Flexible Sensors Use AI for Weather Forecasting to Increase Disaster Preparedness

Weather forecasting has become more accurate over the decades and can readily inform us days and weeks in advance of adverse conditions, but it's still not 100% accurate. The technology behind that science tends to be bulky, expensive stationary equipment such as radar, which impedes timely updates on local conditions for personal use. Researchers from Osaka Metropolitan University are looking to close that gap in information using flexible, inexpensive, lightweight sensors that utilize AI for measuring rain volume and wind speeds. The sensors can be affixed to almost any surface and feature a computational analysis reservoir.

"The findings open up a promising economical approach to weather reporting, contributing to disaster preparedness and greater community safety," stated Professor Kuniharu Takei. The sensors function Rain volume is determined by measuring the electrical resistance generated when a raindrop hits its surface. The sensors are protected by a superhydrophobic silicone sheet of polydimethylsiloxane (PDMS), which is infused with graphene and further processed with a laser. The laser provides texturing that allows constant control and measurement of the behavior of water droplets, no matter if they're stationary, sliding, bouncing, or splitting on the sensor surface.

Of course, wind speed has a significant effect on water droplet behavior, so the scientists used a machine-learning algorithm known as reservoir computing to analyze multiple pieces of weather data. Changes in rain and wind conditions result in resistance changes, which are picked up by the sensor and then recorded as time-series data. The resulting data is then used to train the machine, which predicted the pattern and reported rain volume and wind speed as output information. While there is still further development to increase the sensor's accuracy, the team expects them to become part of next-gen weather sensing.