Research Symposium

BIO

Hi! My name is Josh Briley and I am from Palm Bay, FL. Outside of conducting research I often enjoy practicing my trumpet or going for a run. Last semester I was lucky enough to be a part of the Marching Chiefs and perform at every home football game!

I hope to one day be able to work in data science or ML technology in tandem with performing and teaching music.

Predicting Fire Dynamics Using a Convolutional Neural Network

Abstract

This project seeks to create a Machine Learning (ML) model using a Convolutional Neural Network that can accurately predict fire spread and fire dynamics. A physical model to predict fire dynamics was created on MATLAB as a control/comparison to the ML model that was created and run through a Python Script and TensorFlow. Datasets were created using the MATLAB model that was then submitted to the High-Performance Computing center to create 6,000 datasets. These datasets were then divided into 85% to be used for training the ML model, and 15% being reserved for testing the model. While testing and alterations to the model are ongoing, it has been proven that (to a certain degree of accuracy) a ML model can predict fire dynamics and spread. Therefore, with this knowledge it is reasonable to assume that a further developed ML model could more accurately predict fire dynamics on a large-scale, aiding in decisions on evacuation and fire extinguishing plans.

Keywords: Machine Learning Fire Dynamics