Abstract/Summary

Polymer production is of paramount importance in the chemical manufacturing industry. However, safety concerns are prevalent due to the exothermic nature of polymerization reactions, which can cause thermal runaway. The limitations of the current industry-standard monitoring methods underscore the need for novel techniques to detect faults early. To facilitate the development and evaluation of such algorithms, benchmarks that enable direct comparisons of performance are required. Addressing this gap, the present work first introduces an open-source polymerization benchmark model and associated datasets. Derived from reaction kinetics, mass balance, and energy balance analysis, the differential equation forms the basis of our model. By manipulating relative parameters, we intentionally induce five typical faults that can lead to thermal runaway. As a result, our benchmark model serves as an invaluable tool for advancing and validating algorithms for thermal runaway process monitoring, significantly enhancing the safety of the polymerization process. The effectiveness of the model and dataset is demonstrated by testing multivariate statistical process monitoring algorithms and deep learning algorithms.