Challenges in synchronising high speed full-field temperature and strain measurement


The overall motivation for the research described in the paper is an enhanced understanding of the behaviour of fibre reinforced polymer composites subjected to high velocity loading. In particular, the work described here considers a method that allows the collection of synchronised high speed full-field temperature and strain data to investigate the complex viscoelastic behaviour of fibre reinforced polymer composites material that occurs at high strain rates. The experimental approach uses infra-red thermography (IRT) and digital image correlation (DIC). Because high strain rate events occur rapidly it is necessary to capture the images at high speeds. The paper concentrates on the challenges of the use of IRT and DIC at high speeds to obtain temperature and strain fields from composite materials, and in particular using them in a synchronised manner. In the future such data-rich techniques provide the opportunity for detailed investigation into the viscoelastic behaviour and allow in-depth material characterisation for input to future finite element or numerical models.


Increasing use of polymer reinforced polymer composites in high performance applications, e.g. military structures, is leading to an increased risk of impact or blast events imparting high velocity loading. Whilst the behaviour of such materials subjected to quasi-static elastic loading is reasonably well understood, the response to high strain rate requires further investigation. To reduce and mitigate the risk of failure it is essential that knowledge of the behaviour of these materials under high velocity deformation is established. Hence the subsequent effects of damage on structural performance can be defined. Therefore the motivation for this work is the need to map the effect of high velocity loading on the overall structural performance. High velocity/strain rate deformations are usually accompanied by a temperature evolution. Therefore the material behaviour is a function of time, strain and temperature, so to fully understand the material structural performance the thermomechanical material constitutive behaviour is required. The overarching aim of the current research is to provide thermomechanical characterisations of glass and carbon fibre polymer composite over a range of strain rates, with the ultimate goal of inputting the constitutive behaviour into a finite element (FE) modelling approach.

Hamouda [1] and Sierakowski [2] discuss the range of approaches for high strain rate testing. The preferred technique is the split Hopkinson bar that allows strain rates up to 104 s-1. However in this work a specialised conventional servo-hydraulic test machine (Instron VHS) is used, which allows moderate strain rates up to 102 s-1. Whilst this machine cannot match the strain rates of the split Hopkinson bar, it allows specimens of approximately 25 mm wide by 100 mm long to be used, unlike the much smaller coupons that must be used in conjunction with the split Hopkinson bar. These specimens are of a similar size, and aspect ratio, to those recommended for quasi-static characterisation by testing standards and therefore provides a larger surface for the application of optical measurements techniques.

The complex behaviour of fibre reinforced composite materials lends itself to the use of full-field optical measurement techniques, as information from the entire specimen is obtained that allows the identification of failure zones, loading paths etc. Digital image correlation (DIC) is used to measure strain, and infra-red thermography (IRT) to obtain the temperature evolution. One of the primary advantages of techniques such as DIC and IRT is that they are non-contacting, so the measurand does not affect the measurement by, for example, localised reinforcement or heating. However, it is essential that the images are synchronised temporally with any independent load or strain data collected from other sensors used in the experiment. Therefore the relative image capture rates, time delays and thresholding are important considerations. The aim of the current paper is to discuss the application of DIC and IRT to high velocity testing and the corresponding challenges. DIC and IRT are initially applied separately to both metallic and composite specimens, and some initial results are presented demonstrating the approach. The metallic specimens were expected to undergo higher strains and temperature changes, and were therefore chosen as a good starting point for assessing the abilities of DIC and IRT applied to such tests. This is followed by a discussion of approaches to synchronise data capture from the two optical systems and an independent load measure.


