ALBUQUERQUE, NM – Geologists at Sandia National Laboratories have used 3D printed rocks and an advanced, large-scale computer model of earthquakes in the past, to understand and prevent earthquakes caused by energy analysis orra.
Injecting water underground after extraction of unconventional oil and gas, commonly known as fracking, geo-regulatory energy stimulation and carbon dioxide capture can cause earthquakes. Of course, energy companies do their best to detect cracks – there will be cracks in the earth’s crust prone to earthquakes – but sometimes earthquakes, even shovels of earthquakes, hitting unexpectedly.
Sandia geologists studied how pressure and pressure from injecting water can move through pores in rocks down to fault lines, including previously hidden ones. They also crush rocks with weak points that are specifically engineered to hear the sound of various types of failure, which helps in the early detection of an earthquake.
3D printing variability provides basic structural information
To study different types of fault failure, and their warning signs, geoscientist Sandia Hongkyu Yoon needed a lump of rocks that would break in the same way every time he was stressed – a weight that was not unlike that. the pressure caused by water entering underground.
Natural stones collected from the same place can be very different in terms of mineralization and healing, causing different weak points and types of fractures.
Several years ago, Yoon began using additive manufacturing, commonly known as 3D printing, to make rocks from gypsum-based ore under controlled conditions, believing that these rocks would be more uniform. . To press the rocks, Yoon and his team sprayed gypsum in thin layers, forming 1-by-3-by-0.5-inch rectangular blocks.
However, as he studied the 3D printed rocks, Yoon realized that the printing process also generated minute structural differences that influenced how the rocks broke. This sparked his interest, prompting him to study how mineral textures in 3D printed rocks affect how they break.
“It turns out that we can use that variability of mechanical and seismic responses of 3D printed fractures to our advantage to help us understand the underlying processes of fracture and its impact. water flow in rocks, “said Yoon. This fluid flow and pore pressure can cause earthquakes.
For these experiments, Yoon and colleagues at Purdue University, a university with which Sandia has a strong partnership, made mineral ink using calcium sulfate powder and water. The researchers, including Purdue professors Antonio Bobet and Laura Pyrak-Nolte, printed a coating of hydrated calcium sulfate, about half as thick as a sheet of paper, and then applied a water-based binder. to glue the next row to the first. The binder reabsorbed some of the calcium sulfate to gypsum, the only mineral used in drywall construction.
The researchers printed the same rectangular rocks and a gypsum-based cylinder. In some rocks the mineral layers of gypsum ran flat, while others had straight mineral layers. The researchers also changed the direction in which they sprayed the binder, to create more variation in the mining situation.
The research team sent out the samples until they broke. The team examined the fractured surfaces using lasers and an X-ray microscope. They noticed that the broken path depended on the direction of the mineral layers. Yoon and his colleagues described this fundamental study in a paper published in the journal Scientific Reports.
Sound signals and machine learning to classify seismic events
Also, working with his colleagues at Purdue University,? Yoon observed acoustic waves coming from the printed samples while they were breaking. These sound waves are signals on fast microcracks. The team then combined the audio data with machine learning techniques, a type of advanced data analysis that identifies patterns in seemingly unrelated data, to detect signs of minute seismic events.
First, Yoon and his colleagues used a machine learning method called a random forest algorithm to aggregate the microseismic events into groups caused by the same types of microstructures and identify about 25 important features in microcrack audio data. They ranked these features according to meaning.
Using key features as a guide, they created a multifaceted “deep” learning algorithm – such as the algorithms that allow digital assistants to operate – and applied it to archived data collected from real events. . The deep learning algorithm was able to identify signals of seismic events more quickly and accurately than conventional survey systems.
Yoon said that within five years they hope to implement many machine learning algorithms, such as those and those with hidden geoscience principles, to detect co-excited earthquakes. -linked to fossil fuel activity in oil or gas fields. The algorithms can also be applied to detect hidden cracks that may be unstable as a result of carbon capture or georegulatory stimulation ?, he said.
“One of the nice things about machine learning is the scalability,” Yoon said. “We always try to apply specific concepts that have been developed under laboratory conditions to major problems – which is why we do laboratory. So fast.” in that we have tested these machine learning concepts developed at the scale of the laboratory on archival data, it is very easy to scale it up to great difficulties, compared to traditional methods. “
Stress moves through rock to deep cracks
It was a hidden fault caused by a strange earthquake at a georegulatory stimulus site in Pohang, South Korea. In 2017, two months after the last georegulatory stimulus test came to an end, a magnitude 5.5 earthquake hit the region, the second strongest earthquake in South Korean history recently.
After the earthquake, geoscientists discovered a fault hidden deep between two injection wells. To understand how pressure traveled from a water injection to the fault and what caused the earthquake, Kyung Won Chang, a geologist in Sandia, realized that he needed to consider more than the pressure of water pressing on the rocks. In addition to that deformation pressure, he also had to account for how that pressure moved to the rock as the water flowed through pores in the rock itself in his large-scale computing model. .
Chang and his colleagues described weight movement in a paper published in the journal Scientific Reports.
However, understanding deformation stress and pressure movement through rock pores is not enough to understand and predict some earthquakes as a result of energy exploration activities. The architecture of various cracks must also be considered.
Using his model, Chang analyzed a 6-mile-long, 6-mile-wide, 6-mile-deep cube where a swarm of more than 500 earthquakes occurred in Azle, Texas, from the November 2013 to May 2014. The earthquakes occurred on two intersecting faults, one less than 2 miles below the surface and another farther and deeper. While the shallow fault was closer to the wastewater injection sites, the first earthquakes occurred on the fault farther and deeper.
In his model, Chang found that the water injections increased the pressure on the shallow defect. At the same time, injection-induced pressure moved through the rock down to the deep fault. Because the deep fault was initially under more pressure, the epicenter of the earthquake began there. He and Yoon shared the advanced computing model and their description of the Azle earthquakes in a paper recently published in the Journal of Geophysical Research: Solid Earth.
“In general, we need multivariate models that combine different types of pressure outside just pore pressure and rock deformation, to understand induced earthquakes and link them to energy actions, such as irrigation stimulation and in waste water injection, ”Chang said.
Chang said he and Yoon are working together to implement and scale machine learning algorithms to detect previously hidden cracks and identify geological weight signatures that could trigger the magnitude of an earthquake.
In the future, Chang hopes to use these weight signatures to create a hazard map for induced earthquakes around the United States.
###
His research effort, in addition to Yoon’s original work, was funded by Sandia Lab Lab’s Research and Development program. Yoon received funding from the Department of Energy’s Fossil Energy Office to continue the research.
Sandia National Laboratories is a multicast laboratory run by National Technology and Engineering Solutions of Sandia LLC, a wholly owned subsidiary of Honeywell International Inc., for the Department of Energy’s National Nuclear Security Administration in the US. Sandia Labs has key research and development responsibilities in nuclear deterrence, global security, defense, energy technologies and economic competitiveness, with key facilities in Albuquerque, New Mexico, and Livermore, California.