Computer Modeling as an Aid to the Analysis of Kinetic Data

Abstract

In studies of the eventual fate of toxic chemicals in the environment, it is common to acquire data on the concentration of chemical species of interest as a function of time. A mechanism is often postulated to explain the observed data. To test the proposed mechanism, one must cast the individual reaction steps into differential equations and integrate. In general, the proposed kinetic processes are nonlinear, and no closed analytical expressions are available. Thus, the only way to fit the observed data is to use numerical integration techniques. This paper describes the development of a computer program based on a modified Marquardt algorithm in conjunction with a least-squares differential corrector to estimate the required states and parameters. Actual data from the decay of hydrazine vapor in an environmental chamber are used to test four proposed loss mechanisms. The use of error residuals as a test of the model fit is demonstrated. The importance of sampling time regularity and integration step size is also shown.