Academic Awards 2023 booklet

85 Data-driven modeling and complexity reduction for nonlinear systems with stability guarantees How can the dynamical behavior of complex engineering systems be captured in compact mathematical models? This is a central question that needs to be answered for the development and operation of tomorrow’s engineering systems. Fahim Shakib’s PhD research provides novel fundamental methods for modeling such engineering systems directly from the increasingly available experimental data and for reducing the complexity of these models. A key feature of the methods developed is the guaranteed model stability during the model construction and complexity reduction. As a result, and in contrast to state-of-the-art methods, these methods provide robustness for safe model use, even in scenarios that were not captured by the data. In addition, the novel methods allow for a drastic reduction in the time required for model construction and complexity reduction. The developed methods have been demonstrated in a study on respiratory devices used for breathing support in hospitals. It has been shown that accurate patient and device characteristics are automatically identified from experimental data in a matter of seconds. This information can be used to improve patient treatment and will save valuable clinician time. Fahim is currently a postdoctoral researcher at the Imperial College London, where he is developing complexity reduction methods. These methods allow complex dynamical system behavior to be captured using simple and compact models. The methods developed allow for simple and compact dynamic models that accurately capture the system behavior. These models are found in a fast way from experimental data and are robust to scenarios that were not captured by the data. The developed methods are applied to respiratory devices used to support breathing in hospitals. The figure shows that the obtained model accurately predicts the patient’s airway pressure using the data-driven model.