Assessing the feasibility of monitoring strain in historical tapestries using Digital Image Correlation

Transcription

1 Young Stress Analyst Competition YSA09 Assessing the feasibility of monitoring strain in historical tapestries using Digital Image Correlation Djallal Khennouf University of Southampton, School of Engineering Sciences, Highfield Southampton, SO17 1BJ, Textile Conservation Centre, Park Avenue, Winchester, SO23 8DL 1. Introduction Tapestries are produced by weaving on a loom where closely spaced, highly twisted yarns, known as warp yarns, are stretched and fixed in one direction. Less dense yarns are woven transverse to the warp yarns to produce the pattern; these are known as the weft yarns. On completion the tapestry is hung so that the weft yarns support the weight of the tapestry (Figure 1). Today, historical tapestries constitute a vital part of the European cultural heritage. Warp yarn Weft yarn Figure 1: Illustration of weft faced tapestry weave [1] After extended periods of display, certain regions in the tapestry start to weaken due to self-weight and external environmental factors e.g. relative humidity (RH) and temperature. If the weak regions are left without conservation, they eventually become heavily damaged areas which are both disfiguring and very expensive to rectify (Figure 1). The aim of this research is to study the mechanical aspects of degradation. To achieve this it has been suggested [2] that full field strain measurement is required. This work investigates the feasibility of using Digital Image Correlation (DIC) to monitor strain in tapestries. 1

2 2. Objectives The main objectives of this work are: i. Conduct laboratory-based quasi-static experiments to investigate if the weave pattern can be used as the device for correlation for DIC strain measurements ii. Devise a systematic approach to determine the optimal DIC processing parameters, mainly the subset size and the overlap iii. Develop a methodology for extracting strain from actual tapestries in-situ and devise an approach for automated continual strain monitoring iv. Cooperate with the manufacturer of the DIC system to develop a mapping function that allows the camera position to be changed during the in-situ monitoring process 3. Experimental work, results and discussion 3.1. Using the textile weave pattern as a correlation device Instron test machine Camera Grips Reference subset P Q PTU Test specimen ADC DIC system computer Matching subset Figure 2: Experimental set-up for quasi-static testing 2

3 Normally, DIC measurements require the application of paint coating to the surface of the specimen. However, some materials have natural intrinsic pattern that can be used for correlation. To identify the optimal DIC processing parameters, an experimental program was carried out on representative textile material specimens with standard surface preparation (see Figure 3a). The work showed that the optimal processing parameters for the textile material are 64 x 64 subset size with 0% overlap. To investigate if the weave pattern can be used as the device for correlation further experimental work compared DIC strain measurements from specimen (A) with specimen (B) which had no paint coating applied to its surface (Figure 3b). The results revealed that the textile weave pattern can be used as the device for correlation, meaning that the problematic application of paint coating can be dispensed of. A further part of the work was to investigate if surface detractors that normally arise from sharp changes in tapestry colours have considerable effects on DIC measurements (Figure 3c-d). It was found that the effect of such detractors is negligible. 128x128 pixels Figure 3: Representative textile specimens with four different surface patterns 3

4 3.2. In-situ monitoring of historical tapestries Camera Velcro Tapestry DIC system computer RH and Temperature sensor Figure 4: Typical arrangement for in-situ DIC monitoring An experimental approach was developed to conduct long term monitoring of actual tapestries in a gallery setting (Figure 4). Various experiments have been carried out in closely controlled RH and temperature conditions. Three tapestries have been used which differ in terms of dimension, mass, age, material, and weave density. The length of the experiments was between 48 hours and 300 hours. The results revealed that the tapestry responds rapidly to the variation in RH levels. As illustrated in Figure 5a, a change of approximately 6% in RH has led to significant changes in the longitudinal strain. The scatter diagram of Figure 5b depicts that the tapestry exhibits positive strain as RH exceeds 53% and negative strain as RH falls below 51%. Between 51% and 53% the tapestry exhibits strain levels that are very close to zero. These results, which agree with previous findings, show that RH plays a major role in the degradation process of tapestries. Future work will focus on accelerating the strain cycling depicted in Figure 5a in laboratory-based experiments on representative textile material samples. This will lead to a better understanding of the effect of RH variation on tapestries and lay the ground for a novel approach to artificially age textiles by accelerating the effect of RH. 4

5 A B Figure 5: The effect of RH changes on the global longitudinal strain in the verdure fragment 5

6 3.3. Automated continual monitoring and the mapping function For artefacts on display in historical houses and museums, it is not appropriate to have a camera system installed as shown in Figure 4, as it detracts from the visual impact of the pieces. To overcome this limitation, a method for long term monitoring must be developed where the DIC cameras can be moved between readings. This way the DIC system can be transported to a gallery setting at regular intervals (e.g. every 3 months) to record an image pair, then the collected images (which have different calibration parameters) are correlated to obtain deformation data. However, despite the recent advances in multiple view geometry in computer vision [3], applying DIC on image data with different calibrations while keeping the same strain resolution is still a challenging proposition. Therefore, this feature is not yet available in standard commercial DIC systems. The manufacturers of the LaVision DIC system have devised a mapping function to allow for the movement of the cameras between readings. One of the objectives of the work presented in this summary is to coordinate with the manufacturer to test and improve the performance of the mapping function. For this purpose, an experimental program was devised where quasi-static tensile tests were carried out on representative textile samples. The data was processed using the standard strain computation functions as well as the mapping function. Finally the results were compared and discussed with manufacturer. Based on this work, it has been decided that the resolution of the cameras must be increased in order to achieve the required strain resolution. Figure 7 shows the effect that the replacement of 2MPixel cameras with 5MPixel cameras had on the strain resolution. Currently, the mapping function allows global strain measurement to be conducted with a precision of 0.03% if the cameras have not been displaced by more than 200 mm from their initial position. This will enable the next step in the program, which is to apply the mapping function to actual tapestries in a historical house. 6

7 (a) With 2 MPixel cameras (b) With 5 MPixel cameras Figure 6: The scatter in the mapping function strain results before and after camera replacement 7

8 4. Conclusions The main conclusions from this work are: i. The quasi-static experiments on the representative textile material validated the precision of DIC and showed that the optimal processing parameters are 64 x 64 pixels subset size with 0% overlap. Further work on the representative textile material showed that the weave pattern can be used as the device for correlation and that precise global strain measurements can be obtained even in the presence of surface pattern detractors ii. In-situ DIC monitoring has been performed on three tapestries. The work revealed that the tapestries exhibited strain oscillation that has strong correlation with RH. Positive global strain dominated when RH rose above 53% and negative global strain was dominant when RH fell below 51%; global strain was nearly zero when RH was in the range of [51% - 53%]. This suggests that this range may be the ideal humidity condition for preserving these tapestries iii. It has been shown that the mapping function has the potential to enable monitoring of tapestries in historical houses and museums. The work has shown that the function allows global strain measurement with a precision of 0.03% to be carried out if the cameras have not been moved by more than 200 mm from their initial position 5. References [1] Lennard, F. and Hayward, M., Tapestry Conservation: Principles and Practice, Butterworth- Heinemann (2006), Oxford, UK [2] Dulieu-Barton, J.M., Sahin, M., Lennard, F. J., Eastop, D. E. and Chambers, A.R. Assessing the feasibility of monitoring the condition of historic tapestries using engineering techniques, Key Engineering Materials, 347 (2007) [3] Richard, H. and Andrwe, Z., Three-Dimensional Computer Vision: A Geometric View, Cambridge University Press, (2003), UK 8