Training for design of experiments

Abstract

Many industrial engineers perform one‐factor‐at‐a‐time (OFAT) experiments to examine situations of process improvement and for problem‐solving activities. However, OFAT experiments can prove to be inefficient and unreliable, leading to false optimal conditions. Moreover, they often consist largely of “trial and error”, relying on luck, intuition, guesswork and experience for their success. Design of experiments (DOE) takes an alternative, more structured approach. DOE is a powerful technique for discovering a set of process or design variables which are most important to the process/product/system and then assisting experimenters to determine at what levels these variables should be set/kept to optimise performance. In order to demonstrate the power of designed experiments over the traditional OFAT approach, the authors use a simple catapult experiment. They suggest that such an experiment could act as a powerful weapon in the training of engineers and managers who might be intimidated by a more “up front” statistical approach.