Tuesday 18 September
Session 1C – Defect Characterisation
Simulations of complex geometry defect characterisation by using PEC sensor array
Ilham Mukriz and Gui Yun Tian
School of Electrical, Electronic and Computing Engineering, University of Newcastle upon Tyne.
E: i.m.zainal-abidin@ncl.ac.uk; E: g.y.tian@ncl.ac.uk
A finite element simulation for the characterisation of complex geometry defects using the pulsed eddy current (PEC) technique is presented in this paper. In this work, we investigate PEC inspection on complex geometry defects, ie angular slots, through the magnetic field distribution profile from the eddy current and defect interaction in the sample. The investigation is implemented via time-harmonic and time-stepping finite element method (FEM) modelled in 2D and 3D using FEMLAB. The characterisation of angular slots was made possible by magnetic field mapping from the utilisation of a magnetic sensor array in the simulation. The simulation results indicate the potential of detection and characterisation of angular defects using magnetic sensor arrays for scanning or static inspection.
Comparative quantification of acoustic and magnetic Barkhausen noise data with respect to various excitation waveforms
John W Wilson, Gui Yun Tian
School of Electrical, Electronic and Computer Engineering, Merz Court, University of Newcastle upon Tyne, Newcastle NE1 7RU, UK. E: john.wilson2@ncl.ac.uk or g.y.tian@ncl.ac.uk
In recent years, magnetic Barkhausen noise (MBN) measurement has gained acceptance as a commercial stress measurement technique, but difficulty is encountered in decoupling the effect of microstructure and stress, so for reliable results material specific calibration samples must be employed. The acoustic manifestation of MBN; magneto-acoustic emission (MAE), is also generated by domain wall motion, but is detected by piezoelectric sensors at any location on the material surface, has greater measurement depth and is sensitive to different material properties and stress ranges. The different sensitivities of MBN and MAE open up the opportunity to use simultaneously acquired signals in a complementary manner to provide decoupled information about material microstructure and stresses. This paper details recent findings in an ongoing comparative study of the two techniques and looks at the effect on MAE and MBN of applying different excitation waveforms at different frequencies, including pulsed excitation of varying pulse width. Different materials are studied, including steel plate, steel block and cast iron water pipe and the effect of sample geometry on MAE frequency spectrum investigated. The work concludes that the two techniques have complementary capabilities and that through judicious selection of excitation waveform and signal processing parameters, present MBN based systems can be significantly enhanced.
PCA-based feature selection for X-ray image defect classification
Ying Yin and Gui Yun Tian
School of Electrical, Electronic and Computer Engineering, University of Newcastle upon Tyne, NE1 7RU, UK. E: ying.yin1@ncl.ac.uk; E: g.y.tian@ncl.ac.uk
Automatic defect inspection based on digital X-ray image processing techniques has been investigated at Newcastle University. It is capable of identifying welding defects and therefore is a powerful tool to aid in Non-Destructive Testing (NDT). In this study, defect images undergo a feature extraction process, where information contained in each image is condensed into a set of twelve geometric features corresponding to the shape of the defect and six textural features based on the co-occurrence matrix. These two groups of features are implemented separately. A feature selection process based on principal component analysis (PCA) provides guidance on choosing the most representative features for defect classification. Subsequently, a feed-forward artificial neural network (ANN) using a standard back-propagation learning algorithm is presented as a classifier for defect identification. The classification results are compared against experienced interpreter's judgments and a detailed discussion of the experimental results is included. The conclusions confirm that PCA-based feature selection can be utilised as an effective tool for X-ray image defect classification.
Defect sizing and characterisation using post-processing of ultrasonic array data
Paul D Wilcox, Bruce W Drinkwater, Caroline Holmes and Jie Zhang
RCNDE, Department of Mechanical Engineering, University of Bristol, Bristol, BS8 1TR, UK
The use of arrays for industrial NDT is expanding rapidly, as is the proliferation of new imaging algorithms. This paper examines how these new array imaging approaches, particularly those based around post-processing the full matrix of time traces, can be used to obtain quantitative information about the nature of a defect. For example, the total focusing method that focused on every point within an image is known to provide the optimal resolution for a linear imaging technique. This should then be an excellent the basis for accurate defect sizing. Furthermore, the image can be scaled such that defects of equal size at any location give the same amplitude of response. If the defects are large it is shown to be possible to extract sizing and characterisation information directly. For smaller defects, of the order of a wavelength this is more challenging and alternative approaches must be considered. For example, an array can be used to examine the reflectivity of a defect from a range of angles, even from a single array location. Inspection of a greater range of angles can be obtained by scanning. This angular reflectivity information contains useful information about the angle, size and type of defect. This approach leads to the extraction of metrics that quantify the defect and hence lend themselves to automated defect sizing and characterisation. The sizing and characterisation potential is the demonstrated through a number of industrial application studies.
