S, Clare AT, Clark M, Appl. Sci. Vol 8(10), p 1991; 2018.
- Hirsch M, Dryburgh P, Catchpole-Smith S, Patel R, Parry L, Sharples S,. Ashcroft I, Clare AT, Additive Manufacturing, Vol 19, pp 127-133, 2018.
THEME 4: Monitoring
Magnetic monitoring of corrosion in pipelines (University of Warwick & Imperial College)
An investigation into magnetic monitoring of pipelines for corrosion, initially studying the metal magnetic memory (MMM) method about which major claims have been made but on which there is very little peer reviewed literature and moving on to the magnetic tomography method (MTM). Significant progress has been made on understanding these methods and the indication is that they are very unlikely to be reliable in practice.
- Tests were performed on 23 steels, and also with and without remnant (bias) field, the presence of magnetite and during fatigue.
- Small but measureable effects are seen at low transducer lift-off; no change greater than the earth’s magnetic field has been seen in any of the tests.
- No evidence has been found of cumulative signal growth under elastic or plastic cycling with any material tested unless the previous maximum load is exceeded.
- The effects with remnant (bias) field are exactly as would be expected with magnetic flux leakage.
- Tightly adhered magnetite oxide has a small effect and there was little change during fatigue cycling.
- At 1m from pipe range, a bolt 2m away can give the same perturbation as a typical corrosion defect through 50% of pipe wall, which probably explains the high false call rate in some trials.
- Li Z, Jarvis R, Nagy PB, Dixon S, Cawley P, NDT & E International, Vol 92, pp 59-66, 2017.
- Li Z, Dixon S, Cawley P, Jarvis R, Nagy PB, Cabeza S, NDT & E International, Vol 92, pp 136-148, 2017.
Ultrasonic monitoring of highly textured materials (Imperial College)
The aim was to develop a monitoring technique to compare successive measurements at the same location, so enabling the grain noise that makes standard ultrasonic inspection difficult to be subtracted out, leaving only signals due to damage growth.
- It has been shown that analytical models give a good approximation to the grain noise from large-grained materials seen with typical transducers at low frequencies.
- A methodology for the selection of transducers for permanently installed monitoring of large grained materials has been established.
- A methodology for establishing ROC (POD vs PFA) curves has now been tested thoroughly and used to establish the best processing method for guided wave monitoring data – this has been proved in a blind trial;
- Liu, Y., Liu, C., Van Pamel, A. et al. J Nondestruct Eval, 36: 53, 2017.
Data processing for NDE monitoring (Imperial College)
The aim of this project is to provide a road map for the complex monitoring systems of the future, with generic development of processing to compensate for system, environmental and operational variability so that high quality data can be provided to enable operator attention to be focussed on the areas where action is needed. Conclusions include:
- Permanently installed sensors provide real-time time domain data, allowing identification of trends.
- A blind trial on guided wave monitoring showed a 5x increase in sensitivity compared to one-off NDT – this opens the way to using lower frequencies with better volume coverage and lower noise in highly textured materials.
- Monitoring how the actual component responds to the actual operating conditions removes reliance on assumptions and therefore yields superior estimates.
- The uncertainty bands in remnant life estimates from rate-based monitoring in the example studied are about 8 times tighter than those obtained from periodic inspection.
- To take into account more complex scenarios such as varying loads, it is necessary to incorporate reliance on more assumptions.
- The work to date has looked at the Paris Law regime; further opportunities are expected earlier on in the fatigue life cycle.