Training deep learning model for BET surface area estimation of plant carbon
by learning from Scanning Electron Microscopy (SEM) images harvesting from scientific journal databases for prediction BET

Example Results: WebApp(Prototype)

Abstract


overview

   Currently, the world is facing problems with global warming and energy shortages. Therefore, every sector of society places great importance on alternative and clean energy sources. For example, solar energy and wind energy are receiving a lot of attention as they are renewable and sustainable energy sources. However, in order to use renewable energy, it is necessary to have energy storage devices to store energy for use during times when there is no sunlight or wind, for example. Supercapacitors, which are high-capacity electrical charge storage devices, have a higher power density than other energy storage devices and do not generate heat, making them safe for users. Additionally, carbon from natural sources is also an interesting material because it has good electrical properties that can lead to high electrical charge capacity and natural resource conservation. However, the material used for producing electrical charge storage must have a high specific surface area (BET surface area) in order to be effective. However, in the traditional specific surface area calculation, high resources are required such as time and cost in calculating the BET surface area using Scanning Electron Microscopy (SEM) images. Normally, it takes approximately 36-72 hours for researchers to develop the model for Machine Learning to estimate the BET surface area from SEM images using Deep learning approach to reduce the time and cost of calculating the specific surface area. This will be beneficial for other researches in calculating the specific surface area in the future. The benefits gained from the research on developing a machine learning model for estimating the specific surface area (BET surface area) of carbon from plants using Scanning Electron Microscopy (SEM) images from a research article database through deep learning are that there is now a web application available for predicting the BET surface area value by analyzing SEM images.

Project poster