A heavy lift

Growing up in the remote countryside of China’s Hunan province, Z.J. Wang didn’t see trains or automobiles often, but about once a week he saw military planes flying overhead. “I was totally fascinated by that,” Wang says. “I heard stories that those pilots ate meat every day, but I only had meat three or four times a year. I wanted to be a fighter pilot, even dreamed about aircraft.”

During some physical tests in middle school, though, Wang’s equilibrium didn’t handle spinning very well. Although that ended his dream of being a fighter pilot, his success in science and math launched a new aspiration — to be an aerospace engineer.

Today, Wang is a Spahr Professor of aerospace engineering at the University of Kansas in Lawrence, where he and his students work on predicting flow over the NASA CRM-HL, for common research model high-lift configuration. Wang describes this project as part of “a joint community effort to tackle one of the remaining challenges in computational fluid dynamics — predicting the maximum lift.”

Scientists around the world can contribute experimental or computational data to the  High-Lift Prediction Workshops. “The experimental effort is mostly used to validate the computational simulations,” Wang explains.

The approach Wang’s team uses to predict the maximum lift is called wall-resolved large eddy simulation (WRLES). The flow near a jet’s surface is highly turbulent, involving chaotic eddies with a wide range of length and time scales. In WRLES, even the small near-wall eddies are computed to accurately predict flow patterns and lift. Many in the aerospace community, however, believe such a simulation will not be feasible for another decade.

Wang thought his team could do it much earlier, if he could run his lift-solving algorithm on the Summit supercomputer, a petascale machine at the Oak Ridge Leadership Computing Facility (OLCF).

Turbulence plays a crucial role in lift generation. When smooth airflow — laminar flow — becomes turbulent, the friction on a jet dramatically increases. “To really do a good job of accurately predicting the transition from laminar to turbulent flow and the air friction on a high-lift configuration, you have to resolve the small near-wall eddies,” Wang explains.

He believed that his implicit high-order compressible flow solver would be at least 10 times faster than existing approaches. Implicit means that this method simultaneously solves many nonlinear equations to obtain the solution at a future time, and can do so in larger time steps, which makes the computation more efficient. Using high-order schemes — up to sixth order, instead of, say, a second-order scheme  — increases the accuracy of the output. As a simplified example, a third-order scheme uses a quadratic equation that can more closely fit data than a straight line used in a second-order scheme. Furthermore, Wang’s solver can handle a high-order unstructured mesh, which uses curved lines and surfaces to accurately model a complex shape, such as a wing’s surface. Last, the solver works with the compressible flow that occurs at speeds above about 0.3 Mach, which is nearly 230 miles per hour.

‘The computational results even capture fine details that escape other models of these eddies.’

With funding from the computational mathematics program of the Air Force Office of Scientific Research, Wang’s group has worked on the algorithms and flow solver for more than 20 years. Part of that work produced hpMusic, which was designed to simulate compressible flow. In addition, hpMusic can run on GPU-based machines, such as Summit or OLCF’s Frontier, the world’s fastest supercomputer and the first to achieve exascale, or a quintillion (billion billion) operations per second. As Wang says, hpMusic “has been validated for a wide range of problems.” It even formed the basis of GE Aerospace’s GENESIS code for simulating jet engines.

To optimize hpMusic, though, Wang reached out for assistance. Computer scientists at NVIDIA helped him make hpMusic get the most out of Summit. Plus, scientists at the software company Cadence Design Systems helped generate a high-order unstructured mesh with over 140 million elements.

With the program optimized for a GPU-based high-performance computer and a 2023 INCITE allocation on Summit, the Wang-led team was ready to test hpMusic’s ability to simulate the NASA CRM high-lift configuration.

Although one hpMusic job requires nearly a month to run, Wang’s team was allowed only 12-hour runs on Summit. “So we had to restart it very often,” he says. Even with a job completed, it took days to transport the roughly 3.5 terabytes of data from Summit to KU. Visualizing the results, says Wang, “was a nightmare, because the output produced tens of billions of elements and points.”

With all of the data visualized, Wang notes two key findings. “First, we did it,” he says. The results showed fine details in eddy patterns on the aircraft surface. Second, those patterns closely resembled the “intricate separation patterns near the wingtip” in experiments. The computational results even capture fine details that escape other models of these eddies.

Future simulations could play a part in the aerospace industry’s move to certification by analysis versus experimentation only. “We can use this kind of simulations to accurately predict the maximum lift, and such simulations have the ability to replace some — definitely not all — experiments,” Wang says. If a simulation shows, say, separation of airflow from a jet’s wingtip, which would reduce lift and increase drag, “maybe the wing could be changed to mitigate that, but there is the never-ending quest to improve the accuracy and efficiency of the simulations.” Higher lift and less drag could increase an aircraft’s fuel efficiency and reduce jet noise.

Still, Wang’s use of hpMusic is just getting started. “We’re extending it to more complex problems,” he says. “One of those is airflow over a helicopter’s rotor blades.” He even envisions simulating an entire jet engine, but that would require adding still more physics to a simulation. “I don’t know whether we will be able to do it before I retire, but it gives me a challenge.”

After decades in aerospace engineering, Wang’s childhood fascination for flight continues. Even today, when he sees a jet fly by, he thinks one thing: “It’s astonishing!”