**4. Data processing for SHM**

#### **4.1 Effect of environmental and operating conditions**

Several investigations into the effect of environmental and operational conditions on the recorded ultrasonic signals have been carried out, and change of temperature has been shown to be the main source of signal fluctuations [56–58]. The influence of temperature on GWT is a combination of effects due to the structure's mechanical properties and the effects on ultrasonic transducers and their bonding. Previous study has reported that, for small ambient temperature variations of a few degrees, the effect on transducer performance is much less significant than that on the wave propagation [59]. The UGW signals will undergo changes in the amplitude and phase. The change in UGW signal amplitude is attributable to changes in temperature-dependent properties of the ultrasonic transducer, particularly the piezoelectric materials and adhesives. To minimise this variability, careful selection of adhesives and transducer material for target temperature is recommended. The phase shift in the UGW signal is due to the change in wave propagation velocity due to variation in the mechanical properties of the waveguide [60], i.e. pipe or tank floor in this study. The material properties of relevance include elastic and shear moduli, and the density; which in turn relates to the elasto-acoustic properties of the material, acoustic absorption and ultrasonic wave velocity. Thermal expansion adds to this effect by changing the propagation distance directly and indirectly through changes in the thickness of the plate or the pipe. The relationship between the difference in time of arrival (TOA) of the signal and the change in temperature of the structure can be described as:

$$
\delta t = \frac{d}{\mathcal{V}} (a - \mathcal{Y}) \delta \tag{1}
$$

**67**

**Figure 11.**

*after baseline subtraction [71].*

*Monitoring of Critical Metallic Assets in Oil and Gas Industry Using Ultrasonic Guided Waves*

of selection criteria including mean square deviation [56] and maximum residual amplitude [62] have been proposed. This method has limitations for cases when a large set of baselines is not available and if the temperature of the selected baseline is different from the temperature of the test signal. Baseline signal stretch (BSS) was introduced as a complimentary technique that in its simplest form requires only one single baseline at a reference temperature. In BSS, time domain stretching is performed to adjust the selected baseline and the local coherence is estimated as a function of time. BSS can be performed in both time and frequency domain to achieve similar performance [63]. A number of researchers have explored these methods to provide enhanced temperature compensation with a reduced number of baseline data sets [62–65]. The temperature resolution of the baseline set is defined by the capability of BSS method as the stretching required for large temperature difference leads to distortion of the signal's frequency content. The performance of BSS depends on signal complexity and mode purity and. For practical application, a temperature step of 1–2°C is recommended for baseline dataset [63]. Recently developed modified-BSS (MBBS) method outperformed BSS and is more effective for temperature differences of up to 13°C [66]. BSS can be computation intensive and alternative methods with improved computational speed have been proposed that operate on signals in the stretch factor and scale-transform domain [67]. Physics-based analytical techniques for temperature compensation [68, 69] utilise underlying physical principles such as changes in material properties and thermal expansion (described in Section 4.2) for transducer signal reconstruction at different temperatures. The advantage of these techniques is that it does not require a large set of baseline sensor measurements from the structure. The performance of these analytical temperature compensation models is shown to be at par with the state-of-the-art data driven techniques. They are however limited to simple structural geometries and boundary conditions. Combinations of analytical and data-driven strategies that require fewer baselines are being explored [70] which will offer an efficient, practical and useful approach for temperature compensation.

A method for damage detection must be applied to the corrected data to see whether the structure being monitored has developed any damage. In structures containing high densities of structural elements, the time-traces obtained are often too complex to be directly interpreted due to a large number of overlapping reflections. A popular approach for SHM is *baseline subtraction*, which is based on the comparison of structure's ultrasonic response at original state (baseline) with response at a later stage. The subtracted *residual signal* will remove reflections from pipe or tank floor features and isolate any damage scattered signals as illustrated in **Figure 11**.

*Baseline subtraction of UGW time traces (a) undamaged structure (b) damaged structure (c) defect signal* 

*DOI: http://dx.doi.org/10.5772/intechopen.83366*

**4.3 Damage detection strategies**

where *t* is the difference in TOA of the signals when the change in temperature of the structure is *T*. *d* is the distance travelled by the wave and *v* is the wave velocity. *α* is the coefficient of thermal expansion and *γ* is the coefficient of change in phase velocity. *γ* is generally much greater than *α* and hence from Eq. (1), it can be seen that the main contribution to change of TOA is from change in wave velocity due to temperature variations. Also, since the time shift is directly proportional to the propagation distance, it can be noted that the effect of temperature on UGW will increase with propagation distance. This can be significant for the large propagation distances required for pipes. The inverse relationship between temperature and wave velocity suggests that faster modes will be less affected than slower ones. These temperature induced variations in UGW signals can adversely reduce the defect detection capabilities of the SHM system. An experimental study [61] showed that the effect of temperature variation on UGW from ambient temperature up to 70°C was much more pronounced than the effect of a drilled hole of 1 mm diameter.

