NDE & inspection
Digital image correlation for nuclear23 April 2012
In a pilot project, an optical monitoring technique that uses computer image processing to measure displacement and strain was applied to a US nuclear power plant containment vessel during a pressure test. Results agreed with data from strain gauges and other physical structures (such as a hidden cold joint location). By Paul Bruck, Thomas Esselman and Michael Fallin
Digital image correlation (DIC) is a non-contact, nondestructive method to measure displacements and strains. Because of this, it has good potential for applications in nuclear power plants to perform measurements on both active and passive components and structures. DIC may be particularly useful in nuclear power plant applications where stay times may be limited or the ability to access components may be restricted due to radiation fields or contaminated surfaces.
DIC is an optical method that employs pattern matching and image registration techniques for accurate two- and three-dimensional measurements of changes in the shape of an item being examined. This method can be used to measure shape, deformation, displacement, and strain. The technique has found wide application in engineering and manufacturing methods to measure changes and provide displacement and measurement insights for material and structure analysis, finite element analysis verification, and quality control.
Because of the non-contact nature and relative ease of implementation of DIC, it is gaining greater acceptability as a measuring technique. Driving the method’s usefulness and acceptance are advances in computer technology and computing power combined with similar advances in the resolution (that is, megapixel size) of digital cameras. As the cost of both continue to fall, the cost of investment in DIC measuring systems also decreases. As the cost of initially purchasing a system decreases, it is expected the DIC measuring systems will become more prevalent for industrial applications.
Early research and development of DIC was performed in the early 1980s with continued technological improvement in the ensuing years. Using computer-based processes it is possible to obtain two-dimensional (one camera) and three-dimensional (two camera) full-field information by recording deformation and motion of random speckle patterns on a component or structure’s surface before, after, and/or during deformation of the object under test. Figure 1 shows a schematic representation of this photogrammetric principle. The stereo cameras can accurately track both global and local changes caused by stresses in the x, y, and/or z directions.
The rule of thumb for DIC is to make the camera separation between one-third to one-half the working distance, which will make the triangulation angle between the two cameras between 15 and 25 degrees. For the tests described here, the working distance was approximately 1.5 meters (5 feet), and the camera separation was approximately 600 mm (2 feet). A convenient working distance is set through lens selection. The exact angle between the cameras is determined precisely during the calibration process.
The concept is straightforward. The surface of the specimen is sprayed or marked with a random dot pattern (see Figure 4) and targets are used to calibrate the system (Figure 2). Two digital cameras take high-resolution photographs of the specimen surface. A left-right stereo pair of images at a reference condition is acquired, and the specimen is photographed at regular intervals during the time that a load is being applied or deformation is occurring. The changes over time can then be studied in detail. Pairs of images taken before and after deformation can also be compared to measure the total displacements or strains that have occurred during the loading. Digital images of the specimen contain gray-scale intensity measurements at each pixel location on the digital photograph. The image is downloaded from the camera to computers using either Gigabit ethernet to a laptop for slower frame rates (<15-30 fps) and/or frame grabber devices in a small rack or tower-style computer for higher frame rates (60-1000 fps). Static measurements are made with low frame rates and dynamic measurements are made with faster frame rates. Static measurements provide the possibility of doing long-term condition monitoring as well as capturing relatively slow loading at a rate of 1 frame per second or less. DIC has been used at data acquisition rates of 1 million frames per second, or more, using digital high-speed cameras. It is usual to run the cameras 10-15 times faster than the highest frequency of interest for vibrations, or fast enough to acquire at least 50 images during a short transient.
Using target features and their location, a displacement field is generated. DIC tracks the position of unique, distinguishable physical points from a reference image and a deformed image to determine differential displacement at these points. To achieve this, a subset of pixels is identified on the dot pattern around points of interest on a reference image and their corresponding location determined on the deformed image. In simplest terms, the computer software has a pattern matcher, which locates the four corners of each pattern-matching box, called a facet, by locating it relative to paint dots in and around each box. The corners will still be located accurately even though the box itself will deform due to local strains. The pattern matcher also intrinsically corrects for perspective differences between the two cameras. For example, on a curved surface, one camera may see the facet as a square, while the other sees it as a rectangle or keystone (warped) shape. The digital images are recorded and processed using an image correlation algorithm. Post-processing takes a few seconds per frame on a laptop with a dual-core processor, and less than a second per frame on a laptop with two quad-core or two hexacore processors. The result is the x, y, and z coordinates of thousands of closely spaced points on the test object.
