Benchmark processes were specified in detail with respect to reaction conditions, composition of the feed material, reaction or treatment time, safety limitations etc. Using these defined specifications, hard and soft sensors and models were developed, optimisation frameworks were implemented and demonstrated in the benchmark processes.
Sensor developments
Hard sensor development for real-time process observation was a key objective of the project. Project partners developed temperature sensor for continuous steel melt temperature measurements, DynTemp®. The sensor consists of a consumable optical fibre which is immersed into the melt. Compared to a contactless optical measurement, this immensely increases the measurement accuracy as the sensor is not affected by slag. The sensor allows an in-line temperature measurement with higher accuracy and measurement dynamics, which was proven in a steel plant. Additionally, the sensor was modified and successfully tested for the silicon refining.
Dispersions from emulsion polymerisation have many applications in which particle size and morphology at the nano-scale are important parameters. Different sensor technologies were developed to get online information about these parameters. Though a use for industrial emulsion polymerization is too early for acoustic sensors and inline TEM measurements, the developed Raman technology was successfully implemented at lab and pilot scale.
Process model developments
One of the important work-packages of the project includes development of process models which are based on physical principles, i.e. on heat and mass balances coupled with thermodynamic relations describing, e.g. phase and chemical equilibria. The different models for the three industrial use cases have been fully developed and successfully tested.
• For polymerisation, process models were developed to predict and optimize the polymer structure, particle size and morphology based on real-time Raman and heat-balance measurements.
• For liquid steelmaking, a process model predicting the temperature evolution along the entire chain of batch processes was developed, validated with industrial online temperature measurements and integrated within model predictive and iterative learning control tools for process optimization.
• For silicon refining, a model predicting the dynamic development of the refining process with respect to the chemistry and temperature evolution in real-time was developed.
Demonstration of the developed tools for optimization and process control
The developed sensors and process models were to a large extend implemented in the industrial environments. For all use cases, online monitoring and non-linear model-predictive control were demonstrated successfully.
• For polymerization, it was demonstrated in lab and pilot scale that the morphology of the particles can be controlled at optimum batch time, energy and raw material consumption.
• For liquid steelmaking, the online working temperature sensor and real-time dynamic process models and control algorithms help to optimize the temperature evolution along the entire batch process chain, thus leading to energy and resource savings.
• For silicon refining, it was shown that the online model very well predicts the temperature evolution in the process. It is used online for real-time process monitoring and batch-to-batch corrections.
Exploitation and dissemination
Exploitable RECOBA results are new and updated real-time process control concepts for batch processes, new and better sensors, and extended knowledge on modelling techniques for batch and semi-batch processes. The new sensors and the real-time process control concepts will be offered to the market, and the industrial use case representatives plan to make use of the models and online control techniques to increase the efficiency of their processes and improve raw material and energy consumption and to be able to develop new and improved products. In the second half of the project, the RECOBA partners participated in or organized more than ten events in which the results of RECOBA were disseminated. One patent was issued, 8 journal papers and 19 conference papers were published, and teaching and training was organized in the consortium.
