Three characteristics of the received signals from the FE models will be analysed: the amplitude of the first backwall signal (BW1), the amplitude of the pit signal (PS) and the extracted wall thickness (WT). BW1 and pit signal phase changes have not been considered.
6.3.1 Normalised Amplitude Definitions
Figure 6.11 a) Schematic of a spot measurement setup b) illustration of the windowing used in the method to determine the backwall and pit amplitudes
131 An overview of the method to determine the backwall and pit amplitudes is illustrated in Figure 6.11b, whereby the time trace is segmented into two windows covering the 1st
backwall echo (BW1) and the pit signal. The maximum amplitudes within either the BW1 or pit signal window is taken as the ABW1 amplitude or the APS pit signal amplitude respectively.
Both of the amplitude values are calculated (in dB) relative to ABW1,ND, i.e. the 1st backwall
echo without a defect present. The Hilbert envelope of the raw time traces, senv[n], are
used to calculate the amplitudes. The following outlines the methods which were used to define the BW and pit signal windows as well as how the BW1 amplitude or the pit signal amplitudes, ABW1 and APS, are calculated. One drawback of using the windowing method is
that it creates βdeadzonesβ, see appendix 9.6 for more details.
6.3.1.1 Determination of the First Backwall Amplitude
First, the peak location of the BW1 without a defect is determined using the known nominal thickness, b:
ππ΅π1,ππ·,ππππ= ππππππ₯
π (π πππ£,ππ·[(π πβ ) ππ π < π < (3π πβ ) ππ π ])
(6.1)
where senv,ND[n] is the enveloped signal with no defect present, cg is the group velocity, fs is
the sampling frequency and (π πβ ) ππ π is a threshold which occurs approximately at the
midpoint between the input signal and the BW1 echo. The start of the BW1 window can then be defined:
ππ΅π π€πππππ€ π π‘πππ‘ = ππ΅π1,ππ·,ππππβ πππππ’π‘ π‘πππππ’ππ π‘/2 (6.2)
where Ntoneburst is the sample width of the input toneburst and is equal to πππ¦ππ π πππ
β , Ncy are
the number of cycles and fcen is the centre frequency. Similarly, the end of the BW1 window
is determined:
ππ΅π π€πππππ€ πππ = ππ΅π1,ππ·,ππππ+ πππππ’π‘ π‘πππππ’ππ π‘/2 (6.3)
The limits of the BW1 window are then used to find the maximum value of senv,ND[n] and
senv[n] which occur within the BW1 window. The relative value of the BW amplitude is given
132 π΄π΅π1= 20 πππ10
πππ₯(π πππ£[ππ΅π1 π€πππππ€ π π‘πππ‘ < π < ππ΅π1 π€πππππ€ πππ])
πππ₯(π πππ£,ππ·[ππ΅π1 π€πππππ€ π π‘πππ‘< π < ππ΅π1 π€πππππ€ πππ])
(6.4)
6.3.1.2 Determination of the Pit Signal Amplitude
The window used to define the pit signal starts where the input signal ends and ends when the BW1 signal without a defect begins. First, the peak location of the input toneburst using the known nominal thickness, b, is determined.
πππ,ππππ= ππππππ₯
π (π πππ£,ππ·[π < (π πβ ) ππ π ])
(6.5)
Next the start of the pit signal window is located using the known width of the pulse:
πππ π€πππππ€ π π‘πππ‘= πππ,ππππ+ πππππ’π‘ π‘πππππ’ππ π‘/2 (6.6)
The end of the pit signal window is equal to the start of the BW1 window:
πππ π€πππππ€ πππ= ππ΅π π€πππππ€ π π‘πππ‘= ππ΅π1,ππ·,ππππβ πππππ’π‘ π‘πππππ’ππ π‘/2 (6.7)
Then using the limits of the pit signal window the maximum value of senv[n] which occurs
within the pit signal window is determined. The relative value of the pit signal amplitude is given in the dB scale, relative to the BW1 echo with no defect.
π΄ππ = 20 πππ10(
πππ₯(π πππ£[πππ π€πππππ€ π π‘πππ‘< π < πππ π€πππππ€ πππ])
πππ₯(π πππ£,ππ·[ππ΅π1 π€πππππ€ π π‘πππ‘< π < ππ΅π1 π€πππππ€ πππ]))
(6.8)
6.3.2 Time of Flight Method to Estimate Wall Thickness
In this study cross correlation is used to extract a value of wall thickness. An overview of the method is shown in Figure 6.12. First a threshold, nWT threshold, where the input toneburst
ends is defined, see Figure 6.12a.
πππ π‘βπππ βπππ = πππ,ππππ+ πππππ’π‘ π‘πππππ’ππ π‘/2 (6.9)
nin,peak is determined using equation 6.5. Two new signals are created, one of which
contains the signal before the threshold (input toneburst signal, sin[n], Figure 6.12b), and
another which contains the signal after the threshold (BW1 signal and a potential pit signal, sBW1[n], Figure 6.12c). The removed part of the signal is replaced with zeroes:
133 π ππ[π] = π [π > πππ π‘βπππ βπππ] = 0 (6.10)
π π΅π1[π] = π [π < πππ π‘βπππ βπππ] = 0 (6.11)
The two time traces are cross correlated with the input toneburst, stone[n]6, which is used
to generate the FE toneburst7:
πΆππ[π] = β π ππ[π]π π‘πππ[π + π] 2πβ1 π= 0 (6.12) πΆπ΅π1[π] = β π π΅π1[π]π π‘πππ[π + π] 2πβ1 π= 0 (6.13)
The arrival times of the input toneburst, tin, and the BW1 echo, tBW1, are found using the
argmax() function: πin,max= ππππππ₯ π (πΆππ[π]) (6.14) π‘ππ = πππ,πππ₯ ππΆ (6.15) ππ΅π1,πππ₯= ππππππ₯ π (πΆπ΅π1[π]) (6.16) π‘π΅π1= ππ΅ππΌ,πππ₯ ππΆ (6.17)
where fC is the sampling frequency of the cross correlations (Cin and CBW1). If the sampling
frequency of stone[n] is the same as s[n] then fc = fs. The method of calculating the wall
thickness will change depending on the transducer setup being modelled. If the same set of
6 To simplify the cross correlation s
tone[n] is resampled to have the same sampling frequency as s[n] 7 The input signal is effectively auto-correlated and therefore t
in = 0; however, there are certain setups where this is not a given (such as the waveguide setup) therefore the method has been described in full.
134 nodes generates and receives the signal (i.e. pulse-echo) then the wall thickness, b, can simply be calculated using:
π = ππ.
π‘π΅π1β π‘ππ
2
(6.18) When the transducer setup under investigation has a different set of Tx and Rx nodes (i.e. pitch catch) then the distance travelled by the pulse must take account of the separation, d, between the Tx and Rx node sets [143]:
π =1
2(βππ(π‘ππβ π‘π΅π1)(ππ(π‘ππβ π‘π΅π1) β 2π))
(6.19)
Furthermore the input signal, sin, will not necessarily be represented in the signal received
at the Rx nodes. In the case of the waveguide setup the surface wave between the Tx and Rx node sets will produce tin [143].
Figure 6.12. Illustration of method used to extract wall thickness, using a time trace a) which is divided into two: b) the input signal in the time trace is cross correlated with the input toneburst2 to find a first time value and c) the first backwall signal is cross correlated with the input toneburstto find the second time value.