Math-based frameworks often use laboratory material failure measurements to predict a structure’s resistance to cracking.

But even engineers who know the math sometimes struggle to predict fractures in materials that have complex microstructures or components made with 3D printing. Existing prediction frameworks also do not always work well for ductile metals, such as some steels, that deform and stretch before they fracture.

To compare the different methods, researchers at the Energy Department's Sandia National Laboratory presented three challenges to colleagues: Given the same basic information about the shape, composition and loading of a metal part, could they predict how it would eventually fracture?

Failure prediction models were not available when the Liberty Bell cracked in the 1800s. Source: National Park ServiceAn overview of the Sandia Fracture Challenge was published in the International Journal of Fracture. And, a collaborative community of researchers has formed to refine prediction techniques for engineering reliable structures made from a variety of materials.

The Sandia researchers said failure predictions typically involve multiple rounds of experimental measurements and calculations. This means that the modeling is calibrated to known fracture data. For the Sandia-led challenges, however, participants did not know the outcome until after the end of the competition.

13 teams

The first challenge, held in summer 2012, attracted 13 teams of researchers from universities, national labs and companies to predict crack initiation and spreading in a common stainless steel alloy. They all received the same engineering drawing of the test piece, microscope images of the material’s microstructure, data about the material’s fracture toughness and measurements of how much stress it accumulated when strained. Then, each team applied its own method to predict a crack’s path under a given amount of force.

Meanwhile, researchers at Sandia and the University of Texas at Austin, who were not participating in the prediction competition, fractured the material in their labs. They loaded test pieces into machines and pulled on them until they tore in half. Cameras recorded the crack paths and instruments measured the amount of force on the samples.

It turned out that none of the 13 predictions completely matched the experimental results, although many reportedly worked well for aspects of crack formation. With only one situation for comparison, it was hard to determine which prediction methods were most effective, the researchers said.

Testing titanium

Two years later, the Sandia team issued a second challenge. This time, 14 teams predicted the fracture pattern in a component made of a titanium alloy common in airplanes, spacecraft and medical devices. The teams were asked to predict crack formation from very slow loading as before and under rapid loading, such as that experienced in a car crash.

Rapid loading creates heat in the material and leaves little time for the heat to dissipate. In the second challenge, most teams did not combine thermal and mechanical modeling. Those that did, however, tended to get the details right, the researchers said.

3D printing test

The third challenge, held in 2016, asked researchers to predict cracks in stainless steel machined with a 3D printer. The microstructure of printed metals can be more porous than in the forged metals used in the earlier challenges. The researchers wondered if the internal porosity could make printed metals fracture sooner than expected.

For this challenge, 21 teams received characterization data from tensile tests and detailed microstructural imaging. All of the teams predicted the crack initiation site and the resulting path that was observed during experimental tests.

Challenge participants have gone on to form a community-owned collaboration through the Structural Reliability Partnership. This group of scientists and engineers at universities, industry and national labs is working to improve models of fracture. There are 17 institutions in the partnership, and partners share results with each other before they are published.

Some of the group's initial interests include predicting physical properties of 3D printed metals and studying how hydrogen gas alters metal in hydrogen infrastructure. Predictions like these could help engineers better understand the reliability of shock-loaded springs or bolted joints, which often are overdesigned to compensate for poorly understood fracture behavior.

In the future, the partnership’s efforts could expand to study plastics and ceramics, and examine fracture behavior at the micro-, nano- and atomistic scale.