A Classification Framework for Process Operation Optimization and its Application in a Triazophos Plant

R. Gang, H. Gu, B. Jin and J. Shao

Abstract
Knowledge-based operation optimization methods may suffer from difficulties in modeling the chemical processes and solving the mathematical equations. In this paper, a data-based classification method for operation optimization is introduced. In contrast with other fields, chemical process is characterized by time delay and interaction between upstream and downstream units. By rebuilding historical data and constructing a group of multiple classifiers, both of the characterized problems are overcome. Some qualitative operational advice may extract from the group of multiple classifiers. As a result, the operation of chemical processes may achieve to a reachable optimal state using rolling optimization strategy by updating the classifiers. In addition, some special data-preprocessing techniques are considered to improve the efficiency of the classification. This classification framework customized for chemical process helps a Triazophos plant to improve the productivity of Triazophos from 93.3 % to 95.8 % after implementation of the proposed method for more than one year.

This work is licensed under a Creative Commons Attribution 4.0 International License

Keywords
Operation optimization, process operational data, data preprocessing, data mining, classification