#### **4.2 Temperature compensation algorithms**

The issues described in Section 4.1 led to several investigations within the SHM research community and a number of EOC compensation strategies have been proposed. Their main objective is to achieve UGW propagation time and amplitude correction for enhanced defect sensitivity. These correction strategies can be classified into two techniques: data-driven and analytical physics-based.

The data-driven techniques requires a large set of baseline measurements from the structure at different temperatures. A signal from the 'bank' of baselines is then selected to minimise the difference relative to the test signal for a particular temperature. This method is called Optimum Baseline Selection (OBS). A number

#### *Monitoring of Critical Metallic Assets in Oil and Gas Industry Using Ultrasonic Guided Waves DOI: http://dx.doi.org/10.5772/intechopen.83366*

of selection criteria including mean square deviation [56] and maximum residual amplitude [62] have been proposed. This method has limitations for cases when a large set of baselines is not available and if the temperature of the selected baseline is different from the temperature of the test signal. Baseline signal stretch (BSS) was introduced as a complimentary technique that in its simplest form requires only one single baseline at a reference temperature. In BSS, time domain stretching is performed to adjust the selected baseline and the local coherence is estimated as a function of time. BSS can be performed in both time and frequency domain to achieve similar performance [63]. A number of researchers have explored these methods to provide enhanced temperature compensation with a reduced number of baseline data sets [62–65]. The temperature resolution of the baseline set is defined by the capability of BSS method as the stretching required for large temperature difference leads to distortion of the signal's frequency content. The performance of BSS depends on signal complexity and mode purity and. For practical application, a temperature step of 1–2°C is recommended for baseline dataset [63]. Recently developed modified-BSS (MBBS) method outperformed BSS and is more effective for temperature differences of up to 13°C [66]. BSS can be computation intensive and alternative methods with improved computational speed have been proposed that operate on signals in the stretch factor and scale-transform domain [67].

Physics-based analytical techniques for temperature compensation [68, 69] utilise underlying physical principles such as changes in material properties and thermal expansion (described in Section 4.2) for transducer signal reconstruction at different temperatures. The advantage of these techniques is that it does not require a large set of baseline sensor measurements from the structure. The performance of these analytical temperature compensation models is shown to be at par with the state-of-the-art data driven techniques. They are however limited to simple structural geometries and boundary conditions. Combinations of analytical and data-driven strategies that require fewer baselines are being explored [70] which will offer an efficient, practical and useful approach for temperature compensation.

#### **4.3 Damage detection strategies**

A method for damage detection must be applied to the corrected data to see whether the structure being monitored has developed any damage. In structures containing high densities of structural elements, the time-traces obtained are often too complex to be directly interpreted due to a large number of overlapping reflections. A popular approach for SHM is *baseline subtraction*, which is based on the comparison of structure's ultrasonic response at original state (baseline) with response at a later stage. The subtracted *residual signal* will remove reflections from pipe or tank floor features and isolate any damage scattered signals as illustrated in **Figure 11**.

#### **Figure 11.**

*Baseline subtraction of UGW time traces (a) undamaged structure (b) damaged structure (c) defect signal after baseline subtraction [71].*

*Advances in Structural Health Monitoring*

**4.1 Effect of environmental and operating conditions**

Several investigations into the effect of environmental and operational conditions on the recorded ultrasonic signals have been carried out, and change of temperature has been shown to be the main source of signal fluctuations [56–58]. The influence of temperature on GWT is a combination of effects due to the structure's mechanical properties and the effects on ultrasonic transducers and their bonding. Previous study has reported that, for small ambient temperature variations of a few degrees, the effect on transducer performance is much less significant than that on the wave propagation [59]. The UGW signals will undergo changes in the amplitude and phase. The change in UGW signal amplitude is attributable to changes in temperature-dependent properties of the ultrasonic transducer, particularly the piezoelectric materials and adhesives. To minimise this variability, careful selection of adhesives and transducer material for target temperature is recommended. The phase shift in the UGW signal is due to the change in wave propagation velocity due to variation in the mechanical properties of the waveguide [60], i.e. pipe or tank floor in this study. The material properties of relevance include elastic and shear moduli, and the density; which in turn relates to the elasto-acoustic properties of the material, acoustic absorption and ultrasonic wave velocity. Thermal expansion adds to this effect by changing the propagation distance directly and indirectly through changes in the thickness of the plate or the pipe. The relationship between the difference in time of arrival (TOA) of

the signal and the change in temperature of the structure can be described as:

