Onzekerheidsbepaling van digitale beeldcorrelatie en de impact op materiaalidentificatie

The digital image correlation (DIC) technique is a full-field non-contact optical deformation measurement tool. The technique correlates images of a specimen before and after deformation, so that the deformation can be retrieved. The correlation is obtained by comparing the similarity of the image features (intensity of pixels) before and after deformation to establish the correspondence between undeformed images and deformed images. Since it was proposed in the 1980s, the technique is largely improved and widely applied in different fields. However, as a measurement method, the accuracy of DIC is still a problem. Plenty of studies have been done concerning the performance of DIC, and yet there are no solid solutions to assess the accuracy of the measurement. In this research, effort is contributed to the analysis of the error of DIC both theoretically and experimentally. A comprehensive theoretical analysis is presented for the measurement errors on the displacement field and the strain field obtained by subset-based DIC. In general, the systematic measurement error can not be estimated, but the random measurement error is well studied. For the displacement error, a general solution is provided to assess the random measurement error due to the imaging noise. The solution indicates that the random displacement error is determined by the imaging noise level, the intensity gradient of the interested image, the subset size, the sub-pixel part of displacement, and the interpolation scheme. For the strain error, the analysis suggests that the random error relates to the DIC implementations (i.e. subset size, step size and strain window size), the extent of deformation, and the random error of displacement measurement. As the theoretical analysis can not fully estimate the measurement error, an experimental approach is proposed to study the performance of DIC with virtual tests associated with finite element analysis (FEA). The idea of this approach is to numerically simulate a real DIC test, and assess the measurement error of DIC afterwards. In this approach, the interested experiment is firstly simulated by FEA with prescribed material properties. The obtained deformation fields are extracted from FEA, and are utilized to deform a reference image numerically. DIC is then performed to measure this deformation by correlating the numerically deformed image and the reference image. By comparing the DIC measured deformation with the FEA results, one is enabled to assess the accuracy of the DIC measurements. In the current stage, this approach could be generalized to any 2D DIC applications. Amongst possible applications of DIC, an important one is the study of the the mechanical behaviour of materials, which is often described by phenomenological constitutive equations (e.g. hardening law and yield criterion). The constitutive equations are one of the basis of FEA for material simulations. To obtain the parameters for the constitutive equations, inverse modelling methods can be adopted associated with full field measurements such as DIC. Inverse modelling methods include several different categories, among which the finite element model updating (FEMU) method is intuitive to understand. The FEMU method combined with DIC has been well developed for years. However, the accuracy of DIC measurements is rarely considered. This is rather dangerous since misinterpretation of the measurements leads to meaningless results. In this thesis, the uncertainty of DIC measurement is studied. As such, the measurement error can be taken into account by using a weighted cost function in FEMU. The mechanical properties of steel DC06 are studied. The results suggest that the identified parameters are sensitive to the measurement errors, and more appropriate parameters are expected when the measurement error and the strain profile are considered. The accuracy of phenomenological constitutive equations is limited: indeed they are only valid for the experiments to which they are calibrated. As such, the identified parameters will generally depend on the type of experiments used to calibrate them. Therefore, FEA of the material under specific deformation might not be acceptable when the parameters are identified for another deformation state. Consequently, it is necessary to identify the parameters from tests in accordance with the interested deformation state so that FEA is capable to provide optimized simulations. As out-of-plane deformation are generally involved in the modern manufacturing fields, material properties obtained from 2D tests might not be sufficient for accurate FEAs of these processes. In literature, however, FEMU with stereo tests is seldom discussed. In this work, a method is proposed to associate FEMU with stereo DIC in an Erichsen bulging test. It is found that the optimized constitutive parameters for the Erichsen test are different from the ones obtained from 2D tests, which proves the necessity of the stereo FEMU method.nrpages: 242status: publishe