Digital image correlation is a full-field, optical method to measure the deformations and strains in a material or structure. DIC tracks the movement of a random surface pattern to monitor deformation or displacement. The random pattern is usually achieved by covering the surface of a component with a painted speckle pattern. Images of the deformation process are recorded using either one (2D DIC) or two (3D DIC) charge coupled device (CCD) cameras. The images are divided into discrete interrogation windows (or cells) and the displacement is obtained by tracking features within each cell [3]. Strain values are obtained by taking the measured displacement and dividing by the size of the undeformed cell. Strain resolutions are quoted as being as low as 40 ^strain [4], although this is highly dependent upon application and test conditions, such as lighting and alignment. The DIC technique has been successfully used to analyse the strains in heterogeneous engineering materials such as composites [5] and there is some reported work on the use of high speed cameras to collect images for DIC [6, 7]. Tiwari et al [6] described the use of high speed cameras for DIC, and the inherent limitations of such an approach.

To apply DIC to high velocity testing, commercially available high speed digital cameras are used to record the images. In this work the images are then imported into the DaVis 7.4 (LaVision) software for analysis. The application of DIC to high speed imaging uses the same speckle analysis algorithm as that applied at quasi-static test speeds. The accuracy of the algorithm is the same as quoted above, but additional sources of error are likely due to the acquisition of images using high speed cameras. To obtain images at the highest possible frame rates it is necessary to reduce the resolution of the sensor, therefore the user must accept a coarse strain map or use smaller cell sizes which give greater uncertainty in the strain result. Secondly it is more difficult to obtain well illuminated images with high contrast at high speed as the integration time must be reduced. Therefore it is important to increase the lighting intensity which may have adverse effects such as specimen heating, or heat haze in the images. Three different high speed cameras are used to capture the images for this work, Photron’s SA-1 and 3 cameras, and Redlake’s MotionPro X3 details of each are described in Table 1. The three cameras have similar maximum resolutions (~ 1 MP), and the X3 and SA-3 are capable of recording this resolution up to 2 kHz whilst the SA-1 extends this to 5.4 kHz. The advantage of the use of the SA-1 to collect images for DIC is clear, allowing full-size images to be used at higher strain rates. The X3 only uses vertical sub-windowing to increase the frame rate, leading to an image with high aspect ratio that lends itself to recording test on tensile strips. The SA-1 and SA-3 sub-window in both directions, producing squarer images. In both cases the compromise is always between spatial and temporal resolution.

Table 1 Specification of high speed cameras used in this work


MotionPro X3+

Photron SA-3

Photron SA-1

Max resolution (pixels)

1280 x1024

1024 x 1024

1024 x 1024

Max frame rate at max resolution (kHz)




Sub-windowing (pixels)

Vertical only to 16

Both to 128 x 16

Both to 64 x 16

Max frame rate (kHz)




Storage size (Gb/~full size images)




Steel and unidirectional (UD) GFRP tensile specimens were loaded at both quasi static speed (0.12 m/s, i.e. 2 mm/min), using a standard Instron electro-mechanical test machine, and then at 1 m/s, on a specialised Instron VHS 1000 test machine capable of testing specimens at speeds up to 20 m/s. Steel dog bone specimens with a cross-sectional area in the gauge length of 14.5 mm by 1 mm were used as proof of concept, three specimens at the QS speed and three at 1 m/s. Followed by GFRP specimens manufactured from ACG, MTM28-1\E-glass-200 prepreg with dimensions 200 mm by 20 mm and were 0.4 mm thick. During the tests the load was recorded by the load cell attached to the machine; in the case of the VHS this is a piezoelectric Kistler load cell. For comparison, the strain was separately recorded using Vishay’s CEA-06-240UZ-120 attached to Vishay’s Strainsmart system with a maximum sampling rate of 10 kHz. In addition, Photron’s SA-1 high speed camera was used to capture images during the 1 m/s tests. The specimens were prepared by spraying with black, grey and white paint to provide a speckle pattern. The cameras were recording at 30 kHz with an image size 512 x 256 pixels. The DIC was performed on each image using a cell size of 64 pixels and 50 % cell overlap, therefore providing a strain map with 16 x 8 data points. Figure 1 presents the evolving strain map for a steel specimen tested at 1 m/s, alongside a plot of the average longitudinal and transverse strains across the specimen. The plot highlights the advantages of using DIC. Where the strain gauge has debonded early the DIC measures up to 30 % strain, and allows longitudinal and transverse strain to be measured simultaneously.

