Tapered Grain Geometry and Statistical Learning for Solid Rocket Motor Simulation

Abstract

This thesis investigates applying statistical learning techniques to a tapered grain solid rocket motor simulation. Tapered grain solid rocket motors (SRMs) have application in both defense and space industries. Tapered grain geometries offer an alternative to complex cross sections to control the thrust profile of the solid rocket motor. New analytical methods were developed to accurately model tapered solid rocket motor grain geometries. A tapered grain solid rocket motor internal ballistics code was developed in FORTRAN using Lagrangian grain regression assumptions, 1D flow assumptions, and new analytical methods developed as part of this work. This code can accurately model the internal ballistics of tapered grain motors, specifically for circular perforated and star grain geometries. This thesis will explore the development of analytical equations for tapered grains, the implementation into a code, and accompanying machine learning techniques and results. The SRM internal ballistics code was used to develop large databases for statistical learning. The SRM code contains a Monte Carlo simulation using a Latin Hypercube distribution that allows the user to robustly generate a multitude of SRM designs, and the resultant thrust-time profile based on desired inputs. Once large databases of performance data were generated, statistical learning methods such as regression analysis and neural networks were used to provide regression analysis and surrogate modeling capabilities.