Ultrasonic characterisation of texture validation using EBSD
Stephen Essex, Dr Mark Potter, Dr Richard Dobedoe, Dr Steve Dixon
University of Warwick
Electron Backscatter Diffraction (EBSD) is a microscopic technique that provides detailed crystallographic orientation information for given samples. It enables the qualitative visualisation and quantitative calculation of important microstructural properties such as grain sizes and their aspect ratios, and the average grain orientations, commonly known as the Orientation Distribution Function (ODF).
In order to validate an ultrasonic technique to quantitatively describe texture we are correlating ultrasonic velocity measurements with those from EBSD. The velocity variations as a function of angle from the Rolling Direction (RD) in sheet samples are dependent on such qualities measurable from the EBSD technique.
We describe a modified technique which can predict the S0 Lamb wave velocity profile as a function of angle from the RD directly from Bunge-Euler angle EBSD data, and calculate the three independent Orientation Distribution Coefficients (ODCs) W400, W420 and W440 for cubic structures. The ODCs represent a quantitative description of the texture.
EBSD results, taken from through-thickness scans to eliminate the errors that arise with surface-to-bulk inhomogeneity, for rolled aluminium sheets are discussed and compared to the ODCs and velocity profile obtained using an ultrasonic EMAT-EMAT S0 Lamb wave velocity measurement system.
Investigation of pulse eddy current with pulse width variation
Yong Li, Gui Yun Tian
School of Electrical, Electronic Computer Engineering, University of Newcastle Upon Tyne, NE1 7RU, UK. T: +44 (0) 191 222 5639; E: yong.li@ncl.ac.uk
Pulse eddy current (PEC) inspection has been exhibiting many advantages over conventional eddy current techniques in term of wide frequency band and high inspection efficiency. In general, the pulse width is made as large as possible in order to implement signal analysis right after the response in interest is effectively acquired. Current PEC applies wide pulse width excitation e.g. >1ms to get transient and steady responses. It has found this excitation has limitation of spectrum band and measurement time. In our research, the pulse excitation with relatively short pulse width varying from sub-microsecond to millisecond is investigated via experimental and theoretical study. It has been found that the rising time of PEC responses to conductive samples are different when different pulse width applies. Based on the skin effect and broad-band spectrums of short pulse excitations, it is expected to find more information for quantitative evaluation of defects, microstructures and materials at different depths in comparison with current PEC systems.
Session 2C(1) – Surface Methods
Automated MPI testing quality checks and dye penetrant Tam panel tests
Thomas Vetterlein
ITW TIEDE GmbH, Germany
In the industrial application of the magnetic particle testing method (MT) as well as in the penetrant testing method (PT) reference blocks are used to check the quality of the magnetic test ink or the quality of the penetrant process. In both cases most of the times the operator of the NDT machinery is responsible for the interpretation of the test results of the process quality given by the reference block. The disadvantage of this procedure is easy to see:
- The interpretation result is subject to human error.
- Documenting the quality status of the system and the process independent of the operator is nearly impossible.
Vibro-enhanced fluorescent penetrant inspection (FPI)
Katy Milne1,2, David Wright1, Tony Dunhill1 and Peter Cawley2
1Rolls-Royce plc
2RCNDE, Imperial College
Fluorescent penetrant is widely used by aeroengine manufacturers to inspect for surface breaking flaws in critical components. Even the most sensitive penetrant systems may not find short tight cracks (<1 μm mouth opening). The potential of vibration as a means of encouraging penetrant to fill and bleed-out of cracks, by changing the crack closure or the surface energy of the liquid-solid interface, was investigated. For the technique to remain non-destructive, the threshold stress-intensity ΔKth for crack propagation must not be exceeded. Tests were carried out on rectangular cross-section bars containing tight thumbnail fatigue cracks 0.4 to 1.5 mm in length. To eliminate the subjectivity of a human inspector, an image-processing algorithm was used to automatically detect crack indications. The change in crack mouth opening with static load was measured using an optical method. To improve the penetrant process static cross-crack strains greater than 450 με were required, exceeding ΔKth; shorter cracks require higher strains. The possibility that an improvement could be obtained at lower dynamic strains was then investigated. The samples were vibrated after the penetrant process at around 17 kHz. A bending mode was excited, generating cross-crack strains up to 300 με; at these strain levels no improvement was obtained.
Session 2C(2) – New Technology
Prospects for high density system integration for ultrasonic array transducers
Richard Challis, Vladimir Ivchenko, Albert Phang and Cuthbert Chirenda
Applied Ultrasonic Laboratory, School of EEE, University of Nottingham, UK
This paper reports initial ideas which have resulted from a brief study of the prospects for integrating electronic systems on to ultrasonic array transducers. Currently, ultrasonic array transducers required dense cabling to connect them to an associated transmit/receive electronic platform. It would be advantageous to integrate the electronics on to the array head so that cable density can be minimised, and this would also allow an increased density of array elements.
The paper will consider how this might be achieved using a range of cmos technologies and sets out design constraints in relation to signal levels, noise performance, device dimensions and cost. It will consider both conventional pulse-echo working and also the use of coded sequences combined with correlation techniques. It will also be shown that, in reality, what is actually achievable will be limited by prohibitively high cost.
