Estimation of turbulent states from limited measurements
We are studying a practical system in a wide area environment by the method of “inverse analysis”
that searches the cause from the result. Consider the problem of finding the source of air pollution. Sensors
that measure atmospheric components are installed at various locations. It is easy to identify the source by
arranging many sensors in a small area and observing the concentration gradient, but in the actual environment
where complicated turbulence occurs in a wide area, the method is inefficient.
In order to accurately
estimate the source from noisy data with fewer sensors, we are studying a method of maximum likelihood
estimation.
On the seabed of Japan, submarine hydrothermal deposits have been found in places such as the Bayonace Knoll
(Izu/Ogasawara area) and the Izena Sea Cave (Okinawa area), and groundwater heated by magma erupts in that
area. It is considered to be an important resource in Japan because precious metals and rare metals are
deposited and precipitated by rapidly cooling it with the surrounding seawater.
However, it is not easy
to discover new hydrothermal deposits because the amount of sensor information that can be obtained is limited
in the vast and deep sea that is isolated from humankind.
Therefore, we are studying a method of
measuring and exploring while moving in the sea using an autonomous robot equipped with a sensor.
(Collaboration with the University of Tokyo, Maki Lab)
In the future, we will introduce multiple exploration robots equipped with flow velocity, temperature, and
concentration sensors, and use the observation data obtained at stations installed on the seabed to determine
the three-dimensional distribution of seawater flow, seawater temperature, chemical concentration, etc. and
inversely estimate its time evolution.
Furthermore, based on the results of the back analysis, the robot
moves to a position where it is easier to observe the flow and substance concentration conditions. We aim to
discover the source efficiently by repeating measurement and movement.
This method can be applied in
various ways depending on the environment and fluid, such as using a drone with a sensor to fly if it is the
atmosphere, or using a self-propelled robot if it is on the ground. In recent years, along with the
development of IoT, various sensor information is becoming available in the environment. We hope to integrate
this information and create new services and businesses that estimate environmental conditions.
Reconstruction of a turbulent state based on randomly distributed
sensors
Left) True flow state Right) Estimated flow state by integrating limited measurement
data into simulation
It can be seen that the estimated state in the right figure converges to the true
state with increasing the number of sensing points shown by yellow dots.
Reference: Suzuki &
Hasegawa, J. Fluid Mech. (2017), Liu & Hasegawa (2020)
Scalar source estimation based on limited sensing information in turbulent
environment
(Left) forward simulation, (Right) Adjoint simulation
Reference: Cerizza et
al. (Flow Turb. Comb., 2016), Wang et al. (J. Fluid Mech., 2019)
Poster
