Community Research and Development Information Service - CORDIS

H2020

RECOBA Report Summary

Project ID: 636820
Funded under: H2020-EU.2.1.5.3.

Periodic Reporting for period 1 - RECOBA (Cross-sectorial real-time sensing, advanced control and optimisation of batch processes saving energy and raw materials)

Reporting period: 2015-01-01 to 2016-06-30

Summary of the context and overall objectives of the project

The RECOBA project is a cooperation project of ten European partners working towards developing and deploying robust real-time model predictive control and sensing methodologies. Through their application the energy and resource efficiencies of the considered (semi-)batch processes will be increased and a consistent final product quality will be achieved. The efficiency improvements will enhance the competitiveness of a significant portion of the European batch process industry.

In the RECOBA project the considered example processes are (i) emulsion polymerization of vinyl monomers where the products are used, e.g., in painting applications, (ii) liquid steelmaking and (iii) silicon refining. The partners focus on three different material systems to demonstrate the cross-sectorial applicability of developed sensors, optimization and control methods. While the goals are similar for all three mentioned processes (higher efficiency) the challenges are very different as the physical process conditions differ significantly as well as product properties and utilization. The challenges and objectives are schematically summarized in Fig. 1. The proof of beneficial application of control methods provided by RECOBA’s industrial partners should encourage other European process industries to deploy common advanced sensing and control techniques also in their applications and processes to improve their efficiency.

Let us here introduce particular goals and challenges of the considered processes in detail.

Polymerization process
Polymerization reactions are nowadays operated based on a repetition of a fixed schedule defined in a recipe. Impurities in the reaction mixture or fouling on the reactor wall are common disturbances during the batch run. In turn, the real process conditions can differ from the ones prescribed by the recipe. This can thus lead to variations in a final product quality. The goal of the RECOBA project is to replace the current fixed recipe control by model predictive control (MPC). Such MPC controllers can adjust the process control variables (e.g., temperature, reactant feed rates, etc.) in a real time in order to follow the optimal process trajectory. Therefore, the desired product properties are maintained while minimal energy consumption or batch time in presence of disturbances are achieved. Due to good reaction control the use of resources, which have often varying quality, e.g., bio-based resources, can become realistic for production of products with constant quality. These quality parameters include copolymer composition, molecular architecture of polymer chains and morphology of polymer latex particles. To enable the use of the MPC technology for the polymerization case, it is necessary to develop precise process and morphology models. Together with newly developed sensors, it is possible to gain information about the current states of the reaction mixture and apply suitable control actions. The polymerization process model is being developed by the University of Chemistry and Technology Prague (CZ) and the latex particle morphology model is being developed by the University of the Basque Country (ES). Development and testing of sensors and sensor data processing are carried out in close cooperation between the University of Cambridge (UK), University of the Basque Country and RWTH Aachen University (DE). Process operations optimization and online model predictive control solutions are being implemented by RWTH Aachen University and Cybernetica (NO). The developed concepts will be tested in the pilot plant facilities of BASF SE (DE).

Steelmaking process
The process of steelmaking is characterised by a large throughput of resources and energy. It is performed along a chain of different batch processes performed consisting of several aggregates for heating and metallurgical refining of the liquid steel melt. The objective of the production process is to treat batches of about 260 t of liquid steel with narrow quality targets regarding temperature and chemical composition, and to deliver these batches at a predefined time to a continuous casting plant. Real-time process control of the steelmaking process along the chain of several batch processes is presently limited to spot measurements for melt temperature and composition, performed during and after treatment at the different aggregates. Continuous inline measurements of these important process state variables are currently not available. Due to the lack of detailed analytical models, process control does normally not take into account the interactions and interdependencies between the different batch processes of the complete chain and the thermal conditions of the reactors (steel ladles and degassing vessels) in a predictive manner. Thus control strategies for temperature and composition adjustment are often not well balanced, and exceptional process conditions, which are most critical for meeting the narrow target window for casting, cannot be considered in a predictive way.

Within the RECOBA project the real-time control of the batch process chain for liquid steelmaking with focus on the steel melt temperature shall be enhanced by application of innovative sensor techniques, predictive process models as well as for process control and optimization. Fibre optical temperature sensors with an extremely short response time, high precision and continuous operation for several minutes will allow an in-line measurement of the liquid steel temperature. Detailed process models considering thermal states of the reactors will allow an accurate real-time monitoring and will predict liquid steel temperature evolution along the whole batch process chain of liquid steelmaking. For process optimization and control the process models will be combined with innovative control tools as non-linear state estimation, MPC and iterative learning techniques, to ensure an energy and resource efficient achievement of the narrow target temperature window at the end of the batch process chain.

The sensor development is performed by MINKON Poland (PL) with support of VDEh-Betriebsforschungsinstitut (BFI) (DE). BFI is also responsible for the development and application of process models as well as control techniques. The developed sensors, process models and control concepts shall be implemented and tested at the industrial facilities of one of the ThyssenKrupp Steel Europe steel plants (DE).

