Cutting-edge technologies such as artificial intelligence and supercomputing are increasingly used in earthquake prevention and disaster reduction

Catching earthquakes with a "super brain" (decoding)

Core reading

How to improve earthquake monitoring and risk prevention capabilities? In recent years, cutting-edge technology has often been seen in earthquake-related research.

Through deep learning of historical seismic data, the artificial intelligence seismic monitoring system "smart ground" can detect weak seismic wave signals and report the seismic parameter information within two seconds. With supercomputers, the area for disaster simulation assessment is wider and more accurate. With the development of science and technology, China's earthquake prevention and disaster reduction capabilities have been continuously improved.

When and where did the earthquake occur? How big is the magnitude? When on duty, Du Guangbao, a reporter at the China Earthquake Network Center, and his partner need to calculate the seismic parameter information in about 10 minutes, correct the automatic flash report results, and give a formal quick report.

Du Guangbao's duty cycle is 24 hours. "It's sleepless nights on duty days." He said that whenever the quake hits and the automatic quick-report system is triggered, he will be put into intense work. When aftershocks occur frequently after a major earthquake, the center's reporters will also be dispatched.

Is there a way to give accurate and fast information on seismic parameters? Today, a set of artificial intelligence full-automatic seismic monitoring system-"smart ground motion" is expected to provide a new method for real-time earthquake monitoring.

Deep learning historical seismic data, can calculate parameters such as position in two seconds

The "Smart Ground Motion" system was completed by the team of Professor Zhang Jie of the University of Science and Technology of China in cooperation with the China Earthquake Administration. Since December 2018, the "Smart Ground Motion" system has been trial-run at the China Earthquake Administration and is connected to the Sichuan and Yunnan Seismic Stations to process real-time data from more than 100 seismic stations in the China Seismic Experimental Fields in both places. Based on the results of 446 earthquake assessments in 2019, it was found that the "smart ground motion" system is very close to the results of manual calculations by professionals. The research and development team also used "smart ground motion" to test related seismic data of Oklahoma in the United States, and the seismic positioning error was within 4 kilometers.

"Smart Ground" is an artificial intelligence software system. Zhang Jie introduced that unlike traditional monitoring systems, it has deep learning capabilities and can quickly process ongoing seismic data based on millions of seismic data collected in memory and combined with seismological theory. For example, the traditional monitoring system is a purely theoretical expert, and "Smart Move" is a theoretical expert with deep historical memory. "Using artificial intelligence technology is equivalent to all seismologists on duty to process seismic data together." Zhang Jie said.

Being able to “learn” from the historical earthquake big data, “smartly move” has trained a pair of “eyes of fire” for monitoring earthquakes—it can effectively remove noise in various frequency bands and has strong recognition ability, thus monitoring weak seismic wave signals. Traditional monitoring systems generally deal with earthquakes over magnitude 3, and "smart ground motion" can report information on earthquakes over magnitude 1 and improve the ability of earthquake monitoring. Zhang Jie introduced that small earthquakes may be a precursor to the development of large earthquakes, and being able to automatically identify them is an important breakthrough in earthquake monitoring.

Accurately locating earthquakes and inferring rupture mechanisms are difficult points in seismic monitoring. The "Smart Ground Motion" system can record and infer the position and depth directly from the seismic waveform. At the same time, with the help of a powerful memory database, the system can calculate the location, depth, magnitude and source mechanism of the earthquake in 1-2 seconds.

What does "outcome in two seconds" mean? Zhang Jie said that information can be reported as early as possible after the earthquake, and it can issue early warning to areas where the seismic wave has not yet arrived, and take protective measures for the public and the government to determine rescue plans to seize valuable time.

Continuous improvement during trial operation, and has reached the technical indicators of routine monitoring

The international scientific and technological community has been paying close attention to the application of artificial intelligence technology in earthquake monitoring. Some countries have proposed plans to develop artificial intelligence monitoring systems, but most of them are still in the early research stage. "Smart ground motion" is currently the only real-time artificial intelligence earthquake monitoring system in the world. The core task is to assist the China Earthquake Administration in monitoring earthquakes in Sichuan and Yunnan. The system also shows great potential in earthquake early warning and prediction research.

