Zhou and Matias develop safer control for exothermic batch reactors
New ARC method achieves 0% temperature violations where NMPC fails in fault scenarios
This paper tackles a critical challenge in chemical process control: safely maximizing productivity in cooling-limited exothermic semi-batch reactors. Traditional industrial ARC relies on heuristic signal selection and tuning, while nonlinear model predictive control (NMPC) demands maintained models and online optimization. The authors develop a systematic analysis-to-architecture workflow that merges finite-horizon minimum-time optimality (for productivity) with local safety analysis (for thermal runaway prevention). Under stated assumptions, they translate boundary-seeking optimality into a cooling-demand valve-position-control (VPC) architecture and safety requirements into near-boundary tuning rules.
Testing on a reduced benchmark and an industrial-scale polymerization example, the proposed ARC performed competitively with nominal-model output-feedback NMPC using an extended Kalman filter. Crucially, in adverse scenarios—parameter mismatches and unmodeled faults—ARC maintained 0% temperature-limit violations, while NMPC either exceeded limits or failed to complete the batch. This demonstrates that systematic, theory-guided ARC can offer robust safety without the computational overhead of full NMPC, making it attractive for industrial deployment where model maintenance and online optimization are impractical.
- Combines finite-horizon minimum-time optimality with local safety analysis for systematic ARC design
- Achieved 0% temperature-limit violations in fault scenarios vs. NMPC failures or incomplete batches
- Tested on industrial-scale polymerization benchmark, showing practical deployability
Why It Matters
Safer, more robust control for chemical reactors without heavy computational models—reducing accident risks in manufacturing.