摘要：AbstractContinuous fluidized bed spray granulation processes are known to be prone to non-linear oscillations, which requires a stabilizing controller design. In this article, a purely data-driven Koopman-based system linearization and control is proposed, which does not need any prior model knowledge and uses only systems input-output measurements. The linearization achieved by the Koopman embedding allows for linear-quadratic control design, which does not depend on the operating point. The proposed method was validated in a simulation study of a fluidized bed granulation process and was able to stabilize the particle size distribution in different operating points.
关键词：KeywordsGranulationfluidized bedneural networksKoopman embeddingdata-based control
