Manual UT vs PIMS (Permanently installed monitoring sensors)

Transcription

1 Manual UT vs PIMS (Permanently installed monitoring sensors) F. B. Cegla Non-Destructive Evaluation Group, Department of Mechanical Engineering Imperial College London, SW7 AZ,UK

2 Outline /36 Motivation/Background Corrosion example + Surface Roughness The effect of roughness on scattering Simulation method (DPSM) Results PIMS/ C-Scan Conclusions Future work

3 Actual Thickness (normalised) Motivation/Background 3/36 Corrosion costs several billion $/annum Inspection very important to avoid failures Main tool Manual UT Inspection After: van Roodselar et al., 9 Inspectors Summit 9, Galveston Texas Source: Measured Thickness (normalised)

4 Motivation/Background 4/36 Human factors are potential source of large spread Mechanized scanning Inspection (C-Scan) Permanently installed sensors (PIMS) Source: Source: What is the likely size of measurement errors? What is the likely influence of roughness on the ultrasonic signal?

5 Amplitude Ultrasonic thickness measurement principle 5/36 T 1 t t p p 1 Wall thickness (T) Transmitter.5.1 Pitch (p) Receiver Wave speed (v) t 1 Surface skimming wave.15. Not to scale t =? t Backwall.5 reflection.3 Time

6 What is the effect of roughness on UT 6/36 Ingredients considered: Length scales that have an effect on UT signal Typical surface roughness in the field Transducer geometry Fast simulation technique for statistics Signal processing techniques

7 Rough surface definition 7/36 Gaussian distributed: Uses normally distributed random numbers to generate different surfaces with similar statistics. Correlation length (λ ) RMS height (σ) Correlation length (λ ) RMS height (σ)

8 Scale of roughness that affects UT signal 8/36 Rayleigh parameter: transition to high surface RMS k cos 4 k = wavenumber σ = RMS height θ = incident and reflected angle RMS value ~.mm in steel for S waves (~.5 MHz) or P waves (~5 MHz) But what about horizontal extent? What horizontal scales must be present for scattering to influence the signal? J. A. Ogilvy. Theory of Wave Scattering from Random Rough Surfaces. IOP publishing Ltd. 1991

9 y-axis (mm) y-axis (mm) amplitude (arb) y-axis (mm) y-axis (mm) amplitude (arb) Scale of roughness that influences UT signal 9/36 1 λs=.5mm (.3λ) 1 λs=mm (1.λ) x-axis (mm) 1 λs=8mm (5λ) x-axis (mm) 1 λs=5mm (31λ) x-axis (mm) x-axis (mm) -1 -

10 y-axis (mm) amplitude (arb) Scale of roughness that influences UT Signal 1/36 D Problem, Tx, Rx (.6 width) MHz, 5 cycles, = 1.6mm: sinusoidal surface wavelength =.mm backwall transmitter receiver flat backwall sinusoidal backwall hilbert envelope x-axis (mm) time ( s) NOTE: trough in surface always occurs directly between transmitter and receiver

11 y-axis (mm) amplitude (arb) Scale of roughness that influences UT Signal 11/36 D Problem, Tx, Rx (.6 width) MHz, 5 cycles, = 1.6mm: sinusoidal surface wavelength =.4mm backwall transmitter receiver flat backwall sinusoidal backwall hilbert envelope x-axis (mm) time ( s) NOTE: trough in surface always occurs directly between transmitter and receiver

12 y-axis (mm) amplitude (arb) Scale of roughness that influences UT Signal 1/36 D Problem, Tx, Rx (.6 width) MHz, 5 cycles, = 1.6mm: sinusoidal surface wavelength = 4mm backwall transmitter receiver flat backwall sinusoidal backwall hilbert envelope x-axis (mm) time ( s) NOTE: trough in surface always occurs directly between transmitter and receiver

13 y-axis (mm) amplitude (arb) Scale of roughness that influences UT Signal 13/36 D Problem, Tx, Rx (.6 width) MHz, 5 cycles, = 1.6mm: sinusoidal surface wavelength = 4mm backwall transmitter receiver flat backwall sinusoidal backwall hilbert envelope x-axis (mm) time ( s) NOTE: trough in surface always occurs directly between transmitter and receiver

14 amplitude change (db) Scale of roughness that influences UT signal 14/36 Look at max. amplitude of backwall reflection as surface wavelength increases compared to max. amplitude of flat backwall reflection: sinusoidal surface wavelength ( s / )

15 Scale of roughness that influences UT signal 15/36 The simulations show: For rough surfaces with horizontal length scales (L) roughness distorts the signal if: o vertical RMS height > λ o horizontal FFT of surface contains.8 λ < L < 5-1 λ

