The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project will be the acceleration and improvement of battery manufacturing and production. Forecasting battery safety and lifetime is largely an unsolved problem in the battery industry. For manufacturers, this uncertainty increases cell cost through control measures during production as well as the precautions taken to avoid warranty events. This project proposes "data science-as-a-service" for battery formation to address both issues. By streamlining the battery formation, test, and grading process, manufacturers benefit from reduced work-in-progress (WIP) inventory waiting for final inspection, reducing facility space requirement to store WIP cell, and reducing scrap rates and increasing manufacturing yields. The impact of these improvements will potentially enable wider spread adoption of electric vehicle applications, a major driver for battery demand.
This Small Business Innovation Research (SBIR) Phase I project focuses on developing information technology infrastructure and algorithms for the prediction of battery performance during cell production. By combining state-of-the-art machine learning techniques with data management and manufacturing execution systems, battery cell manufacturers will greatly reduce the cost to operate and manage cell formation and test - an environment which has been largely underserved for innovation. The proposed project objectives will be achieved through two developing battery classification and prediction machine learning algorithms to improve early detection of battery failures. Novel implementation of the proof-of-concept algorithms in battery production environments will improve the key performance indicators of these battery manufacturers. Regression and clustering models will be used as often as possible, and the bulk of the technical work will be dedicated to the feature engineering required to elucidate changes in the change and discharge voltage profile during the first few cycles. New features will be developed by a) modelling physical processes (e.g. growth of the solid-electrolyte interphase layer) expected for a given cluster group or b) employing dynamical systems techniques like time-delay embeddings.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
