Digital Image Correlation (DIC) is a non-contact, full field displacement optical measurement technique.
It is most often used in the following applications:
- Material characterization (CTE, glass transition temperature, Young’s Modulus, Poisson’s Ratio)
- Sample testing – fatigue and failure (in situ monitoring of displacements and strains)
- Applications where displacement or deformation measurements are needed
- High speed/frequency applications (i.e. crash testing, vibration)
Digital image correlation technique involves multiple steps. It works by tracking the movement of a group of pixels within a speckled pattern that has been applied to the surface of the sample/material. Because the movement is tracked by two cameras, stereo triangulation allows for capturing three-dimensional behavior. The rest of the work is done by software that can compute displacement and strain.
DIC has huge advantages over other measurement techniques. Unlike strain gauges, it is a wide-area solution. The lack of a physical connection also prevents the measurement from inadvertently influencing behavior of soft materials, which is a concern with strain gauges, thermo-mechanical analyzers (TMA), extensometers, and dilatometers.
If DIC is almost turn-key and so useful, what can go wrong? A LOT! Like all other measurement techniques, a comprehensive understanding of the limitations of each DIC step is necessary for a complete confidence in the output.
To better understand the process of Digital Image Correlation and its possible pitfalls let’s examine the initial DIC steps.
Step one is the speckling. Speckling involves applying a pattern of black and/or white features. It can be applied in a variety of ways, but the most common is either through a stamp, roller or spray.
The pattern can be applied to the area of interest or the entire sample. The pattern can be black and/or white because sometimes, primarily for high expansion translucent polymers, it is better to apply a random pattern of white and then place the polymer on a black background (see image below).
The key to success with speckling is the size, contrast, and consistency of the features. 5-10 pixels per speckle is recommended to optimize spatial resolution. A lower number of pixels per speckle can be ok if the features are consistent. Below 3 pixels per speckle, resolution starts to be an issue and can lead to inconsistent results.
Examples of these challenges are shown below. The first speckle pattern on the left is an example of inconsistent spot size. The second speckle pattern in the middle is an example of poor contrast. Both samples gave incorrect results when measured under DIC. Only the third speckle pattern on the right, with consistent spot size, small spot size, and good contrast, provided the right material properties.
When the speckle pattern is perfect, or close to it, the test setup can also be important. One area of debate for measuring coefficient of thermal expansion (CTE) is whether increasing the temperature should by a fast ramp and hold or by a much slower continuous ramp. Interestingly, work by ANSYS-DfR has shown that DIC measurements under both circumstances are about the same.
The next step in understanding DIC is capturing the repeatability and reproducibility (R&R) of the CTE measurement. Repeatability measures variation due to the equipment and reproducibility measures variation due to the operator (which includes applying the speckle pattern). Quantifying repeatability and reproducibility of material data is critical for successful and relevant simulation. As every good machinist knows, there is never one number for a measurement. Robust analysis increasingly requires simulation schemes that assign a tolerance to every geometric dimension and material property (such as CTE).
For repeatability, ANSYS-DfR had the same person perform two measurements on the same sample with the same speckle pattern. For materials with moderate to high coefficient of thermal expansion (CTE > 10ppm), the repeatability averages around 1% and is no higher than 5%. For materials with low CTE (CTE < 10ppm), the measurement variability can be much higher. Silicon, with a CTE of 4ppm/C, had a repeatability average around 5% while low expansion glass Zerodur, with a CTE of 0.1ppm/C, had repeatability average around 50% (see below). This behavior is understandable and is due to an increase in the noise-to-signal ratio as the CTE value decreases (see below).
By comparison, reproducibility had more variation. For reproducibility, ANSYS-DfR had different people perform multiple measurements on the same material, but different samples (and therefore, different speckle patterns). As with repeatability, there was a dependence on CTE. Materials with moderate to high coefficient of thermal expansion (CTE > 10ppm) had an average reproducibility of around 3%, with a maximum reproducibility around 6%. By comparison, silicon had an average reproducibility around 25% and a maximum reproducibility of 35%.
What does it all mean? If you do everything right in these two steps: speckling and CTE calculation, the measurements of polymers and metals should be within +/- 5% of actual. DIC can provide can provide a lot of useful information for a wide range of materials and samples, and it is more advantageous than other techniques. However, as you’ve learned the results are dependent on the skill of the person performing DIC, consistency and methodology. Small discrepancies can greatly affect measurements and in turn – simulation results.
To learn more about more Digital Image Correlation, register for one of the upcoming webinars on materials characterization.
THE VALUE OF DIGITAL IMAGE CORRELATION IN ELECTRONIC DESIGN AND ROOT CAUSE ANALYSIS