Titanium alloys serve as cornerstone materials for the aerospace industry - stronger and lighter than steel, resistant to rust and corrosion and resilient past the melting points of most other metals. Companies such as RTX, formerly Raytheon Technologies, rely on these sturdy alloys to build such vital machinery as jet-engine turbine blades, landing gear and exhaust ducts.
But this workhorse comes with some expensive trade-offs: The blazing heat necessary to process a titanium alloy into usable components typically wastes as much as half the raw metal as chips.
On top of the initial cost, the metal grains that make up most titanium alloys run in a single direction, like the wooden splinters that make up a twig. Under enough heat and pressure, components, like twigs, can crack and break during production.
"They're useful materials but extremely expensive because we lose so much in the machine forging to fabricate the part to its final geometry," said Tahany El-Wardany, an RTX senior technical fellow for advanced manufacturing. "We knew there had to be a better way, but certifying a new material for use could have taken 10 years or more. That's after the trial and error to perfect the ratio of elements for a new alloy."
Simulations performed on the Summit supercomputer at the Department of Energy's Oak Ridge National Laboratory are cutting through that time and expense by helping researchers digitally customize the ideal alloy. ORNL distinguished scientist Balasubramaniam Radhakrishnan, a specialist in computational modeling of materials, worked with the RTX team and used Summit to develop a predictive model that streamlined the experimental process to point to the most promising possibility.
The results hold promise for improvements across the aerospace field and beyond. The new alloy could cut the annual $273 million production costs of machining titanium components in half and save the company as much as 2.5 quadrillion British thermal units in energy costs by 2050, according to the research team's estimates.
"Thanks to Summit, we have a candidate for an improved titanium alloy," El-Wardany said. "Now we can begin work to manufacture a physical part. We still have a lot to do as far as real-world testing to verify the findings, but Summit's predictive simulations shrank a decade of physical testing into what we hope will be 2 or 3 years."
Mix-and-match metallurgy
RTX relies on such alloys as Ti-6Al-4V - a blend of titanium, aluminum and vanadium - for most of its aircraft components. Forging those alloys requires extreme heat and precision that make the process vastly more difficult than typical manufacturing.
The research team, drawing on recent studies in the field, theorized adding copper to raw titanium could yield an alternate alloy and that 3D-printing techniques could be used to knit the alloy's grains into an equiaxed, or latticed, microstructure with grains running horizontally and vertically that would hold up under greater stress. Molten metals could be mixed and printed to produce the desired components, which would cut out many of the traditional production steps.
"When we have the grains running equally in both directions in this equiaxed microstructure, fractures are much less likely. Then once the part is printed, it's ready to go," Radhakrishnan said. "But first we needed a way to simulate this microstructure and see what causes this pattern to form and how stable it is. These machine parts have complex geometries that are built layer by layer during 3D printing. Different layers could experience different cooling rates during 3D printing. Could we reproduce this new structure under the necessary conditions for these complex geometries?"
Finding a way to produce the latticed microstructure in a titanium alloy comparable to the industry standard would yield not just an improved product but huge savings. Such parts could be 3D-printed using powdered titanium made from the scraps left over from traditional production.
"These parts could be made for roughly half the cost," El-Wardany said. "If we could fine-tune production to achieve these results via 3D printing with this new alloy, it would be fantastic. These kinds of simulations are extremely time-consuming and computationally intensive, so they were beyond the capability in-house."
Confident calculations
The team turned to ORNL for help and received an allocation of time on Summit, the Oak Ridge Leadership Computing Facility's 200-petaflop supercomputer. Radhakrishnan used phase-field modeling, a computationally demanding and time-consuming mathematical technique that rigorously captured the physics of the process. The approach sought to simulate the complicated dynamics of melting and solidifying for various alloy blends of titanium, copper and the metal niobium.
Radhakrishnan used MEUMAPPS, an open-source code designed at ORNL for phase-field modeling, to perform the study. The computational power of Summit allowed the team to simulate microstructures down to the nanometer - about a millionth of a millimeter - under a wide range of extreme conditions, including various stages of heating and cooling and the evolution of millions of microscopic metal grains.
"We had to capture in the simulations not just the thermodynamics but these millions of convoluted interactions that evolve during production," Radhakrishnan said. "When an alloy like this forms, we see these interactions between the solid grains and the surrounding liquid. The simulations captured these details quite nicely."
The results suggested higher ratios of copper and niobium made the equiaxed microstructure more likely to form. Cooling rates, the simulations indicated, could be tailored to produce the stronger alloy.
RTX technicians sought to take the results from simulation screen to factory floor. The team identified a part - an integrated blade rotor, or bladed disk used in compressors - and proposed a 3D-printing approach to build the part.
The simulations captured so much detail that RTX crews found their production process needed tighter standards to reproduce the study's findings. Current production standards couldn't keep pace with the simulations, which led to inconsistent results in microstructure formation.
"The samples we produced didn't match the published findings in terms of reproducing this microstructure, but we believe that's due to a production issue," El-Wardany said. "During production, we feed the titanium powder through hoppers, and it's difficult to control the consistency of how that powder's fed. We trust the computational results over the production results in this case, and we're fine-tuning our production to meet the computational standards."
The team expects continued studies on Frontier, the OLCF's exascale supercomputer and fastest computer in the world, could yield further insights.
"ORNL has given us a model, and we can carry on with further simulations for digital design," El-Wardany said. "Once we overcome the physical difficulties with the manufacturing process, we can perfect the production process over small runs of parts and scale up to larger runs. These simulations have already saved us years of work, and we can continue to apply this approach as we seek better, more advanced materials and designs."
Support for this research came from the DOE Office of Science's Advanced Scientific Computing Research program and from the DOE Office of Energy Efficiency and Renewable Energy's Advanced Materials and Manufacturing Technologies Office High Performance Computing for Manufacturing program. The OLCF is a DOE Office of Science user facility at ORNL.
UT-Battelle manages ORNL for DOE's Office of Science, the single largest supporter of basic research in the physical sciences in the United States. The Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit https://energy.gov/science.
