(Texas Tech University, 1965-08) Carradine, William Radell

Show more

This investigation was concerned with the fitting of Hougen and Watson type catalytic reaction rate equations to adiabatic data. A particular equation corresponding to the case in which the surface reaction is the rate controlling step was chosen, and data were mathematically generated there from. Steepest descent methods were used to fit the equation (in the least-squares sense) to the data.
It was found that the sum-of-squares response surface for this particular equation is quite rough. The steepest descent search for the correct set of parameters led to a number of sets depending on the starting point, A parameter perturbation search was proposed which aided the steepest descent: search.
Even though error free data were used, it was found that this combined search procedure is not robust enough to accurately determine parameter values without considerable previous knowledge of their values. It can be used to obtain a data fir over a relatively wide range of experimental conditions such that the average error in the calculated rates is about fifty percent.
The combined search was also applied to a set of data from the literature. The classical linearization technique used by the original experimenters provided a solution which resulted in an average error of thirty-five percent. From a starting point corresponding to the final linearized solution the combined search technique was used to reduce the average error to nineteen percent. When other starting points were used, it let to solutions such that the error was about the same as that obtained by the linearization procedure.