Optimisation Based Composites Materials Characterization

AMET s.r.l. has developed a technique for the simulation of full-scale crash tests on structures built with composite materials. AMET has already assessed the performance of the virtual simulation in predicting the behaviour of these heterogeneous materials on various applications, with good results.
However, composites show a peculiar behavior, especially in presence of damages. In fact they follow failure paths uncommon for conventional materials and their main damage mechanisms are fiber breakage, matrix cracking and delamination at interfaces. The onset and the propagation of damages are able to absorb and dissipate a great amount of incoming energy; as a result, composites are the best candidate for energy absorbing structures. With the available commercial FEM code just part of these mechanisms can be successfully simulated, since no existing material model can give a reasonably accurate prediction of delamination. Consequently, the technique here proposed works well when composite structures have to be studied, but they miss to be accurate when the delamination is the main failure occurring, like e.g. in crash boxes. In these cases, even if the material model is well characterized, it cannot be exhaustive, because an efficient treatment of delamination implies an accurate definition of interlaminar stress field, difficult to achieve with the shell elements actually available. Nevertheless, for the other cases, in which matrix and fiber failures are dominant, this technique can give good results and can drive the design toward an adequate exploitation of composites.

The present technique focuses on the characterization of composite materials, in unidirectional and fabric configurations, according to two key features: first of all, the characterization should be based on classical experimental tests, as for standard materials (i.e. traction and compression tests); moreover, the material, identified by the experimental curves fitting, should be accurate enough to be introduced in the full-scale virtual model, without shape factors or correction parameters. Thus, the technique requires six experimental curves as inputs, i.e. traction and compression tests at 0°, 90° and 45°.  Once the experimental tests are available and the FEM code has been chosen, two FEM specimen models are prepared to simulate the experimental tests, one for the compression tests, the other for the traction tests. The models should be built on the real experimental specimens, whose shapes and sizes are reproduced; the virtual specimens are generally modelled with shell elements, since they are used for full-scale simulations. The sample models are loaded and constrained like the physical samples. The aim of this phase is the reproduction of the 6 experimental tests by the simulations. Through the use of optimization techniques, an analysis campaign is performed to obtain the material parameter values that bestfit the experimental tests, on all the loading conditions at the same time. On the basis of the optimization results, the material model card can be filled and its parameters defined. Both fabric and unidirectional composites can be characterized by the present technique, even though different procedures should be followed. In any case, the material model defined in the characterization phase can be directly introduced in a full-scale simulation, without additional considerations, allowing to simulate the behavior of the real material even when embedded in complex structures and subject to complicated loading, such as crash tests.

This methodology makes the introduction of composite materials possible in structural car components with a limited effort in terms of time and costs; a prototype software to deal with the complete process has been developed and it will be released shortly.


The Author

Michele, Rabito Crescimanno    
Manager of Numerical Simulation Department