ML Database for low flammability polymers

We propose PolyFlamExMLDB, a highly modular and extensible ML and data lake system. It will leverage QVÆS existing proprietary ML experimentation and data storage platform (ML-X) developed during the Phase 2 CyberNeuro-RT DOE project. QVÆs ML-X, will provide a full suite of ML capabilities including modular & extensible support for custom data encoders, models, architectures, and training parameters, automated pipeline for encoding, feature selection, model selection, training, tuning, and testing support, and integration with Weights & Biases, MLFLow, and Tensorboard for model insights and experiment tracking.Ā Our tool will be usable in both a forward sense to predict ASTM-E1354 results including peak heat release rate, time to ignition, and smoke density for a given input composite such as a carbon fiber reinforced polymer and in an inverse manner to predict ideal composite constituent materials given desired flammability index values or ASTM-E1354 test results. As an SBIR, Quantum will develop, and test our advanced PMC discovery tool and models to determine performance, effectiveness for the target use case. We can additionally integrate our work into other branches of the Defense department as the tool becomes commercially viable.