Distillation yield from simulation-recommended setpoints
Replaying 2.5 years of campaigns through the site's simulator found about 4 percentage points of average yield headroom — worth around EUR 1.3 million a year.
Challenge
A European re-distiller produces high-purity solvents in campaigns, with by-products sold near naphtha price, and yields varied widely. Operators steered on purity with no view of achievable yield, adjusting setpoints in small steps paced by four-hourly lab tests; flowsheets used an estimated feed while the real feed shifted; lab results lived in spreadsheets.
Solution
Arundo proved a simulate-recommend-adjust loop on 2.5 years of plant data: about 940 lab samples from 39 campaigns replayed in simulation, predicting iso-octane purity within 0.14–0.16 percentage points, and setpoint sweeps on the actual feed composition to find the highest-yield point within purity and equipment limits. The agreed architecture feeds PI and LIMS data through Arundo Foundation to a control-room app that calls the simulator machine to machine, with the shift leader making the decision.
Impact
About 4 percentage points of average yield headroom (about 8 on iso-octane), worth around EUR 1.3 million a year against an investment of about EUR 300 thousand.