Abstract
This article presents a case study of a chemical compound acting as a subsystem of a delay mechanism for starting a rocket engine. The objective of this study was to investigate the proportions of mix components that enable a previously specified burning time. Thus, a linear regression model with normal responses was fitted, but later considered inadequate, as there was evidence that the response variance was not constant. Models fitted by the quasi-likelihood method were tried then. Through the developed model, it was possible to determine the proportion of each component to accomplish the process optimization. For the process optimization, besides considering a specific burning time, it was possible to consider the variance minimization for this time prediction as well.