There is little research reported in the literature on the use of high speed IR; however the few examples found used specifically designed systems. Noble [9] described the use of a thermal scanning camera to measure the temperature change occurring during high strain rate test on ductile iron at a rate of 1600 s-1 in a split Hopkinson bar rig. The scanner was an AGEMA 880LWB that used a liquid-nitrogen cooled CdHgTe detector with an accuracy of ± 2 K. The camera was only capable of scanning at 2500 Hz, and therefore was not fast enough to record temperature evolution during the test. Instead the camera recorded the temperature change approximately 0.5 ms after specimen fracture. The nature of the material tested, and the high speed applied, produced temperatures up to 573 K at the necking site. Three possible error sources were identified that could account for uncertainty of ~ 100 K; movement of the specimen with respect to the camera, change of specimen orientation and changing emissivity during deformation. Improvements in electronics allowed Zehnder [10, 11], to produce a system capable of 1 MHz with 64 HgCdTe detector elements in an 8 x 8 plane array. Studies of the temperature rise near the tip of a notch in high strength steel sample subjected to an impact showed the system was capable of a temperature resolution of approximately ± 2 K. Finally, more recently, Ranc [12] used a bar of 32 InSb infrared detectors to measure a line of temperature points on a high strain torsion test on titanium. This also sampled at 1 MHz, but was measuring temperatures of the order of 100s K.

In the current work infrared data was recorded using a Silver 480M (FLIR systems) detector. The Silver 480M uses a dual layer InSb sensor with 320 x 256 pixels. At maximum resolution it is possible to capture data at 383 Hz, and by windowing down to 48 x 4 pixels it is possible to achieve 20 kHz. It was decided to operate the detector at 15 kHz with a window of 64 x 12 and an integration time of 60 ^s as a compromise between spatial, temporal and temperature resolution. When operating the detector outside of its standard configuration it is necessary to perform calibration and non-uniformity correction procedures different to those provided by the manufacturers. These processes set-up the detector and the internal electronics for use at higher speeds, and altered window size to allow calibrated temperatures to be measured. Full details of the calibration and non-uniformity techniques, and their requirement, are discussed in a separate paper [13]. Tests were performed on the Instron VHS machine at an actuator speed of 10 m/s initially on a steel specimen and then a glass fibre chopped strand mat (CSM) specimen. The tests were performed as described in the DIC section, but the surface of the specimen was sprayed only with matt black paint to provide a constant emissivity.

Figure 2 is a full window image of the steel specimen before and after the test to demonstrate the location of the sub-windowed data, shown by the red box. By varying the stand-off distance it was possible to view the entire gauge length of the specimen, and it is clear from the post test image that the failure section was captured. The temperature evolution with time is plotted in Figure 3 at the six points displayed on the final image from the series show next to the graph. The data is displayed in its raw format as the number of digital levels measured. For reference the load signal from the Kistler load cell on the VHS machine is included shown by the blue line. While it is noisy there is a definite change in signal as the specimen fails at approximately 3 ms into the test time. The temperature at all six points increases steadily at a similar rate as the load is initially applied to the specimen, but towards the end of the test the temperature of the point at the failure location diverges from the rest as it rises rapidly. Therefore there are two distinct stages in the temperature evolution during the high speed test. The first represents temperature change as the specimen is stressed, whilst the second demonstrates large temperature increase due to damage at failure. The apparent cyclic temperature change at point six after failure has been identified as out-of-plane movement of the specimen.

The temperature evolution during the test on a CSM specimen is plotted in Figure 4. The data is plotted from the same points used for the steel specimen. It should be noted that the timescale of the temperature evolution is far shorter than for the steel specimen. The temperature rise is much sharper, approximately 0.1 ms, whilst the rise for steel was spread across 2 ms. Secondly the data from the composite is a couple of orders of magnitude less than the steel. It is evident that the shorter timescale and lower data values produce noisier IRT. Point six is near to the failure zone, and as for the steel specimen, there is a sudden temperature rise as the specimen fractures. However, point one on the CSM plot also shows a temperature increase lagging slightly behind point six. From the images taken during the test, and from the remains of the specimen post test, it became clear that two failure sites were present. The failure at point six was first to initiate, shortly followed by a fracture at point one. This highlights the more complex nature of tests involving composite materials. CSM specimens are particularly weak, and therefore during failure there is less energy and consequently lower temperature evolutions. It is expected that applying IRT to stronger composites, e.g. UD glass fibre specimens, will provide higher energy failures and higher temperatures.

