No abstract

The Thermographic Signal Reconstruction (TSR) technique, based on the logarithmic polynomial fitting of the temperature time-evolution, improves the efficiency of pulse-stimulated thermography. However, the non-destructive evaluation of the inspected structure requires the analysis of the full temperature-time sequence. This drawback impacts all transient thermal NDE method whatsoever. A new post-processing development of the TSR data, based on the use of the very limited number of polynomial coefficients, avoids this time-consuming task by producing a unique composite image of the defects, containing information on their depths. This image results from an RGB-basis projection of three of the most pertinent coefficient images.The performance of the method is compared with that of the classic TSR method, in the case of a carbon/epoxy coupon containing artificial defects simulating delaminations. Comparisons are also made with three other well-known data processing techniques: Pulse Phase Thermography (PPT), Principal Component Thermography (PCT) and HighOrder Statistics (HOS). In each case, the relevance of the new approach is proven: not only is it simple and rapid, but it also guarantees a defect detectivity at least as good as the one provided by the other techniques. Moreover, the RGB projection, successfully applied to all four methods (TSR, PPT, PCT and HOS), is shown to be an efficient operation to compress and synthesize the thermal data which was, so far, scattered in several images. It may be considered as a general post-processing technique useful for any such method currently used in NDE.

No abstract

a b s t r a c tThe mechanical behavior of a four-layer plain weave glass fiber/epoxy matrix composite is modeled at the mesoscopic scale, taking into account the dry fabric preforming before resin injection, the relative shift and nesting between fabric layers, and the characteristic damage mechanisms, i.e., intra-yarn cracking and decohesion at the crack tips. The surface strain fields obtained numerically are similar to the strain fields observed at the surface of the specimen. Damage is modeled by introducing discrete cracks in the FE mesh of the representative unit cell of the composite. The crack locations are determined using a stress based failure criterion. The predicted locations are similar to those observed experimentally. The effects of intra-yarn cracks on the macroscopic mechanical properties show the same trends as the experimental data. Good quantitative agreement is obtained if yarn/yarn or yarn/matrix decohesions at the crack tips are taken into account.