"All models are wrong but some are useful"
— G.E.P. Box, "Robustness in the strategy of scientific model building." Robustness in statistics. Academic Press, 1979

The fundamental question of how to leverage available process data – i.e., measurements – to improve system operation is much older than the most recent shift of research foci towards AI and data science. Indeed, in systems and control this topic has been discussed under various names and labels, ranging from iterative learning control (mechatronics) via real-time optimization and measurement-based optimization (process systems engineering) to data-driven control.

In this context, the group has substantial expertise on data-driven methods for run-to-run optimization of dynamic systems. This includes method design and implementation alike.

Selected publications:

• Shukla, Harsh A., Tafarel de Avila Ferreira, Timm Faulwasser, Dominique Bonvin, and Colin N. Jones. "Convergence Certificate for Stochastic Derivative-Free Trust-Region Methods based on Gaussian Processes." arXiv preprint arXiv:2010.01120 (2020).
• Milosavljevic, P., Marchetti, A. G., Cortinovis, A., Faulwasser, T., Mercangöz, M., & Bonvin, D. (2020). Real-time optimization of load sharing for gas compressors in the presence of uncertainty. Applied Energy, 272, 114883.
• de Avila Ferreira, T., Wuillemin, Z., Faulwasser, T., Salzmann, C., & Bonvin, D. (2019). Enforcing optimal operation in solid-oxide fuel-cell systems. Energy, 181, 281-293.
• Faulwasser, T., & Pannocchia, G. (2019). Toward a Unifying Framework Blending Real-Time Optimization and Economic Model Predictive Control. Industrial & Engineering Chemistry Research, 58(30), 13583-13598.
• Marchetti, A. G., François, G., Faulwasser, T., & Bonvin, D. (2016). Modifier adaptation for real-time optimization—methods and applications. Processes, 4(4), 55.