16 Sulphidation corrosion example 16/36 o RMS height =.1-.5mm o FFT of surface contains scales of L between 1-1mm Picture from: Taylor-Hobson

17 Typical Transducer parameters 17/36 Parameter C-SCAN PIMS Transducer Area 6mm diameter 1x1mm rectangular Operational frequency Operational wavelength (steel) Beam footprint (at 1mm depth) 5 MHz MHz ~1 mm ~1.5mm ~6mm diameter ~8-1mm Corrosion RMS height =.1-.5mm ~.1-.3λ Corroded surface FFT scales = 1-1mm or.8λ < L < 1λ

18 Operation in region where roughness influences signal Statistical Simulations 18/36

19 DPSM: basic principle 19/36 Fundamentally based on Huygens principle Propagating wave front can be discretised into contributions from many point sources. Field at a single target point is then the summation of contributions from all point sources Propagating wave front Point Sources D: P m m () m r A H k r Target point n n f n Free Space Greens function 3D: P m m An m r exp ik r n r m n f n Placko, D. and Kundu, T. DPSM for Modeling Engineering Problems. (7)

20 DPSM: Matrix formulation /36 Equations cast into set of linear equations All contributions calculated in a single step P T N source points A 1 A A 3 r 1 m r m r 3 m r N m P 1 P P m P M Q TS A S Q TS exp ikf r 1 r exp f r1 exp f 1 M r1 1 1 exp ik r f 1 1 N ik r exp ik r exp ik r 1 f r M M M ik r exp ik r exp ik r r f M r 1 exp ik r r f 1 f N r f M N r 1 N N N M target points A N P Q T TS A S

21 1/36

22 amplitude (arb) y-axis (mm) D simulation case study different models /36 Single acoustic wave transceiver point source (λ=1.6mm) σ = 3λ c /16 x-axis (mm) Backwall FEM DPSM Kirchoff time ( s) Method Nodes Time taken (s) FEM DPSM Kirchhoff Chosen for its very high speed and ability to simulate multiple scattering and shadowing

23 Permanently installed sensor simulations 3/36 Transducer field l f l f l f l f l f MHz SH wave.5x1mm contact Average surface z y x

24 Permanently installed sensor simulations 4/36 Transducer field MHz SH wave.5x1mm contact z y x Inner surface Footprint width

25 amplitude (arb) Extracting a thickness from the simulated signal 5/36 Wall Thickness Example Envelope peak algorithm is used to evaluate the range of wall thicknesses that would be measured from many surfaces with the same roughness/surface statistics. TOF Envelope Peak (EP) time ( s)

26 Effect of roughness on thickness measurement? 6/36 1 surface realisations at each RMS for correlation length.8mm Peak to peak timing algorithm 1: Jarvis, A.J.C. and Cegla, F.B., Application of the Distributed Point Source Method to Rough Surface Scattering and Ultrasonic Wall Thickness Measurement, JASA (1). : Jarvis, A.J.C and Cegla, F.B., (13) Scattering of SH Waves by Sinusoidal and Rough Surfaces in 3D: Comparison to the Scalar Wave Approximation, manuscript in peer review process 13

27 C-scan transducer field 7/36 6mm 1mm 8mm 8mm

28 Mean of thickness estimates (mm) C-scan transducer results, 1 surfaces 8/36

29 Sampling due to the footprint 9/36 Transducer only probes a small area of the surface Transducer How much of this variation is due to this sampling effect? 9.8mm 1mm

30 Standard dev. of thickness estimates (mm) C-scan transducer results, 1 surfaces 3/36

31 31/36

32 Conclusion 3/36 Vertical RMS >~1/1 incident wavelength UT signal can be distorted Horizontal CL~.8-1 incident wavelength UT signal can be distorted For simulations with RMS >1/1 UT measurement std > surface RMS spread due to interaction of signal processing with the scattered signal Awareness of this important: measurement spread and uncertainties not due to UT setup and equipment but due to structure property itself.

33 Actual Thickness (normalised) Conclusion 33/36 All UT measurements influenced by the physics: manual, automatic scanning and permanently installed After: van Roodselar et al., 9 Inspectors Summit 9, Galveston Texas Measured Thickness (normalised)

34 Future Work 34/36 Link more temporal and spatial information Link to underlying corrosion mechanisms (general vs pitting corrosion)

35 Acknowledgements 35/36 For contribution: Dr Andrew Jarvis Mr Attila Gajdacsi Mr Daniel Benstock Sponsors:

36 QUESTIONS 36/36 QUESTIONS?