Obtaining the kinetic data for some systems can be difficult for various reasons, but a major concern is the need for chemical analysis. In most cases, a chemical analysis probe (eg. IR) is not suitable because of the chemical compounds or the analytical speed required. One option is taking samples into a small cylinder with a kill agent to stop the reaction. A manifold of such cylinders can be used to obtain several samples during a batch run. The analysis is then conducted after the run is complete.
I have developed an alternative method that appears promising for a number of exothermic systems. The method uses a semibatch, bubble column reactor. The only data needed to determine the kinetic parameters is the flow rate of the gas used to control the reactor temperature. This gas may be a reactant or not.
In the example studied (via computer simulation), the five parameters for the five-step series hydrogen of glucose to hexanediol were obtained exactly using an order-of-magnitude initial guess. In the procedure, the glucose aqueous solution is charged to the reactor, and the hydrogen is started with manual control. After the reactor temperature exceeds the desired reaction temperature, the temperature controller is turned on. The controller adjusts the hydrogen flow rate to maintain the partial pressure of the boiling water. The catalyst is charged (at 30 min) to initiate the reactions.
The blue curve shows that the reactor temperature was controlled within 2 C of the desired 155 C during the reaction. The red curve is the only data required for the parameter estimation. The Minerr optimization routine (Levenberg-Marquardt option) in Mathcad 15 was used to estimate the parameters.
The figure above shows the liquid composition during the run. Had a chemical analysis method been used, it would have been a daunting task to obtain and analyze the samples needed to capture the changes in composition.
The main requirements for this method are as follows:
The presence of other boiling components adds some complications that will change the method, but this will be covered in a future blog. Also to be covered later will be the effects of model errors on the parameter estimation.
Please contact me if you wish assistance in implementing this technique.