Full window images of steel specimen before and after test

Fig. 2 Full window images of steel specimen before and after test

Plot of the temperature evolution of the specimen at six points during the test

Fig. 3 Plot of the temperature evolution of the specimen at six points during the test

Plot of the temperature evolution of the specimen at six points during the test

Fig. 4 Plot of the temperature evolution of the specimen at six points during the test


The overall goal of the research is to obtain full-field temperature and strain information from composite materials subjected to high velocity loading. Initial tests on the separate use of DIC and IRT at high speeds have shown promise in the two techniques although further work is required to improve consistency and obtain a better understanding of the error sources. The next step is to develop an approach that will allow both systems to capture data concurrently. In the literature there is little mention of the use of white light imaging and IRT together, particularly at high speed. Noble [14], used the IR system mentioned above in the IRT section and a white light high speed camera during a test on iron in a split Hopkinson bar rig. The white light camera was used to obtain deformation information by monitoring the change in shape of the specimen, and did not provide full-field strain. The IR system did not operate at a high enough sampling rate to capture temperature evolution during the test. Instead it was triggered shortly after the specimen failure to measure the maximum temperature at the fracture site. These systems were still effectively used separately to obtain information at different stages of the test.

To fully characterise the viscoelastic behaviour of the composite material subjected to high velocity loading it is necessary to measure load, temperature and strain at the same temporal location. Therefore an approach is required to synchronise the three data types together. The first challenge to overcome is to ensure all data capture systems are initiated at the same time; this is achieved using specifically designed LabView code for operation on National Instruments Compact Rio hardware. The Compact Rio monitors the voltage signal from the Kistler load cell and at a user defined threshold triggers the capture of load data and generates a logic signal that is sent to the two camera systems. The white light cameras and IR detector can be triggered from a digital pulse with a known jitter of the order of 100 ns. With a known frame rate it is a simple matter to find a corresponding load value for each white light and IR image. The second challenge is an artefact of the different performances of the white light and IR systems. Whilst the white light cameras can achieve frame rates over 100 kHz (although with much reduced spatial resolution), it has been shown that detector sensitivity will limit the IR system to approximately 15 kHz. Both systems are to be initiated at the same time, but with different frame rates they will quickly lose phase. It is envisaged that using the white light system at a frame rate that is a multiple of the sampling rate of the IR detector would ensure that for each IR image there is also a DIC strain image. It is advantageous to operate the white light camera at higher frame rates to assist the DIC algorithm in obtaining accurate strain information. The final perceived challenge is in the physical collection of the white light and IR images. DIC requires the specimen surface to be sprayed with a speckle pattern, whilst, to obtain the best results, IR requires a surface preparation of matt black paint. Further tests are required to investigate the effect of using a painted speckle pattern on the collection of IR data. The illumination required for DIC will also have an effect on IR data, particularly with specimen heating. ‘Cold’ light LEDs are being investigated to illuminate the white light images. It may also be possible to obtain DIC data from the front of specimen, and IRT data from the rear. Consideration must be taken though that any localised effects from damage may affect the relevance of the two data sets.


It has been shown that DIC and IRT with commercially available cameras, detectors and software are feasible techniques to obtain full-field strain and temperature information from high strain rate testing on composite materials. Further work is required to improve consistency and fully understand the errors involved. A system is in place that allows the data capture to be synchronised, even accounting for differences in sampling rates of the DIC and IRT. Consideration is required of the physical aspects of collecting data simultaneously from two techniques that have some conflicting requirements.

Next post:

Previous post: