Evaluation of "parameter design" methodologies for the design of chemical processing units.

Abstract

'Parameter design' is a procedure to identify settings of design variables that minimize variation in performance of a processing unit. However their application to the design of chemical processing units has not been widely reported. The objective of this study was to investigate the feasibility of using 'parameter design' methodologies to determine values of the design variables for a non-isothermal Continuous Stirred Tank Reactor which minimized the sensitivity of product quality to disturbances. The relative merits of each approach and the potential for application of 'parameter design' to chemical unit design are discussed. In summary, Response Surface Methodology associated with Monte Carlo simulations was found to be very efficient. Error Transmission Analysis did not perform well because of the inaccuracy of the first-order approximation adopted in the transmitted variation function. The use of the Taguchi Method was not always justified due to the universal logarithmic transformation, and the use of the marginal means. Sequential Elimination of Levels was found to be generally unreliable. (Abstract shortened by UMI.)