On Characterizing Strain Fields in Impact-Damaged Composites using Shape Descriptors

One approach to damage assessment in composite structures is to examine the strain field induced by service loads and compare the fields measured during maintenance with those observed in pristine or virgin components. Levels of damage that cause appreciable changes in the strain field observed in service are significant since it is very likely that they will cause premature failure and the measured strain field provides the input data for predicting the reduced life. While visual observation and comparison of strain fields is possible, it does not offer a very viable approach to large-scale inspections. The automation of the process is not straightforward because data collected in-service during maintenance inspections may not be located in the same spatial coordinate system, at the same scale or orientation as the benchmark data acquired from a virgin component. In addition, each data set may consist of the order of million pieces of information rendering quantitative comparisons cumbersome and potentially expensive. Image decomposition allows such images to be represented by less than a hundred coefficients through exploiting the redundancy present in most images. In this research, two-dimensional fields of strain, i.e. maps of strain, have been treated as images; thereby, enabling the use of geometric moments that are invariant to rotation, scale and translation as a means of representing strain data. The image descriptors for a specific strain field can be grouped in a feature vector so that the feature vectors for strain fields in damaged and virgin component can be compared using the Euclidean distance between the vectors in multi-dimensional space. The Euclidean distance between feature vectors represent strain fields in undamaged and damage laminates has been found to correlate well with the level of damage measured using ultrasound and extremely well with the energy of impact used to induce the damage.