Data-based modeling and control of nonlinear process systems using sparse identification: An overview of recent results

This research project looks at using Sparse Identification of Nonlinear Dynamics (SINDy) to build data‑driven models and controllers for complex, nonlinear chemical processes. Industrial plants often have processes that run on very different time scales and that suffer from noisy sensor data. To handle this, the method combines modern system‑identification tools with robust control techniques. For systems with both fast and slow dynamics, we first create a reduced‑order model. SINDy learns the slow dynamics, while nonlinear principal component analysis (NLPCA) algebraically links the fast states to the slow ones. This coupling reduces model stiffness and avoids ill‑conditioned controllers. The reduced‑order model is then placed inside a Lyapunov‑based model predictive control

(LMPC) scheme. The LMPC guarantees closed‑loop stability as long as the fast and slow dynamics are sufficiently separated. To cope with noisy measurements, we apply subsampling and denoising before running SINDy, which gives more reliable parameter estimates. The project also provides step‑by‑step guidance for simulating these data‑driven modeling and control strategies. It shows how to evaluate controller performance, state‑estimation accuracy, and closed‑loop stability in multi‑scale chemical processes.

Message: Hi FE, I need help with “Data-based modeling and control of nonlinear process systems using sparse identification: An overview of recent results” in “Chemical Engineering”