Online model selection approach based on Unscented Kalman Filtering

► Successful model selection requires the design of experiments providing informative data to distinguish between models. ► Optimal design of experiments is usually done offline. In this work, an online approach is presented. ► The method is demonstrated for a biochemical system but may be applied to a wide range of technical processes as well. ► Three different test case scenarios are analyzed going from very idealized assumptions up to a “true-to-life” case.