Well-optimized input parameters provide very accurate results. In a one meter square area, the in-plane displacement accuracy is 5 microns (0.2 mils, thousandths of an inch) and the out-of-plane accuracy is 15 microns (0.6 mils). With this accuracy, in-plane deformation of less than 0.025mm (0.001 inch) can be measured.
In-plane and out-of-plane displacements are determined at discrete points in the data grid by subtracting the initial coordinates from each time sample during the test. To determine the strains, virtual strain gauges are centred on each data point in the image processing software. Multiple strain gauges in different directions are defined for each data point, so that both directional strains and principal strains (maximum strain independent of direction) can be determined.
The raw data collected shows all movement relative to the cameras. This can include rigid body movement of the item being monitored or it can be tilting or translational motion of the cameras. The DIC software first quantifies rigid body movement (assuming the cameras are stationary) and then can automatically remove the rigid body movement frame by frame, leaving only the local deformations. This ability to remove rigid body motion is critical for condition monitoring, since the cameras cannot be accurately placed over repeated measurements. This allows the cameras to be taken down and then re-set for measurements of multiple locations with a single camera pair.
The Ginna test
One element of a research program on the long-term operation of light water reactors is associated with the ageing of concrete structures. In support of one of the aspects of this programme, consulting engineers Lucius Pitkin, Inc. provided services utilizing DIC as a means to augment normal containment inspections and testing. The objective of the augmented inspection was to supplement the industry’s understanding of long term operation of a concrete containment structure, with the goal to develop comprehensive containment inspection guidelines.
The R.E. Ginna (Ginna) plant, which is owned and operated by Constellation Energy Nuclear Group (CENG), was one of the plants selected for testing. The Ginna plant is a 2-loop pressurized light-water reactor designed by Westinghouse. The plant is one of the oldest operating reactors in the United States, with commercial operation commencing in 1970. The plant is currently licensed by the USNRC for 60 years of operation; 20 years beyond the 40 year original licensed life (period of extended operation). The Ginna concrete containment structure is a vertically post-tensioned design, and includes the containment dome, cylindrical walls, with vertical tendons connected to embedded rock anchors Figure 3). One safety-related design function of the containment is to provide the capability to withstand the pressures and temperatures of a design-basis accident that occurs within the containment.
At nuclear plants in the United States, containment interior and exterior inspections are performed in accordance with American Society of Mechanical Engineers (ASME) Section XI, Subsections IWE/IWL. A periodic simultaneous Integrated Leak Rate Test (ILRT) and Structural Integrity Test (SIT) are also performed. Ginna has committed to perform two SITs during the period of extended operation concurrent with the ILRTs. The SIT evaluates the integrity and functionality of all containment components. The first SIT was performed in 2011. The SIT test pressure for this containment was 59.8 psi (0.41 MPa), compared to standard atmospheric pressure of 14.7 psi (0.101 MPa).
During the SIT, DIC was utilized on selected containment concrete surfaces. The objective of the testing was to accurately and quantitatively measure the behaviour of the concrete while being pressurized, and to record that information. The tests can then be repeated in the future and the results compared to the recorded information. Changes in the performance of the concrete when loaded by the internal pressure of the SIT, like increased magnitudes of local strains or increased crack opening displacements, could indicate age-related degradation of the concrete. No changes in the behaviour of the concrete would provide confidence that age-related degradation was not occurring in the containment.
Three locations were selected on the containment structure to perform strain and shape monitoring using DIC just prior, during, and for a short time following the pressurization of containment during the SIT. The locations selected were near the equipment hatch (outdoors), near the personnel hatch, and at a location near the top of the cylindrical containment (see Figure 3). The locations near the two hatches were chosen because they are discontinuities in the structure, and so are likely to experience extra strain. Near the upper cylindrical containment location, fibre-optic strain gauges were mounted on the containment concrete and an exposed rebar to measure strain directly, and provide the ability to correlate with the DIC results.
The DIC tests were performed in cooperation with Trilion Quality Systems, Inc., provider of the Aramis optical inspection system, which included the cameras and associated processing hardware and software, all manufactured by GOM of Germany. Lucius Pitkin, Inc. provided overall project management, plant co-ordination, and interpretation of the results, under contract by the Electric Power Research Institute (for its Long Term Operations Programme) and the US Department of Energy (as part of the Light Water Reactor Sustainability programme).
