Ultrasonic Non-Destructive Evaluation
Ultrasonic non-destructive evaluation (NDE) consists of transmitting ultrasonic waves into the test object and evaluating the backscattered response. The ultrasonic waves going into the object are governed by the laws of wave propagation including reflection and refraction at surfaces between two different materials. The reflection and refraction effects produce the information about the inspected specimen in the received signal. The response may contain, for example, information regarding structure of the material, possible shape and location of defects within the object.
The use of elastic waves in the evaluation of materials is a complex problem. Different wave-types take different shapes and paths. The bulk waves propagate through the material while the surface waves are constrained to the boundaries of the object. Furthermore, the bulk waves have two modes, pressure (longitudinal) waves and shear (transversal) waves. For these waves both the particle displacements and the sound speed are different. Moreover, mode conversions will occur at boundaries giving rise to several wave types, even when the transducer generated only a single wave type.
The identification of defect responses in the ultrasonic signal is the main objective of ultrasonic evaluation. As the waves may both be reflected and refracted, the signal will – disregarding the noise – consist of peaks corresponding to the reflections at different inhomogeneous surfaces. These surfaces can either be irregularities, i.e., holes or simply be the bottom, sides or the top of the test object. The amplitude of the peak is partly connected to the size of the reflecting surface. A small hole or crack will reflect little and instead refract most of the pulse, while a distinct border between two different materials will reflect more of the pulse leading to higher amplitude in the signal response. Since the speed of waves in the test object is known the location of the peaks on the time axis simply becomes a measure of distance to the irregularity that gave rise to the peak in question. Figure 2.1 illustrates this principle.
Fig. 2.1 Measurement setup for ultrasonic testing of a specimen. The ultrasonic data consists of a set of signals backscattered at different boundaries within the test-object.
Ultrasonic Array System
The conversion of an electric pulse into elastic waves can be achieved using a variety of transducer types. Figure 2.2 shows the system used at Signals and Systems group at Uppsala University that is used for gathering ultrasonic data used throughout this report .
The most commonly used transducer1 is the piezoelectric transducer that usually requires some form of couplet connecting it to the object being tested. The couplant can be either water, as is the case in Figure 2.2 – called immersion technique – or a thin layer of gel – called contact technique (water can also be used in the contact method).
Most modern piezoelectric materials are ceramics. The thin disk is the most common shape of piezoelectric ceramics used for non-destructive testing applications. An electric field is applied to the disk causing a change in the thickness of the disk that is proportional to the intensity of the field that lay upon the disk. Depending on the polarity of the electric field, the disk can either contract or expand. The opposite of this energy conversion, i.e. converting mechanical pulses into an electrical signal can also be done using the piezoelectric transducer. This means that a single transducer can be used as both the source of the pulse and the receiver of the backscattered signal, this use of the transducer is called pulse-echo mode.
There are means for reducing the noise, for example, by refining the equipment used, or more precisely, the transducer used. The ultrasonic data gathered using one single transducer includes grain noise that depends on the path taken by the ultrasonic beam through the test object. If it is possible to focus the ultrasonic beam on a single point from several different positions and use the average of all those signals as output it will most certainly lead to better resolution in that point. Another way of collecting multiple data from one point is to use array systems. A linear array is a series of transducers connected in a line; this system is shown in Figure 2.3.
Fig. 2.3 Schematic view of a linear array. A series of transducers are connected to a controller to build an array.
The elements are controlled electronically to allow for beam focusing. The focusing is achieved by introducing different delays to different transducers in the array. The superposition of the pulses from the different transducers will then generate a wave with the direction controlled by the delays. Figure 2.3 presents a simple scheme of a linear array having 10 elements. Common array sizes are 16, 32 or 64 elements.
There are advantages in using an array system compared to a single transducer. Scanning an area with a single transducer requires the use of a mechanical device that controls the position of the transducer. Since ultrasonic inspection sometimes requires very fine resolutions it is necessary to keep two points of measurement as close to each other as possible. This leads to time being consumed simply moving the transducer to the next measurement spot. Using an array the scanning can be performed electronically2 leading to less time being consumed by data acquisition. That is why ultrasonic evaluation applications are today often performed using arrays instead of single transducers.
To further improve the SNR signal-processing methods can be applied to the acquired ultrasonic data. These methods include digital filters used to suppress the noise. The general idea of all signal-processing methods used for noise reduction is that the signal corresponding to a defect has a distinct pattern while the noise spectrum does not. The signal-processing methods will be discussed more thoroughly in Chapter 3.
