Wavelet analysis: application to the magneto-inductive testing

Size: px

Start display at page:

Download "Wavelet analysis: application to the magneto-inductive testing"

Diana Murphy
5 years ago
Views:

1 11th European Conference on Non-Destructive Testing (ECNDT 214), October 6-1, 214, Prague, Czech Republic Wavelet analysis: application to the magneto-inductive testing More Info at Open Access Database A. CANOVA 1, E. FURNO 2, A. BUCO 2, M. RESSIA 2, D. ROSSI 2, B. VUSINI 2 1 Politecnico di Torino, Corso Duca degli Abruzzi 24, 1129 Torino Tel Fax , aldo.canova@polito.it 2 AMC Instruments, C.so Re Umberto 42, 1128 Torino Tel fax , info@aemmeci.com Abstract The filtering of signals detected in a non-destructive testing can be effective in order to automatically delete the noise and extract a clear signal that contains the most significant information. Regarding filtering technique, the Fourier transform is one of the most known and used techniques although often, in the presence of non-stationary signals or with transient behavior, are more effective techniques such as wavelet transformation. In the present work the wavelet analysis is applied to the magneto- inductive control and in particular to signal called LF ( Localised Fault) which by its nature contains information that are well suited to be highlighted with this technique. The LF signal shows to the operator a value of electric voltage generated by a magnetic field sensor: coils or Hall effect sensors. When an external or internal fault happens, eg. broken wire in the case of wire ropes or lack of material for corrosion in the pipes, an instantaneous variation of the magnetic field with respect to a background value occurs. Wavelet analysis allows highlighting these discontinuities of the signal even when the background noise tends to hide them. After validating the efficacy and safety offered by wavelet filtering it is also possible to adopt this technique during the certification of a system for controlling magneto- inductive. The conventional test for the certification of magneto-inductive device is the EN :24 Safety Requirements For Cableway Installations Designed To Carry Persons - Ropes - Part 8 : Magnetic Rope Testing. In the paper the main results that have been obtained by applying the wavelet transform are presented. The presented examples regards different control systems primarily in the field of magneto- inductive wire ropes but also in the control of steel pipelines. Keywords: Wavelet, magneto inductive testing, Localised fault, metallic ropes 1. Introduction In the contest of magneto-inductive testing (MIT) the signal called Localised Fault (LF) is the most important indicator for fault detection. LF signal in the metallic rope controls show the presence of broken wires constituting the rope s strands [1-2]. Such signal is influenced by several factors including: Location of fault, inside or outside the rope Size of fault: e.g. more than one broke wire in the same location Length of the fault, that is the distance between the two ends of broken wires Because of the natural background noise produced by the vibration associated with the stray magnetic fields, generated by the detector or from the magnetic anisotropy of the rope, the LF signal due to a broken wire is sometimes drowned in the noise band. The non-detection occurs especially when the fault is internal or when the ends of broken wires are very close.

Figure 1: Inside and outside fault, small gap fault It s also possible, through an in-depth field experience, detect other phenomena such as corrosion and abrasion, but in this paper only main

