Process Optimization via Conventional Factorial Designs and Simulated Annealing on the Path of Steepest Ascent for a CSTR

This work determines the efficiency of sequential algorithms for automatic optimization of a chemical process. A method of steepest ascent and an integrated approach between the method of steepest ascent and Simulated Annealing, are compared on a simulated continuous stirred tank reactor (CSTR) with various levels of signal noise. The results suggest that the method of steepest ascent seems to be the most efficient on the CSTR surface at the lower levels of noise. However, the integrated approach with the Simulated Annealing element works well when the standard deviation of the noise is at higher levels. Although the average, the standard deviation of the greatest actual concentration of the product and percentage of sequences ended at the optimum from the integrated algorithm are better, it needs more runs, on average, to converge to the optimum when compared.