On-Line runaway detection in batch reactors using chaos theory techniques

In this work nonlinear time-series analysis using delay coordinate embedding was applied to simulated temperature data from isoperibolic batch reactors to develop an early-warning detection system of the runaway. In the first part of this study an early-warning detection criterion, that is, when the divergence of the system becomes positive on a segment of the reaction path, was developed and compared with previous criteria. This criterion has the property of being preserved under phase-space reconstruction and, hence, it is possible to calculate it when only temperature data are available. This reconstructed divergence is compared with the theoretically calculated divergence using temperature and conversion data. The comparison demonstrates the validity of such an approach finding a new path, lying between the simple use of measured variables and the complex model-based reconstruction techniques, to assess in advance hazardous situations in chemical reactors.