2 Figure 1: Inside and outside fault, small gap fault It s also possible, through an in-depth field experience, detect other phenomena such as corrosion and abrasion, but in this paper only main function of the LF signal is analyzed: to provide a reliable indication of broken wires. The improvement of the signal/noise ratio (SNR), which is the main performance required during the MIT, can be reached by acting on both the detector either on the recording system. The detector is composed by two parts: the magnetizer and the sensor box, and there are many possible configurations depending form the size and shape of permanent magnets. Regarding the recording system in the past were all analogic but in the last years have been replaced by digital apparatuses. It is obvious that a good design of detector is the most important step to reach a satisfactory SNR ratio but this aspect is out of the issue of this work. Using digital recording system it is possible to acquire the signal with high number of samples and postpone the chose of the filter at the end of the acquisition step. In this paper we have focused mainly on this second point evaluating how the Wavelet Analysis [3] can improve the quality of the LF signal cutting informative contents that are not associated with the phenomena of breakage of wires. Regard to the nature of the control taken into consideration, a resolution of about 1 mm is sufficient. On the basis of this specification and on the rated speed of the rope inside of the detector is defined the sampling frequency of the system that usually does not exceed a few khz. The noise on the signal is mainly due to several factors: interference conducted by the electronic components of the acquisition system; radiated noise due to external electromagnetic sources; mechanical noise tied to the vibrations of the sensor or to fluctuations in the detector; magnetic noise due to the anisotropy of the rope. These noise components are characterized by very different frequencies and therefore can be reduced by using various techniques. Regarding the electromagnetic noise (conducted or radiated), they may be limited by acting on both the shielding of the sensor and by filtering the signal with a high-pass filter (for example, to remove interference due to the mains frequency) or low-pass filter (being careful not to filter the signals due to faults). Regarding the filtering of the other components which contribute to the noise the main problem is related to their non-stationary: in case of signals characterized by a frequency content not placed in time, the analysis of signal characterizes by the use of the Fourier series is notoriously not effective and is instead promising the use of Wavelet [3]. In the next paragraph will be described how to use the technique of wavelet to improve the quality of the LF signal. 2. Wavelet and LF signal

As known from the literature the application of the wavelet transform is used to determine a set of coefficients that allow

These coefficients associated with the low frequencies are known as approximation and those associated with high frequencies

To comply with the sampling theorem and don t make aliasing errors it s obviously necessary to limit the number of

2 shows an example of this subdivision diagram. Fig. 2.

3 As known from the literature the application of the wavelet transform is used to determine a set of coefficients that allow to reconstruct the source signal. These coefficients associated with the low frequencies are known as approximation and those associated with high frequencies are known as detail. To comply with the sampling theorem and don t make aliasing errors it s obviously necessary to limit the number of coefficients of high and low frequency to no more than half of the source samples. Fig. 2 shows an example of this subdivision diagram. Fig. 2. Example of decomposition of the signal using wavelet The decomposition process can be repeated several times (there are different criteria for defining a maximum number of decompositions) according to the diagram shown on Figure 3. Fig. 3 Tree of wavelet decomposition According to this scheme the LF signal is processed during the post-processing and the information of detailed elements can provide the detection of broken wires. Due to the

similarity between the shape of the selected wavelet and the LF signal, the latter is highlighted through the elements of detail that have higher values of the coefficients.

4 similarity between the shape of the selected wavelet and the LF signal, the latter is highlighted through the elements of detail that have higher values of the coefficients. For example Figure 4 shows the application of the wavelet to a signal obtained during the control of a wire rope. In this example the detail charts D8 and D9 provide information useful for the identification of defects. 3. Results Fig. 4. Approximation charts and detail of the LF signal on wire rope A first example of wavelet relates to a case of noise reduction. Fig 5 shows the trend of the LF signal that detects the presence of a break in the ring of a metal chain. From the source signal appears the breaking of the chain ring but there is a significant noise due to the change of magnetic reluctance brought by the rings of the chain. Fig. 5 shows the filtered signal with the wavelet in which noise reduction is evident. A second relevant and effective example of clear identification of defects regard signals from magneto-inductive open type instruments. Such devices are characterized by the presence of a low frequency noise associated to the flux fluctuation inside the rope. Fig. 6 shows for example a system in which there is the need to use an open type instrumentation. It s a system of overhead cable transport by conveyor belt used in mining, in large construction sites and in any case where it s required a medium-high hourly material flow. The conveyor belt, in the application presented, assumes a shell shape to better contain the material and slides on a roller fixed to a frame itself supported by ropes. The high number of closed type hooking of the frames to the ropes requires the use of open type detectors.

