[ad_1]
Researchers at the Tokyo Metropolitan University applied machine learning techniques to obtain fast and accurate estimates of local geomagnetic fields using data collected at several points of observation, potentially allowing the detection of changes caused by earthquakes and tsunamis. A Deep Neural Network (DNN) model was developed and trained using existing data; The result is a fast and efficient method for estimating magnetic fields for an unprecedented early warning of natural disasters. This is essential for the development of effective warning systems that can help reduce the number of victims and widespread damage.
The damage caused by earthquakes and tsunamis leaves no doubt that an effective means of predicting their impact is of paramount importance. Admittedly, warning systems already exist just before the arrival of seismic waves; However, it often happens that the wave S (or secondary wave), that is to say the later part of the earthquake, has already happened when the warning is given. A faster and more accurate way is absolutely necessary to give local residents time to seek safety and minimize losses.
Earthquakes and tsunamis are known to accompany localized changes in the geomagnetic field. For earthquakes, this is essentially what is called a piezo-magnetic effect, where the release of a huge amount of stress accumulated along a fault causes local changes in the geomagnetic field; for tsunamis, it is the vast and sudden movement of the sea that causes variations in atmospheric pressure. This in turn affects the ionosphere and then changes the geomagnetic field. Both can be detected by a network of observation points located at various locations. The main advantage of such an approach is speed. By reminding us that electromagnetic waves move at the speed of light, we can instantly detect the impact of an event by observing changes in the geomagnetic field.
However, how do we know if the detected field is abnormal or not? The geomagnetic field at various places is a fluctuating signal; the whole method relies on knowing the "normal" field of a site.
So, Yuta Katori and Assoc. Professor Kan Okubo of the Tokyo Metropolitan University has developed a method to take measurements in different parts of Japan and to create an estimate of the geomagnetic field at different points of observation. Specifically, they applied an advanced machine learning algorithm known as deep neural network (DNN), modeled on how neurons are connected inside the human brain. By feeding the algorithm with a large amount of inputs from historical measurements, they allowed him to create and optimize an extremely complex and multilayered set of operations that map as efficiently as possible the data to what was actually measured. With the help of half a million data points taken in 2015, they were able to create a network capable of estimating the magnetic field at the point of observation with unprecedented accuracy.
Given the relatively low cost of computing DNNs, the system can potentially be associated with a network of high-sensitivity detectors to quickly detect earthquakes and tsunamis, providing an effective warning system to minimize the damage and save lives.
Source of the story:
Material provided by Tokyo Metropolitan University. Note: Content can be changed for style and length.
Source link