Earthquake information is no small matter. At present, countries around the world are very cautious when adopting new seismic standards and systems, and it usually takes 5 to 10 years of testing. In addition, the seismic information released by the National Network Center of China has been carefully calculated and corrected by experts. "Smart movement" is an unmanned system. Therefore, some experts believe that when the machine gives a result, if there is a problem, it is difficult for people to go back and find the cause.

Zhang Jie admitted that this may indeed be a deficiency of artificial intelligence systems. In order to avoid the impact as much as possible, in actual operations, researchers combine artificial intelligence and traditional computing methods to analyze. If there is a big difference between the two, the monitoring system can analyze the cause of the difference and determine the best result. No worse than traditional methods. "

"We hope to improve the operating capability of the system and accomplish things that cannot be done manually." Zhang Jie said that after more than a year of test operation, the "smart ground" system has reached the technical indicators of conventional earthquake monitoring and earthquake early warning. Claim. "In the future, the system will need a long period of trial operation and improvement before it can replace 24-hour staff."

New technologies such as simulation systems help disaster assessment be more accurate and faster

In recent years, in addition to artificial intelligence, a series of cutting-edge information technologies have also been applied to earthquake-related research.

After the earthquake, which part of the city was most damaged and to what extent? The traditional method is to rely on manual surveys to investigate the damage status of buildings in various places. After 6 years of research, Lin Xuchuan, a researcher at the Institute of Engineering Mechanics of the China Earthquake Administration, developed a simulation system. The system relies on key geographic information system data such as house height, structure type, and construction year to quickly and accurately model the city. As long as you input the monitored seismic data, you can quickly generate a map of house damage in your area.

"Modern urban systems are very complicated. The damage of individual bridges during an earthquake may affect half of the city's traffic." Lin Xuchuan said that combined with real-time seismic monitoring data, the simulation system can give preliminary information about the city's damage immediately after the earthquake. Provide decision-making reference for disaster relief. In addition to real-time post-disaster assessment, the simulation system can also provide support for urban planning. Lin Xuchuan introduced that through the simulation system, scientific researchers simulated scenarios when various earthquakes with different magnitudes occurred, and given the destructive state of each house, so as to provide refined suggestions for urban disaster prevention and mitigation planning. For example, where should it be, how many shelters, fire stations, hospitals and so on.

As the amount of data increases and the degree of simulation refinement continues to increase, Lin Xuchuan feels that general computing systems can no longer meet the requirements and must use faster and more powerful supercomputers or cloud supercomputing platforms. At one time, he was going to carry out a detailed elastoplastic simulation analysis of a city with millions of orders of buildings. If he uses a daily computing platform, it may never be counted. Can be placed on "Tianhe No. 2", and the results are output in a few minutes. "Earthquake relief is precious every second. Using supercomputers can win more time." Lin Xuchuan said.

In general, a large earthquake affects more than just a single city. The analysis of earthquake damage to urban agglomerations within a region is very important to coordinate rescue resources. Large-area, refined, and dynamic simulations place higher demands on computing power. Lin Xuchuan believes that supercomputers will be an important strategic resource for intelligent society, and its application can provide a leap opportunity for simulating earthquake disasters.

■ Reporter's Notes

Use new technologies to respond more

As we do not know enough about the geological structure and stress changes inside the earth, we cannot directly enter the birth and occurrence of earthquakes in the earth. Earthquake prediction is still a worldwide scientific problem.

But in the face of potential earthquake disaster risks, we can not only respond passively, but we can plan ahead.

Whether it is using artificial intelligence technology to monitor earthquakes, issuing early warning information as soon as possible for protection time, or using supercomputers to make urban earthquake-resistant refinement planning, we can help us to be more calm when facing disasters. We can also take the initiative to deduce the underground situation by making controllable artificial earthquake sources, etc., find out the "temper" of the earthquake, and improve the response level.

It is not easy for ordinary people to directly observe the progress of seismic technology. Although human beings are unable to fully grasp the law of earthquakes in the short term, our understanding of earthquakes is constantly deepening. The application of some new technologies and new methods has provided scientists with more confidence in dealing with earthquakes.

Yu Sinan