Silicon refining process
The current methods and procedures for producing refined silicon is in large based on relatively old technology and empirical data. There are relatively few process parameters being measured, the most important being temperature and chemical composition. The current method is acceptable when the input conditions or product specifications do not change, however, in today`s market there is a growing trend to make speciality products designed to fit the specific need of each customer. This requires a great deal of flexibility at the producer`s end. As such, we need to develop process control systems that can both deal with changing product specifications, but also with changes in initial conditions such as raw material composition.

The basis for a new process control system is a complete time-dependent mass/heat balance model for the entire refining ladle, including refractory walls and molten silicon. Elkem is developing an offline model for educational and analytical purposes, while Cybernetica is implementing the fundamental physics of the model into an online model predictive control system, with a generic structure similar to that of polymerization.

In addition to model development we also aim at integrating state-of-the-art sensor technology to produce robust measurements of critical state variables of the refining process. The integration of sensors into existing refining infrastructure will yield both valuable input/corrections to the model, as well as enhanced safety for operators. The joint effort of Elkem, Minkon and BFI is vital for the successful implementation of new sensor technology.

The expected outcome of the project is a significantly more robust process control system that maximizes the yield of silicon given the initial composition, the temperature during the process and the product specifications. It is necessary to both verify sensor technology in our process as well as developing accurate and efficient models for silicon refining.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

In summary, the project has progressed according to project plan with meeting of all milestones.

Emulsion polymerization process
Regarding the polymerization process significant achievements include:
• Process model of emulsion copolymerization including Monte Carlo simulation for detailed description of polymer chain architecture has been developed and validated with experimental data.
• Model for simulation of latex particle morphology evolution has been developed. Large sets of experimental data for the estimation of morphology model parameters are being collected.
• Different types of the low-cost acoustic sensors for latex particle characterization have been manufactured and tested. The evaluation of sensor data is still ongoing.
• Flow cell for online TEM microscopy (transmission electron microscopy) for monitoring of latex particle morphology has been developed, manufactured and is recently tested. The potential of the method for online monitoring of latex particle morphology will be evaluated based on future experiments.
• Raman spectroscopy has been tested for monitoring of concentrations of considered monomers and latex particle sizes in the reaction mixture.
• The developed process and morphology models are being introduced into the optimization and control software enabling model predictive process control.

The experimental focus in the first project period was mainly on the preparation and analysis of samples which went hand in hand with the development of sensors for process monitoring. The researchers have synthesized samples with different particle morphologies, cf. Fig. 2, to enable testing of various sensing concepts which seem to be promising for online characterization. Acoustic and online transmission electron microscopy sensors were tested on their responses regarding the latex particle morphology. Although the TEM technique provides directly visual results, their automatic evaluation is challenging due to low contrast between soft and hard polymer phases. As the first TEM results are promising the development of the online TEM method will continue also in the second project period. In case of acoustic sensors we have proven that there is a difference in sensor responses on samples with different morphologies, nevertheless the work on processing of sensor signal into morphology information is not finished yet and will continue as well. Further, Raman spectroscopy was tested for online determination of level of residual monomer concentrations in the reaction mixture. The challenge, which still remains, is in the determination of the very low monomer concentrations in case of the so called starved feed polymerization. The results and collected experience regarding sensors for the polymerization process case are described in the public deliverables which can be found on the official RECOBA website (deliverables 3.3 and 3.4)

Modeling efforts in the first project period were focused on (i) the development of accurate process model coupled with Monte Carlo simulation describing the evolution of monomer and polymer concentrations and polymer molecular weight distribution and on (ii) development of polymer morphology evolution model. While the process model was validated with the experimental data, the collection of experimental data necessary for parameter estimation and model validation is in the case of the morphology model still ongoing. Both models meet computational requirements for the online use. They are being implemented into the optimization and control software to enable the process intensification and control within the next project period. Further, within the project, a model involving particle interactions and their effect on latex viscosity or coagulum formation is being developed. The model results show good agreement with experimentally obtained data and brings thus insight into polymer latex flow phenomena which is crucial for prevention of fouling or coagulum formation.

Liquid steelmaking process
Regarding the liquid steelmaking process significant achievements in following fields were obtained:
• A prototype of fiber optical sensor for online monitoring of liquid steel temperature (DynTemp®) was built and successfully tested in a pilot plant environment.
• A coverage efficiency sensor resistant to a harsh environment has been built up and successfully tested for monitoring of temperature loss during steel treatment.
• A high-temperature sensor for refractory temperature monitoring (Fiber-Bragg sensor) has been developed and tested in lab. The tests provided conditions under which the sensor can be operated in the long term.
• The DynTemp® system was installed in the industrial plant environment in the steel plant and successfully tested in the operation.
• A process model describing in detail the process state (temperature and composition) evolution of the steel melt throughout the chain of the steelmaking batch processes was developed.
• Process data of more than 100 industrial heats were collected to enable process model parameter estimation and model validation.
• A model predictive/iterative learning control concept for online temperature control of the liquid steel production was developed and tested.

