Detection of stealthy outliers in thermal-kinetics modeling

A method is presented to uncover outliers in data from small-scale process-safety studies. Points along an adiabatic runaway reaction, especially at the onset and conclusion of the experiment, may be vaguely jittery and difficult to recognize as outliers. Even though they fit well visually, their stealthy misalignment may have a profound impact on the estimate of kinetic parameters, such as reaction order, pre-exponential factor and activation energy. The proposed technique combines temperature and its time derivatives to magnify the effect of hidden outliers. Once identified, these points can be excluded from regression, in order to generate kinetic parameters with minimal distortions. This method is also suited to recognize autocatalysis.