Development of an intelligent digital twin for prediction and control of the process flow by means of transient flowsheet simulation using the example of fluidized bed spray granulation
Xiye Zhou, M.Sc. and Dipl.-Ing. Robert Kräuter
Motivation
In order to minimize unwanted process status and defect production, process monitoring and control are critical in industrial particulate production. In this trilateral project involving TUHH SPE, Fraunhofer IFF and Pergande Group, an intelligent digital twin model should be developed and tested for a fluidized bed spray granulation process, with the goal of improving process engineering efficiency by predicting future process behavior and ensuring reliable process control. The structure of the digital twin is schematically shown in Figure 1. Based on the dynamic model for the fluidized bed spray granulation implemented in Dyssol, the corresponding knowledge module is established, which interacts with the communication interface for data exchange and adjustment. Afterwards, an optimization algorithm is developed to automatically determine the setpoint value. Based on the algorithm and the simulation parameters, recommendations for operation can be integrated into the process control system and provided to operators.
TUHH SPE is in charge of modelling the fluidized bed spray granulation (subproject A) and development of the knowledge module (subproject B) based on the lab-scale fluidized bed granulators.
Subproject A: Dynamic flowsheet simulation of the fluidized bed granulation
The dynamic process of the fluidized bed spray granulation is studied experimentally and implemented in the open-source flowsheet simulation software Dyssol. The Dyssol model serves further for subproject B.
Methodology
- A model of batch fluidized bed dryer without water spray is implemented in Dyssol based on a previous work.
- Experiments are carried out for the batch fluidized bed granulation with bottom spray of water. Process parameters, such as water flow rate, temperature and velocity of fluidizing gas, are varied to examine the resulting bed temperature.
- Inline sensors are installed in the granulator. Experiments are carried out with regard to varied process parameters. The sensor calibration is carried out between the online and offline measurements.
- A multi-zone model for granulator is implemented in Dyssol based on the experiments with dynamic change of process parameters. The heat and mass transfer, the particle growth, breakage and attrition are considered in the model.
- Simulations in Dyssol are carried out for the experiments in step 3 with varied process parameters.
- Sensitivity analysis regarding to the dynamic change of granulation process based on the experiments in step 4.
- Simulation of step 6 is carried out in Dyssol considering the sensor calibration function in step 3.
Subproject B: Development of the knowledge module
The knowledge module executes the Dyssol models from subproject A with the current process parameters and uses the simulation results to develop control strategies for the fluidized bed.
Methodology
- Definition of the knowledge module and implementation of the communication with Dyssol.
- Implementation of the communication with the fluidized bed and integration of calibration functions and soft-sensors.
- Definition and implementation of heuristics for the drying potential of the fluidization gas, the balances of water and mass and the pressure drop.
- Definition of error functions for the comparison of simulation results and process measurements.
- Development of control strategies based on the heuristics and error functions.
- Testing of control strategies on the experimental data of subproject A.
Project funding and starting date
Beginning from February 2022, this project is funded by the German Research Foundation (DFG).
Cooperation partners
- Fraunhofer Institute for Factory Operation and Automation (Fraunhofer IFF)
- Pergande Group