The developments in the sensor field in the first project period were focused on (i) the set-up of the DynTemp® based sensor for in-situ monitoring of liquid steel temperature, (ii) on testing of the coverage efficiency sensor for monitoring of temperature losses during treatment of the steel and (iii) on testing of fiber Bragg sensors for monitoring refractory temperatures. The basic functionality description as well as the first experience with all the above mentioned sensors are described in the public deliverables available on the RECOBA official website (deliverables 3.7, 3.8 and 3.9). The data from the sensors will serve for model validation and for closing a control loop regarding temperature during liquid steelmaking treatment.

The modeling work was focused on the development of a detailed process model describing the evolution of the liquid steel state during its treatment, with focus on the temperature. The model contains a detailed description of the relevant metallurgical reactions coupled with thermodynamic relations describing chemical equilibria, and considers the thermal state of the refractory-lined reactors (steel ladle and vessel for vacuum treatment). The developed model is capable to fit the process data with sufficient accuracy and the calculation time is also suitable to use the model for online control. The model is now being implemented into the developed control environment for process state estimation, prediction and optimization.

Silicon refining process
Regarding the silicon refining process between significant achievements belong:
• Detailed process model of silicon refining was developed and validated with process data. The model describes evolution of temperature and chemical composition of the reaction mixture. Model involves chemical kinetics and thermodynamic equilibria of all key components.
• A key measurement and data transmission methods for real-time determination of refractory mass were identified and tested.
• Methods for temperature measurements applied for the steelmaking process are identified as applicable also for the silicon refining process.

The main modeling effort in the first project period was aimed at the development of the process model suitable for online application, i.e., dynamic model which can be simulated in order of few seconds. Such model was developed and includes the detailed heat balance connected with the description of proceeding chemical reactions and chemical equilibria. The process model was validated with accessible process data. Further, a model describing heat state and status of fouling of the refractory ladle between silicon refining batch runs was developed. The detailed knowledge about refractory ladle heat state and thickness of the formed silicon “scull” on the ladle wall can bring significant savings of energy needed for heating of the reacting melt as well as savings due to longer lifetime of the ladles.

The challenges on the sensor field are similar to the steelmaking processes. Therefore, Elkem will test the in the project developed sensors also in the silicon refining case. The knowledge about the temperature evolution during the refining is crucial for achievements of energy savings as well.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

At the project end the partners will demonstrate the utilization of the model predictive control methodology in production of emulsion polymers, steel and silicon, respectively. To enable this, partners will work on development and utilization of:
(i) innovative process models
a. of emulsion polymerization including reaction kinetics and Monte Carlo simulation for description of polymer’s molecular architecture,
b. model for description of latex particle morphology development,
c. detailed model of the chain of batch processes for melt temperature in liquid steelmaking, including reaction kinetics and important thermodynamic phenomena,
d. detailed model of silicon refining process including reaction kinetics and important thermodynamic phenomena,
(ii) and sensors and data evaluation methods
a. for monitoring of latex particle morphology (flow TEM, acoustic sensors),
b. for monitoring of reactant concentrations in the reaction mixture,
c. temperature monitoring sensors applicable at high temperatures and harsh environment (fiber optical sensor, coverage efficiency sensor, fiber Bragg sensor).
As the project progressed excellently the involved partners are already working on the demonstration of the developed concepts and sensors in the plant environment.

The usage of MPC together with the state of the art sensors will enhance the energy-, resource- and cost-efficiency of the respective semi-batch processes. To quantify the project results the industrial partners will compare standard process control methods with the new ones developed and implemented within RECOBA and discuss environmental, economic and social impacts of MPC usage for production. It has to be noted that the intention of the project is not just to bring financial benefit to the technology end users (BASF, ThyssenKrupp, Elkem) but also to suppliers of software and modeling tools and sensor developers and manufacturers. From the perspective of SMEs involved in the project it is an opportunity to get experienced in the new application fields and exploit their technologies to broader customer portfolio. With the introduction of new control concepts and sensor technology new businesses along the value chain will arise. The complexity of the control concept will also lead to the need for training, maintenance as well as performance monitoring and optimization services. This business can be generated by the involved SME partners, but also by the academic partners with regard to training.

As the RECOBA’s goal is to significantly improve efficiency of batch processes in various industries the competitiveness especially of the energy-intensive European process industry will be enhanced in short to medium term. In the medium to long term some of the new control and sensor technologies will also be transferred to continuous processes, can be implemented in modularized production concepts and facilitate the development and implementation for advanced & improved processes which would further strengthen the European process industry. Higher competitiveness of the European industry will preserve existing jobs and will create new highly-skilled jobs as a result of the innovation process.

In case the MPC technology will be spread into all intended areas significant savings of energy and raw material can be achieved. The advanced sensing and control technology will also improve safety of the considered processes. The overview about the intended project impact is visualized in Fig. 3.

Related information

Record Number: 192977 / Last updated on: 2016-12-13
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top