Welds and the grain structure of the inspected material give rise to a more complicated problem than one may first expect. An important factor limiting the quality of tests is noise. Here – ultrasonic inspection – a primary source3 of this noise is called grain noise and is the result of the pulses backscattered at grain boundaries. Figure 2.4 presents the transmitted signal and the corresponding backscattered signal that has been distorted due to the structure of the material.4
Visualization of Ultrasonic Data
The acquired US-data are stored in vectors containing received signals from the different positions of the transducer. These A-scan vectors can be presented in different forms. The basic approach is plotting the amplitudes as a function of either time or distance. An example of such presentation is shown in Figure 2.5.
The peaks in the A-scan in Figure 2.5 indicate the front echo, the back echo or echoes coming from defects within the test object. The information available in an A-scan corresponds to a single measurement. In order to enhance the visual effects, several A-scans taken from different position are put together in one plot to form so-called B-scan. The B-scan represents a two-dimensional surface within the test-object. Then, each horizontal line in the B-scan represents an A-scan with the peaks in the A-scans coded as light or dark areas, depending on the imaging method. An example of a B-scan taken from the CAN1 test block is shown in Figure 2.6 (the CAN1 test block is described in detail in Appendix A). The echo within the marked areas – black rings – in Figure 2.6 originate from side-drilled holes within the test block.
Each B-scan covers a vertical slice of the test object. Several B-scan can be taken over the test specimen to cover the entire test object. This results in a three-dimensional US-data covering the entire test-object.
There is a third way of presenting ultrasonic data where the output is the maximum amplitude within a specific time gate. This method is called a C-scan. The C-scan is taken over a set of B-scans within the time gate (or depth). These points are put together to form an image representing flaws at different depths. The image in Figure 2.7 shows an example of a C-scan taken from the CAN1 test-block.
To facilitate the understanding of the relations between A-, B-, and C-scans Figure 2.8 presents the data volume and the different areas representing the different scan-types. The C-scan extraction operates on a data volume as shown in Figure 2.8.
Introduction to Signal Processing methods in US-Toolbox
Grain Noise Reduction [7, 8, 9]
This toolbox contains three different methods for grain noise reduction, these are split spectrum processing (SSP) Minimaization, SSP Consecutive Polarity Coincidence and finally Common Component Rejection (CCR). The common idea among these methods is assuming different models for the target echo and noise and under those assumptions applying methods to discriminate the noise. Based on the assumption that there are a large number of scatters at random positions within a resolution cell, i.e. the smallest volume that can be resolved, the material noise within an A-scan can be regarded as an interference pattern with maximum and minimum values. The location and amplitude of these extremities depend on the phase differences and intensities of the interfering waves. These quantities depend on the position and the impedances of the reflectors giving rise to the noise pattern. The conclusion is that changing the position of the scatters relative to the inspection beam or changing the size of the scatters relative to the wavelength, i.e. changing inspection frequency, will lead to a change in noise pattern. Two methods of altering the noise pattern are known as the spatial diversity method (SDM) and the frequency diversity method (FDM). The SDM method applies an averaging on signals representing the same point but taken from different positions, while the FDM method applies an averaging in the frequency domain instead. Assuming that defects have a larger cross-section than a single material grain, the FDM and SDM methods can be used to separate between the noise pattern and the target echo pattern.
The SDM is applied when collecting ultrasonic data. Although the FDM could be applied during data collection as well a common way is to use post processing on a single ultrasonic signal (i.e. one A-scan). The idea is to run the signal through a filter bank to simulate the collection of a set of signals using different inspection frequencies. This approach is adopted by the two SSP methods implemented here. The statistical processors available in the toolbox are described in detail in the following sections. The CCR method utilizes a similar assumption as the SSP method, i.e. that noise and target echo spectrums differ. However it makes no assumptions on the spectrum of the target echoes directly but instead estimates the noise spectrum and suppresses the frequencies corresponding to the estimated spectrum. The CCR method is further described in section 2.8.
Table of contents :
1.2 Purpose and goals
1.3 Tasks and scope
1.4 Outline of Thesis
2.1 Ultrasonic Non-Destructive Evaluation
2.2 Ultrasonic Array System
2.3 Grain Noise
2.4 Visualization of Ultrasonic Data
2.5 Introduction to Signal Processing methods in US-Toolbox
2.5.1 Grain Noise Reduction [7, 8, 9]
2.6 Split Spectrum Processing
2.6.1 The Extraction Process
2.7 Common Component Rejection
2.8 The Wiener Filter
3 ULTRASONIC IMAGE AND SIGNAL TOOLBOX
3.2 Graphical user Interface
3.3 Load and Save functions
3.4 Processing Methods
3.5 Image presentation formats
3.6 C-scan presentation
3.7 Polynomial fit- and Adaptive Extraction
4.2 Common Component Rejection
4.3 SSP Minimization
4.4 SSP Consecutive Polarity Coincidence
4.5 Wiener deconvolution
5 CONCLUSION AND FURTHER WORK