This Scientist Took A Deep Dive Into A Pool Of Sewage Treatment Plant Data. Here’s What He Fished Out
May 24, 2016
The thought of sewage treatment plants is enough to make most people hold their noses. But Jason Nichols, who works as a chemical scientist at GE’s Global Research Center, decided to jump into the pool of data that such plants produce. He and his team will use it to build computer models — what GE calls “digital twins” — of the treatment process, in hopes of finding out what’s really going on inside the murky waters. “A well-operated plant smells like fresh soil, or compost,” Nichols says. “A fetid smell is an indication that something is going wrong with a plant.”
The digital twin could help identify the source of a nasty smell — and much more. It uses computerized real-world information and algorithms to re-create the plant (minus the odor) in the data cloud. The insights will allow Nichols and his team to make wastewater treatment more efficient. “Few people think about it, but some really expensive stuff happens after you flush the toilet,” Nichols says. “We’ve got great treatment technology, but we don’t always operate it in the most efficient fashion. If we built a digital twin of every sewage treatment plant, for example, we could save $4 billion to $6 billion globally over the next decade.” Since GE is also in the wastewater treatment business, connecting its membrane bioreactors and other technology to the twin could save customers between $200 and $300 million.
Sanitation is an expensive business. U.S. municipalities spend nearly $100 billion per year in tax money installing, upgrading and operating water and sewage treatment systems. With costs expected to keep growing, town councils are desperate for savings.
Nichols, who has a PhD in organometallic chemistry and did postdoc work at the University of California at Berkeley, smelled an opportunity. His team has spent months analyzing data from physics and biokinetic models of sewage treatment plants. They concluded that many of them waste a small fortune by pumping too much air into ponds of partially treated sewage.
Nichols and his team intend to fish for insights to make the process more efficient by first installing chemical sensors inside a plant and then building cloud-based algorithms that suck in data and mimic the biochemical work taking place inside the plant. The team has quickly started getting some clarity using biochemical and physical simulations as the first step in creating a digital twin. Although the amount of sewage that arrives at the plant varies considerably over days, weeks and months, most facilities set the air pumps that feed their waste ponds to maximum. “It makes sense,” Nichols says. “If you can’t reliably predict how much sewage you’ll be dealing with, you’ll pump in the maximum amount of air so you don’t oxygen-starve your microbes in their treatment ponds.”
Using the digital twin, Nichols expects to discover hidden patterns in plant operations, anticipate the amounts of sewage coming in and calibrate the oxygen level needed. “We can see reasons why certain things might be happening and propose solutions,” he says.
For example, an early warning about high nitrate or phosphorus levels could protect municipalities from polluted waterways. The digital twin would not only alert plant operators to a problem, but identify what portion of the plant is most likely the cause and reduce response times for repairs. Early warning and rapid response in such cases would save plant operators from paying costly fines and ensure cleaner waterways for communities.
The digital-twin technology could have other valuable benefits. Operators can use it to monitor the health of bacteria in their systems, schedule the right time for maintenance or even predict the cost of future plant expansions to accommodate a growing community.
Several hundred GE scientists like Nichols are building digital twins of jet engines, wind farms and subsea blowout preventers. The company wants to build a digital model of every machine it makes, to make them work more efficiently and save customers money from unplanned downtime. Based on data from its digital twin, for example, a jet engine that might normally be overhauled every 24 to 36 months may not end up requiring such a service until after 38 months.
How soon might wastewater treatment plants start embracing Nichols’ proposals? He recently made a presentation to water utilities and plant operators and is in discussions with a number of plants to find a suitable facility to become a development partner in what will be the first practical demonstration of the technology. “Once we show them what their data can do for them, we’ll be in business,” he says.