Olivares says that such tools are needed because NIAR currently must use very generic models that don’t allow it to capture the details of various fibers and matrices. “Right now, we compensate for that lack of fidelity of models with a lot of testing at the coupon and the component levels, so that we can fine tune the models using those specific tests. But it will be very helpful in the future to have better materials models that are able to capture these failure mechanisms,” says Olivares. Researchers are currently working on micro, meso and multi-scale path analyses for composite materials, which could result in more predictable models that will not depend as heavily upon the coupon and component testing.
Modeling all the complex mechanics of failure in the matrices or fibers would take a very long time to compute, Olivares says. But he remains hopeful that such models will eventually reduce the need for some testing.
“If this approach takes off, we will be able to generate mechanical properties as part of simulation models and that will make things move toward the virtual wall that will enable the fast design of materials in initial life,” he says.
Accounting for Unique Properties
Engineers who want to design with composite materials often wish for a database of composite properties similar to the ones that exist for various metal materials. But Pipes says that’s not realistic since each combination of a fiber and continuous base creates a new material. Change the percentages of materials in a composite, change the shape of the fibers or change the way that they are manufactured and you get a material with different properties.
“That’s a problem, because now the guy that wants to use it doesn’t know what its properties are,” explains Pipes. Two-phase composites – a matrix and a fiber – are complicated enough to define; when you add in another phase – the interphase, or interface between the matrix and fiber – there are millions of combinations (and variations in properties) that could result.
Simulations, rather than databases, are the answer, says Pipes. “If I can simulate the properties, if I can predict them from the composition and the makeup of the composition and the meso structure of the composition, then I don’t have to measure them. I can predict millions of combinations, then pick what I want and do a few tests to substantiate the properties of that combination. But that’s not where the world is today,” Pipes adds.