Applying the science of air flow analysis to the prediction and prevention of viral droplet infection

Fugaku Supercomputer Simulations and Mask Experiments

Professor Iida reflects on what led to the opportunity: a March 2020 conversation with Professor Makoto Tsubokura (Kobe University, Riken) on the way back from a research conference.

"As the pandemic approached, we talked about how this wasn’t the time to be doing normal research," says Professor Iida. Professor Tsubokura was researching phenomena relating to automobile engine gasoline inflow.
"For decades, we had been engaged in joint research to address issues such as automobile noise reduction. Since we had a computer program for droplet movement, we thought it could certainly help contribute to preventing infection from droplets. While there are significant differences in conditions between engines and humans, gasoline and saliva, and the speed that droplets travel, the fundamental ideas and relevant calculations are the same," he says.

Coincidentally, at the same time, MEXT and Riken issued permission to use the computing resources of the Fugaku supercomputer (which was set to launch in 2021) on a trial basis in order to combat the novel coronavirus. With that, they decided to investigate how the virus spreads in the air.

Their focus was on mask efficacy. At that time, while manufacturers provided performance ratings for mask filters, there was no systematic evaluation of the effectiveness of actually wearing the masks. On top of that, at the time the WHO was not recommending mask usage for those without symptoms, believing that masks were not effective. So why were medical staff wearing masks if they weren’t effective? Was there an etiquette-based reason to wear masks when coughing? To answer these questions, Professor Iida and the team decided they would need to clarify what masks could actually do for us.

Performance of masks (measured values) (left) and experiment movie ( right)

However, the miniscule fibers of masks are measured on a micron scale, which would make it very challenging to calculate a realistic computer analysis including all conditions. To address this, they decided to conduct an experiment to obtain data prior to performing calculations with the Fugaku supercomputer.

"Droplet calculations were handled by Professor Tsubokura, while Kyoto Institute of Technology Professor Masashi Yamakawa examined how the virus moved in the air, and I oversaw the mask experiments. Coincidentally, Assistant Professor Tsukasa Yoshinaga was researching human pronunciation in my lab, and as he had just purchased a mannequin head for his research, we decided to put a mask on it to conduct experiments," says Professor Iida.

The experiments commenced in April. Through these experiments, Professor Iida and the team determined mask pressure loss (degree of breath resistance) and permeability. and modeled the mask by measuring how much droplets a person would expel while taking into account the gap between the face and the mask.

Showing Risks in a Variety of Cases

Receiving cooperation from companies such as Kajima Corporation and Daikin, the team then used a calculation method devised by Professor Tsubokura to conduct simulations on droplet spread in locations such as offices, elementary schools, hospitals, and train cars. Superimposing the experimental mask data was, the Fugaku supercomputer then calculated the degree of mask efficacy in each location using numerous cases.

As early as June, a video was created for presentation to the press showing the efficacy of non-woven masks, droplet spread from coughing and conversation at offices and restaurants, the efficacy of opening windows in train cars, and the efficacy of partitions. These results received extensive news coverage and were widely adopted to make the infection countermeasures.

"We were in a rush for about two months leading up to the first announcement because we wanted to quickly dispel public uncertainty over the efficacy of masks and what conditions affected the risk of infection. To be honest, there was some wishful thinking that the pandemic might be reined in by summer," says Professor Iida.

From June onwards, Professor Iida and the project team surveyed around 200 types of commercially available mask, and repeated simulations across various daily scenarios. As a result, by August, the project team had been conducted over 2,000 analyses.

"To be honest, 50% of these analyses were failures. The cause of the failures was rooted in the fact that, as opposed to models for vehicles, the droplets move slower and remain in the air longer. Despite this, thanks to the Fugaku supercomputer, we were able to progress the experiment quickly through a process of trial and error. Even if a calculation failed after coming up with an idea, we were able to keep our motivation since we could immediately move on to the next idea. This is what made our research possible," says Professor Iida.

They subsequently publicized numerous guidelines showing the efficacy of masks made from non-woven fabrics, gauze and urethane, the effects of double masking using non-woven fabric and urethane masks, and infection risk from speaking or singing loudly.

"For example, loudly conversing at restaurants or karaoke produces 10-14 times more droplets than regular conversation, so ample precaution is required.
On the other hand, our demonstration of mask efficacy created new issues, such as the peer pressure placed on people who would not wear non-woven fabric masks. Wearing a high-performance mask can sometimes lead to heat stroke, and also increases the amount of carbon dioxide in the mask, which increases the risk of anoxia.

To start, masks are only helpful when there is someone infected within the group, so for the most part, their purpose goes unfulfilled. Nonetheless, to reduce infection, it is best to wear a mask. Cloth masks are also effective to a certain extent, and I think it is important to continue to take reasonable measures based on each situation as long as the risk is present"

Experiments and Simulations in Conjunction

Professor Iida and the team are continuing their research into coronavirus countermeasures. Recently, an experiment that generates droplets using artificial vocal cords has shown that droplets significantly increase when the speaker converses by vibrating their vocal cords, compared to when their vocal cords do not vibrate. Based on this finding, the team applied a formula to determine infection probability, which examined the degree of virus inhalation required for infection, and showed the infection risk for each event per infected person.

"One hour in a karaoke bar increases the probability of infection by 25%. At 90 minutes, the risk rapidly increases, so when dining it is best to finish up as quickly as possible. In addition, we recently discovered that seat configuration at restaurants affects the infection risk potential."

In the future, the team intends to create computer models for the throat, respiratory tract, and lungs to investigate how the virus attaches and how infection spreads when inhaled. There will also be an investigation of vaccine efficacy.

"This research should prepare us for the next infectious disease. Ultimately, I would also like to continue with my main research on vehicles and wind turbines. In particular, my research into giant ocean-based wind turbines, which I began some time ago, is a challenging topic that will be essential to utilizing natural energy. That said, since the foundations use aerodynamics, the approach is the same.” Addressing the appeals of his research, Professor Iida says that “while not all of my research is immediately helpful to the world, the reward of fundamental research is that it can be applied to various issues."

Reporter's Note

Professor Iida was admitted to Toyohashi University of Technology after graduating from Tokyo Metropolitan College of Aeronautical Engineering (now Tokyo Metropolitan College of Industrial Design). With a desire to work with vehicles, upon graduation he began research at Hitachi Ltd., where he used aerodynamics to develop bullet trains.

"It was a good company, but with corporations, as the years go by, it is hard to do your own work, since you may move into a managerial position or the company may restructure. I thought it would be boring if I could no longer conduct experiments, so I moved to a university," says Professor Iida.

While his skills with experiments ended up helping to evaluating the performance of masks, his research intentions at the university were not solely focused on applications for society.

"If you study the fundamentals, it can end up being helpful in some way. I mean, at one point I found myself researching the flight of dragonflies. It’s difficult to know what that could lead to," says Professor Iida with a laugh. The fun of fundamental research is that it can lead to discovering problems that are shared across various phenomena.