Golden-setpoint quality prediction for a superabsorber plant
A feasibility study to predict a key quality indicator from process data at the moment the lab sample is taken, as the base for recommending setpoints that hit the target.
Challenge
A superabsorber producer's quality indicator (centrifuge retention capacity) is known only when the lab result comes back. The plant wanted to predict it from process data and, from there, the process parameters that give the desired result.
Solution
Under a feasibility order running from November 2025 to January 2026, Arundo reviews two years of batch process and lab data, reports on data quality and stable-production periods, tests regression approaches that map production parameters and engineered features onto the quality measurement and its uncertainty, assesses process time lags, and proposes how to feed the model live on Arundo Foundation; a second phase deploys the model online with an interface comparing predictions to measured values.
Impact
A model-feasibility and performance report, and the architecture for online prediction; the optimisation phase that recommends setpoints follows a go decision.