A Novel Approach for the Autonomous Control of Fluidized Beds with Intelligent Digital Twins Based on Transient Flowsheet Simulations

Fluidized beds have been industrially established for decades. However, the process monitoring and control of such production plants are continuously confronted with new challenges to avoid undesirable process conditions and faulty production. Therefore, new control strategies which increase the efficiency of fluidized beds are of high importance. Such strategies can be provided by intelligent digital twins. A new framework for intelligent digital twins of fluidized beds, which is based on transient flowsheet simulations, is currently under development and is presented in this contribution.

The core of each digital twin is a so-called knowledge module which is a data processing unit for measurements and simulations. First, the knowledge module consolidates all online measurements of the sensors in the plant and detects possible outliers in the data. This data then serves as input for the transient flowsheet simulations. Second, a sensitivity study of the flowsheet simulations based on the input data and available control parameters is set up and is executed in parallel computation and significantly faster than real-time. Lastly, the simulation results are validated against the most recent measurement data. This procedure enables the knowledge module to accurately predict the future of the process and to calculate the necessary changes in the control parameters to ensure the desired process behavior.

The transient flowsheet simulations are carried out in the open source software Dyssol, which allows the implementation of any process and process chains in the field of solids process engineering [1]. For Instance, the data processing and consolidation of simulation results by the digital twin are shown for a lab-scale process of fluidized bed spray granulation.

[1] Skorych, V., Dosta, M., Heinrich, S.: Dyssol – an open-source flowsheet simulation framework for particulate materials, SoftwareX, Volume 12 (2020), 100572, DOI: 10.1016/j.softx.2020.100572