One Aramis camera system was mounted at the personnel hatch to monitor behaviour during the entire pressurization test. Another was used to measure the displacements in the high containment location during the time that pressures were changing between hold points. That camera was then moved to record the displacements at pressure hold points at the location near the equipment hatch. All three locations were permanently marked using so-called ‘monuments.’ These monuments will enable monitoring in future years to be performed at identical locations.
Dot patterns were applied to the surfaces to be measured (Figure 4). These dot patterns do not need to be maintained as they can be reapplied for future tests. The DIC technique works with regular or random patterns, with semi-regular or semi-random patterns being preferred. Each paint dot is approximately 3-7 pixels diameter, with approximately 50% coverage, so the finished pattern looks like a mix of equally-sized white and black dots. There are numerous materials and application methods which are very well-proven for DIC, including spray paint applied directly and through a template, brush paint, ink and rubber stamps or rollers with raised dots. The dot patterns at Ginna were applied by hand with marking pens. A pretest was done by taking two pairs of images and running the image correlation process. This verifies the completeness of the data set while also producing a noise floor map that shows the accuracy of the displacements and strains.
Measurements at Ginna included radial displacements, in-plane displacements, strains in the circumferential and vertical directions, and any crack-opening displacements.
Results and discussion
A sequence of plots taken from data collected at the high containment location is shown in Figures 5-12. The examination of the principal strain, as seen in Figure 5, showed ‘fingers’ of high strains. The ‘principal strain’ is the maximum strain independent of direction. These fingers are indications of surface crack locations in the concrete, caused by hoop stress from containment pressurization. The higher strain value on the left side means that the crack is wider there. Figure 6 and Figure 7 are the circumferential and vertical directional strains, which also show the crack location. The horizontal crack seen in Figure 7 was found to correspond to a cold joint location, a discontinuity in the vertical pour of the concrete containment wall that is visible to the naked eye. Note that there is higher strain indicated at the left end of the crack.
Displacement data confirms the crack indications of the strain plots. Figure 8 is the vertical displacement overlay at the 35 psi hold point with the rigid body motion removed, leaving only the local relative displacements. Note that the green colour represents zero movement, since the global rigid body motion has been subtracted out. There is upward motion above the crack and downward motion below the crack due to the crack opening. In Figure 10 at full pressure, relative vertical displacements have increased. Figures 11 and 12 are a sequence of initial and rigid body motion-corrected displacements for the horizontal direction, which indicate the presence of vertical surface cracks as colour discontinuities. Movements to the right (red) and left (blue) clearly indicate that surface crack openings are occurring.
The strain ‘fingers’ are indications of hairline cracks in the surface of the concrete. Cracking on the surface of the concrete is expected for a reinforced concrete structure. The reinforcing bars are approximately 7.6 cm (3 in) beneath the surface of the concrete. The outside surface of the containment is in tension and concrete has low strength in tension and cracks will form. Colour contrasts in the displacement results (for example, Figure 12) indicate the extent of crack opening. Crack opening displacements can directly be measured in the DIC software by placing a virtual extensometer across a crack. This was done for some of the cracks at Ginna in order to provide a quantitative measure of crack opening for future pressure test comparisons.
Figure 13 is the out-of-plane time history for two points taken once per minute up to the 35 psi hold point, indicating gradually increasing expansion of the containment cylinder. The two points taken were the bottom right and top left of the field of view.
The strain measurements recorded by the concrete strain gauge, adjacent to the free field DIC measurement location, showed good correlation between the two measuring systems.
The recording of this data, with similar strain and 3D shape measurements at the recorded test locations, will enable a direct comparison to future data. For the next SIT, which is planned in approximately 10 years, using the DIC measurements together with other augmented inspection methods will enable any differences in containment performance to be directly measured and quantified.
Paul Bruck and Tom Esselman, principals, New York City-based consulting engineering firm Lucius Pitkin, Inc., 304 Hudson Street, New York, NY 10013. Michael Fallin is a principal engineer, asset management, Constellation Energy Nuclear Group, 100 Constellation Way, Suite 200-C, Baltimore, MD 21202.
This article was published in the April 2012 issue of Nuclear Engineering International magazine.