Researchers from the National Laboratory. Lawrence in Berkeley applied an innovative method to study the processes of degradation of lithium batteries with quick charging. Scientists have created a unique algorithm for machine learning based on X-rays of chemical processes in batteries and added data from other studies on this topic to the model. The “scientific AI” created in this way is able to easily understand the problem than scientists.
In its study, scientists used an improved light source radiant in the range from infrared to X-ray. At the same time, dozens of experiments were carried out using chemical visualization methods or scanning translucent X-ray microscopy (STXM). The lithium battery was cyclically charged and the discharge, and the equipment simultaneously monitored the physical and chemical processes in the hundreds of battery particles in 50 different energy levels.
The 5,000 images “rape” specially created by the machine learning algorithm obtained during the experiment. And it was not a simple algorithm that could extract certain data from the pictures. It is argued that the algorithm analyzed the images based on a number of pre-entered data and models obtained in other studies in similar subjects. In fact, repeat, scientists argue about “scientific AI”, which could be “meaningful” to analyze the necessary, and not accidental information.
The algorithm created in the laboratory, combining machine learning and X-ray microscopy, promises to save years of research, relatively quickly conducting an analysis of the behavior of particles in batteries under load and clarifying the fundamental processes that occur in the materials inside the batteries.
Patrick Herring (Patrick Herring), Senior Researcher at the Toyota Research Institute, who supported this work as part of its program of accelerated design of materials, said: “Understanding fundamental reactions occurring in the battery, we can extend its service life, to provide a faster charging and, Ultimately, design batteries from the best materials. “