Analysis of optimization methods on chemical processes with high noise level

This study explores optimization techniques for processes with significant variation, or noise. Classical techniques are not effective when the response improvementto-noise ratio is small. In this realm of optimization, statistical techniques are necessary to achieve optimal conditions. Statistical methods such as the Student's t-test and Evop (Evolutionary Operation) were studied extensively.

The utility of optimization of processes with high noise was illustrated by performing statistical optimization methods on a furfural manufacturing process. Furfural manufacture had a high variability due to wide fluctuations in raw material quality, raw material type, and process economics. The process response as yield was increased significantly by surfactant addition to the reactors and by changing the catalyst concentration. Optimization of the furfural process should be a continuous one due to the dynamic response effect.