Case studies
Value Proposition
- Automated computation of numerical “fingerprints” for mixtures of solids, liquids, and additives
- One-click training and deployment of machine learning models
- Support for user-specified custom performance metrics
- Scalable platform for handling datasets of any size
- Models trained in the platform can be automatically applied to >200,000 candidate materials aggregated from multiple databases
- Candidates can be combined into a nearly infinite number of unique formulations
- Cloud-based data storage for instant recall of historical screening results
- Vendor and pricing information for candidate materials
- Immediately access and deploy the latest models as they’re published by the scientific community
- Save the time and effort required to reproduce models from the literature
- Add in new in-house data and re-train models to make them your own
- Optimal starting point for teams with minimal machine learning expertise
- Identification of the most important features driving performance
- Compare the relative importance of atomistic features versus processing steps
- Quantify the predictive power of your datasets
- Highlight areas of materials space where models are most and least reliable
Liquid Li-ion Electrolyte Optimization
Case Study
Liquid Li-ion Electrolyte Optimization
Optimizing Electrolytes for Cycle Life
Using the Aionics platform to train a model on performance data from 200 unique electrolyte formulations and screen thousands of new formulations, one Aionics customer was able to identify promising new formulations with a 10x acceleration over random guesswork.
Liquid Electrolyte Optimization for Non-Li Batteries
Case Study
Minimal Degradation Formulations
After spending over a year sampling over 80 difference electrolyte formulations looking for the one that would meet both performance metrics, one cell manufacturer began using the Aionics platform to assist in the discovery process. Three of the first ten new formulations suggested by Aionics scored as high performers on both performance metrics.
Liquid electrolyte formulation B
After spending over a year sampling over 80 difference electrolyte formulations looking for the one that would meet both performance metrics, one cell manufacturer began using the Aionics platform to assist in the discovery process. Three of the first ten new formulations suggested by Aionics scored as high performers on both performance metrics.
Solid Li Superionic Conductor Discovery
Case Study
Small Data, Big Improvement
Aionics Founder Dr. Austin Sendek's Ph.D. research at Stanford University focused on developing new data-driven methods for identifying solid lithium superionic conductor materials. Based on a training set of only 40 examples, Dr. Sendek and colleagues discovered a model capable of identifying new superionic conductor materials three times more effectively than guesswork and orders of magnitude faster than experiments or simulations.
Learn moreData-Driven Lifetime Prediction and Charging Optimization
Case Study
Predict Cycle Life in 10 Cycles
Leveraging a data-driven cycle life prediction algorithm inspired by Severson and Attia et al.’s work in Nature Energy, one battery manufacturer is now using the Aionics platform to predict the long-term cycle life of new cells after cycling them only ten times, representing a 10X+ acceleration in the otherwise time-consuming testing cycle. These models continue to improve through work at the Stanford StorageX initiative, co-directed by Aionics advisory board member Will Chueh.
Learn more
Leveraging a data-driven cycle life prediction algorithm inspired by Severson and Attia et al.’s work in Nature Energy, one battery manufacturer is now using the Aionics platform to predict the long-term cycle life of new cells after cycling them only ten times, representing a 10X+ acceleration in the otherwise time-consuming testing cycle. These models continue to improve through work at the Stanford StorageX initiative, co-directed by Aionics advisory board member Will Chueh.
Learn more