En Jämförelse mellan OLS och Bayesiansk Regression:Modellering av Maximalt Gastryck i Innerballistik

This bachelor thesis investigates whether it is possible to reliably indicate how maximum gas pressure at elevated temperatures depends on measurable predictors, based on firing data collected at nominal temperature. Two regression approaches are compared: a classic frequentist multiple linear regression model employing Ordinary Least Squares (OLS) estimation with hypothesis testing and confidence intervals; and a Bayesian linear regression model that integrates prior information to yield full posterior distributions for parameters. The underlying data is collected from Saab Dynamics in Karlskoga, and are derived from internal ballistics, with the data collected from shootings, and include variables such as the number of rounds fired, charge weight, muzzle velocity, recoil (all measured at nominal temperature), and the observed maximum pressure at higher temperatures. Both regression models demonstrate that charge weight, muzzle velocity, and recoil at nominal temperature significantly influence the maximum pressure at elevated temperature.