Test optimization for the identification of elastic and orthotropic parameters
Topic(s) : Experimental techniques
Co-authors :
Antoine VINTACHE (FRANCE), Pierre BAUDOIN (FRANCE), Renaud GRAS (FRANCE), Ludovic BALLERE , Nicolas SWIERGIEL (FRANCE), François HILDAbstract :
Identification of constitutive parameters for composite materials often means relying on a large number of tests, particularly when identifying damage models. To be able to design shorter test campaigns, this work aims at identifying as many elastic parameters as possible for a layered composite through a single Iosipescu test. The composite is symmetrical, with layers alternating between two orientations (ɑ and β), for a total of 20 plies. Using two finite element simulations, a virtual study is proposed. The first simulation is used to get pseudo-experimental data, namely, 3D displacement field measurements on a single surface and load levels. These data mimic global StereoDIC measurements and load cell readings in a real experiment. To approximate an actual experiment more closely, uncertainties on the measured data are introduced. These data and their uncertainties serve as inputs of the identification procedure. A subsequent simulation represents the ideal model and deviates from the pseudo-experiment simulation solely on the value of constitutive parameters. This model will be updated by the identification procedure to better match the pseudo-experimental data.
A weighted Finite Element Model Updating algorithm is employed to calibrate the material parameters. This algorithm has been chosen as it accounts for the uncertainties of experimental data. Moreover, the measured data are weighted according to their spatial and temporal densities. A preliminary sensitivity analysis determines which parameters are identifiable with respect to uncertainties. To improve the number of identifiable parameters, the geometry of the sample is modified. In this study, three geometrical parameters are optimized: the distance (L) between the two clamps of the Iosipescu setup and the orientations (ɑ, β) of the two groups of plies (figure 1). Through exploration of the parameter space, maps of the number of identifiable parameters are determined as well as a set of optimized geometrical parameters (figure 2). Last, the elastic and orthotropic parameters are identified using the optimized geometrical model.
This approach shows the effect of geometrical parameters in the Iosipescu setup to identify as many parameters as possible, thereby allowing for a better characterization of the material behavior. This method can be used to design test campaigns by maximizing sensitivities of target parameters, while minimizing the number of tests that are necessary to calibrate these parameters.
A weighted Finite Element Model Updating algorithm is employed to calibrate the material parameters. This algorithm has been chosen as it accounts for the uncertainties of experimental data. Moreover, the measured data are weighted according to their spatial and temporal densities. A preliminary sensitivity analysis determines which parameters are identifiable with respect to uncertainties. To improve the number of identifiable parameters, the geometry of the sample is modified. In this study, three geometrical parameters are optimized: the distance (L) between the two clamps of the Iosipescu setup and the orientations (ɑ, β) of the two groups of plies (figure 1). Through exploration of the parameter space, maps of the number of identifiable parameters are determined as well as a set of optimized geometrical parameters (figure 2). Last, the elastic and orthotropic parameters are identified using the optimized geometrical model.
This approach shows the effect of geometrical parameters in the Iosipescu setup to identify as many parameters as possible, thereby allowing for a better characterization of the material behavior. This method can be used to design test campaigns by maximizing sensitivities of target parameters, while minimizing the number of tests that are necessary to calibrate these parameters.