5 Fig. 5. LF signal unfiltered and filtered using wavelet technique In this instruments the LF control is limited to a sector which does not cover more than 15. It should be noted how passing the test as proposed below, and similar to those for approval of open type instruments is more complex than closed type instruments but reachable if there is a good filtration of the LF signal. Shown below are the results obtained on a rope test in which the faults (absence of material) are inside according to the picture shown in Fig 7. The two faults have a cylindrical shape and have a diameter and a length equal to about a tenth the diameter of the rope and are distanced by a length equal to the diameter of the rope. The ropes of the test have respectively a diameter of 3 mm for the detector called FB3 and 6 mm for the detector called FB6. Fig. 8 shows the waveforms of the LF signal during the test relative to the detector FB3, in the absence (blue curve) and in the presence of Wavelet filtering (red curve). The filtering parameters used in the previous analysis are reported in Table 1.

6 Fig. 6. Conveyor belt and open type instrumentation Tab. 1. Filtering parameters used during the testing of detector FB3 on 3 mm diameter rope Detector Wavelet Level Reconstruction FB 3 Bior3.5 9 D6+D7+D8

7 Fig. 7. Diagram of the fault used during the test.2 Original Obtained by Wavelet.15 Signal (V) x 1 4 Fig. 8. LF filtered and unfiltered signal in the test detector FB3 on 3 mm diameter rope x Fig. 9 shows the waveforms of the LF signal during the test relative to the detector FB6, in absence (blue curve) and in the presence of filtering Wavelet (red curve). The filtering parameters used are reported in Table 2. Finally it s reported the application of wavelet to LF signal detected by a closed type instrument called ROPE35 (for 3 mm ropes). Fig 1 shows two tracks, the original filtered using FFT and filtered through wavelet, which shows how the wavelet technique in this case can t achieve a significantly better result compared to a standard filtering technique. The original signal has a relatively low noise characterized by a frequency content of the same order of magnitude of the signal due to the fault. To further filter this signal should reduce the number of coefficients that reconstruct the waveform. This choice allows to obtain signal/noise ratios very high but it s sometimes risky and unreliable. All results submitted use the maximum number of coefficients calculated in function of the number of samples acquired. Tab. 2. Filtering parameters used during the testing of detector FB6 on 6 mm rope Detector Wavelet Level Reconstruction FB 6 Bior3.5 9 D6+D7+D8 Tab. 3. Filtering parameters used during the testing of detector ROPE35 on 3 mm rope Detector Wavelet Level Reconstruction Rope 35 Bior3.5 9 D7+D8

8 .3.25 Original Obtained by Wavelet.2 Signal (V) x 1 4 Fig. 9. LF filtered and unfiltered signal in the test detector FB6 on 6 mm rope x.8 FFT Obtained by Wavelet.6 Signal (V) Fig. 1. LF signal filtered with wavelet and FFT (ROPE35 on 3mm rope) 4. Conclusions In the paper we present some results obtained by applying the wavelet transform to LF (Localised Fault) type signal that are based on the measurement of the leakage flux obtained by magneto-inductive instruments on wire ropes. x

9 Due to the similarity between some of the wavelet and the classic shape of leakage flux due to breakage of strands it s possible to improve the signal to noise ratio by increasing the sensitivity and reducing noise. The technique were particularly efficient in those cases where the phenomena of magnetic noise at low frequencies due to causes of mechanical nature, such as the oscillation of the instrument relative to the rope, hide the LF signal associated with anomalies. In particular it was shown how for open type instrumentation this technique can significantly improve the identification of faults. Future evolutions of this work are the identification of the different types of fault through the cretion of appropriate wavelet functions and the possibility of introducing the wavelet transform in real time during acquisition and in post-processing. References [1] A.Canova, B. Vusini, Magnetic Analysis of Non Destructive Testing Detectors for Ferromagnetic Ropes COMPEL; Vol. 27, No 4, 28, pagg [2] Weischedel H.R., The Inspection of Wire Ropes in Service: a Critical Review, Material Evaluation ASNT, vol. 43, No. 13, pp , 1985 [3] Christopher Heil, David F. Walnut, Fundamental Papers in Wavelet Theory, Princeton University Press

Analysis of metallic ropes magnetisation during magneto-inductive testing

11th European Conference on Non-Destructive Testing (ECNDT 2014), October 6-10, 2014, Prague, Czech Republic Analysis of metallic ropes magnetisation during magneto-inductive testing More Info at Open