*d*

where *t* is the difference in TOA of the signals when the change in temperature of the structure is *T*. *d* is the distance travelled by the wave and *v* is the wave velocity. *α* is the coefficient of thermal expansion and *γ* is the coefficient of change in phase velocity. *γ* is generally much greater than *α* and hence from Eq. (1), it can be seen that the main contribution to change of TOA is from change in wave velocity due to temperature variations. Also, since the time shift is directly proportional to the propagation distance, it can be noted that the effect of temperature on UGW will increase with propagation distance. This can be significant for the large propagation distances required for pipes. The inverse relationship between temperature and wave velocity suggests that faster modes will be less affected than slower ones. These temperature induced variations in UGW signals can adversely reduce the defect detection capabilities of the SHM system. An experimental study [61] showed that the effect of temperature variation on UGW from ambient temperature up to 70°C was much more pronounced than the effect of a drilled hole of 1 mm diameter.

The issues described in Section 4.1 led to several investigations within the SHM

The data-driven techniques requires a large set of baseline measurements from the structure at different temperatures. A signal from the 'bank' of baselines is then selected to minimise the difference relative to the test signal for a particular temperature. This method is called Optimum Baseline Selection (OBS). A number

research community and a number of EOC compensation strategies have been proposed. Their main objective is to achieve UGW propagation time and amplitude correction for enhanced defect sensitivity. These correction strategies can be classi-

fied into two techniques: data-driven and analytical physics-based.

*<sup>v</sup>*(*α* − *γ*)*δ* (1)

*t* = \_\_

**4.2 Temperature compensation algorithms**

**4. Data processing for SHM**

**66**

**Figure 12.**

*Transmit-receive matrices for the imaging algorithms; (a) common source method (b) synthetic aperture focusing technique and (c) total focusing method [72].*

For sensor arrays Full-Matrix Capture (FMC) is a data acquisition process which records all possible transmit-receive combinations of UGW data. This data collection matrix is symmetric due to reciprocity (**Figure 12**) and only the lower and upper triangular parts of the matrix need be recorded. This data can then be used to obtain tomography images of the structure or perform sound energy focusing techniques to improve SNR.

For complex structures and if the data corresponding to the damage state is not known *a priori*, damage detection strategies based on unsupervised algorithms are used. One such strategy is based on the Outlier Analysis (OA) algorithm which extracts damage sensitive features from the UGW signals and aims to identify if they have deviated from their baseline distribution using Mahalanobis squared distance [73]. OA can be applied as univariate and multivariate depending on a number of features. For univariate implementation, root mean square (RMS) of the signal has been successfully used as a damage sensitive feature for detection of corrosion type defects in plates [56] and pipes [74]. To increase the damage sensitivity, multivariate OA is recommended, where a number of features are extracted from the UGW signals and classical methods of multivariate statistics such as principal component analysis (PCA) are applied. For UGW, the features of interest include time-of-flight, frequency centres, energies, modes of scattered waves, and time-frequency spread. A review of the feature extraction approaches based on time-frequency representations such as short-time Fourier transform, Wigner-Ville distribution, Hilbert-Huang transform, and wavelet transform can be found in [75]. Recent advances in the field of artificial intelligence led to researchers formulating defect detection as a machine learning problem. A study using an Artificial Neural Network (ANN) based strategy was applied for damage classification [73] and was reported to outperform OA for damage detection using just one feature. Such supervised machine learning strategies will however require data from the structure with known types and levels of damage, which may not always be present.

#### **5. Conclusions**

This chapter presents the advances in guided wave technology for structural health monitoring of two of the most critical metallic assets, pipelines and storage tanks, in the Oil and Gas industry. These SHM technologies support cost-effective asset integrity management by enabling a condition based maintenance model, moving away from conventional routine inspection. The advances in SHM technologies of pipes and tanks are presented. Operational requirements of these SHM systems

**69**

**Author details**

Anurag Dhutti1

2 TWI Ltd, Cambridge, UK

provided the original work is properly cited.

, Shehan Lowe1

1 Brunel University London, Uxbridge, Middlesex, UK

\*Address all correspondence to: tat-hean.gan@brunel.ac.uk

© 2019 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium,

and Tat-Hean Gan1,2\*

*Monitoring of Critical Metallic Assets in Oil and Gas Industry Using Ultrasonic Guided Waves*

are discussed with a thorough review of the state-of-the-art and fundamentals of pipelines and tank floor inspection using UGW. Limitations of SHM for high temperature pipelines have also been identified for future research and development.

*DOI: http://dx.doi.org/10.5772/intechopen.83366*

*Monitoring of Critical Metallic Assets in Oil and Gas Industry Using Ultrasonic Guided Waves DOI: http://dx.doi.org/10.5772/intechopen.83366*

are discussed with a thorough review of the state-of-the-art and fundamentals of pipelines and tank floor inspection using UGW. Limitations of SHM for high temperature pipelines have also been identified for future research and development.
