DYN Season's Greetings 2020

Season's Greetings

Dear co-workers, project partners, colleagues, friends, and former members of the dyn group!

No need to say again that 2020 was a special year. As everybody, first in spring and then again in October, we had to adapt our plans almost daily. We had planned the final meeting of the CoPro project that was coordinated by our group for early April in Leverkusen. Then the team at Covestro and I went through the whole trajectory from sticking to the plan over hybrid with a live panel discussion to finally a fully online event within less than two weeks. In the end, it worked out very nicely, with an attendance higher than expected for the physical event, great technical presentations and a stimulating, more politically/ strategically oriented second day with high-level speakers. Many thanks for all who contributed, especially to the Covestro team!

Similarly, within a few weeks, our teaching had to be switched from the usual lectures, tutorials and labs to being fully online. It was a great achievement of the dyn group to realize this transition without a drop of the quality. A huge amount of work of the whole group went into the preparation of the online lectures and video clips, the introduction of remote or simulated lab experiments, etc. The dyn team also pioneered setting up doctoral exams with partial presence of the examiners in a smooth and nicely interactive manner. Great team spirit, great results, big big thanks to everybody involved! Looking back, it was (and is) a lot of work for all of us, but I think that at the end the quality of our teaching even increased and several elements will for sure be preserved in the future. Interestingly, in the beginning we were warned that the communication bandwidth of the university would not be enough for live video teaching, but in fact this was not a problem and this experience provides interesting options for the future.

What of course also changed was conferencing. Zero physical conference participation since February! The group members had to deliver a large number of presentations by videos, and also this has a positive side: Our conference presentations are now available everywhere and anytime. You may want to check our video channel.

I expect that travelling will resume quite quickly once the pandemic is over or at least well under control, and also that conferences will take place again with physical presence, because both respond to humans needs. The same holds for physical meetings, we will not become complete Zombies. But on the other hand, my experience is that for certain kinds of discussions, videoconferencing turned out to be very productive and of course easier to arrange than physical meetings. So some elements will enrich our communication portfolio (and reduce emissions) in the long term. TU Dortmund University will not change into a distance learning university, but if online lectures work as well as live lectures in the auditorium, according to student feedback, why should everybody commute, maybe only for 2 lecture hours on a specific day? Face-to-face elements will surely return, but why not go for hybrid lecturing (with audience and video) in the future?

On August 1, 2020, the dyn group had its 30th birthday, I hope you all got our newsletter on this occasion. As reported there already, and in the Season’s Greetings website, we had again a very successful year. Most notably, after some period of drought, this year five group members passed their doctoral exams: Simon Wenzel, Benedikt Beisheim, Lukas Hebing, Lukas Maxeiner and Sankaranarayanan Subramanian. Please check their excellent work in the related papers or ask for a copy or pdf of their dissertations! I am hoping for at least the same score in 2021! Besides his Dr.-Ing, degree, Simon Wenzel also received the Namur Award 2020 for his dissertation. Big congratulations, and big thanks to him and to the others for many contributions besides their core research!

In 2020, we started two new projects, the BMWI-funded project KEEN which addresses the application of Artificial Intelligence in the Process Industries, and OptiProd. In KEEN , I am responsible for the domain of “self-optimizing plants” (i.e. control and real-time optimization using machine learning techniques). OptiProd, is funded by the state of NRW from EU EFRE funds, and deals with optimal production planning in complex batch plants, in a cooperation with Inosim and Bayer. Currently four scientists from our group are working in KEEN and one in OptiProd, some continuing, some starting their work in the group.

Several group members were offered good positions in industry in 2020, despite the slow-down cause by Corona. This shows that our field of research is in high demand and provides excellent chances on the job market. And, of course, that they are excellently qualified. The two secretaries of the group, Simone Herchenröder and Lisa Guckenberger, both moved on to other, more attractive positions in the department or university. Simone Herchenröder now is the secretary of the Dean of the department, and Lisa Guckenberger works in the 3rd party funds administration department. Both are still supporting us, but they have left a gap that still remains to be filled. Big thanks for their dedicated work and good spirits in the past years!

Furthermore, I am very happy to report that the BCI department now temporarily has two professors whose interests are centered around advanced control and optimization: Dr. Sergio Lucia joined on October 1 as the new professor for Process Automation Systems. Please see his introduction to his group on the Season’s Greetings website! We are working closely together and will implement a smooth transition. Unfortunately, we had to postpone the celebration of the 30 years anniversary of the dyn group and of the inauguration of the PAS group. As you probably all do, we hope for a summer of 2021 full of opportunities for activities which currently are not possible.

I would like to thank all group members, project partners and colleagues for the pleasant and rewarding collaboration and the great work and outstanding results in these difficult times! I wish you enjoyable holidays and a successful and happy year 2021 in good health! Finally, special thanks to the Season’s Greetings team!

Sebastian Engell

dyn Video Portal

The Covid-19 situation created new challenges for all of us as almost all conferences and symposia in 2020 were modified into virtual events. Therefore, video presentations were prepared for most of the 2020 conference papers which are available via the dyn Video Portal.

Please click here to access the DYN Video Portal.

dyn @ CDC 2020

From December 14th to 18th, Yehia Abdelsalam attended the 59th Conference on Decision and Control. As most other conferences this year, the CDC2020 took place in a fully virtual format instead of on Jeju Island, Republic of Korea. V. Abdelsalam presented the following paper:

Y. Abdelsalam, S. Subramanian, S. Engell: Asymptotically Stabilizing Multi-Stage Model Predictive Control

dyn @ PAAT 2020

On November 9th and 10th, the PhD students Stefanie Gerlich, Tim Janus and Robin Semrau represented the dyn group at the PAAT2020 (Jahrestreffen der ProcessNet-Fachgemeinschaften “Prozess-, Apparate- und Anlagentechnik”). Due to the Covid-19 situation, it was organized as a virtual conference. Over 200 chemical engineers, plant constructors, process engineers, and technical chemists from science and industry had the opportunity to present research results, discuss requirements from industrial practice, and jointly develop solutions for new processes in the chemical process industries and other sectors. They DYN contributed the following talks to the program:

S. Gerlich, H. Arab, S. Engell: Online Prozessüberwachung in SMB Prozessen
T. Janus, A. Lünners, S. Engell: Kürzere Optimierungszeiten für Prozessfließbilder in Aspen Plus durch den Einsatz von künstlicher Intelligenz
R. Semrau, F. Tamagnini, A. Tatulea-Codrean, S. Engell: Dynamische Modellierung und Zustandsschätzung eines kontinuierlichen Coiled Flow Inverter Copolymerisationsreaktor

dyn @ SIMS 2020

From September 22nd to 24th, Prof. Engell attended the 61st International Conference of Scandinavian Simulation Society (SIMS 2020) which was arranged as a virtual conference. Prof. Engell gave a plenary presentation on “Real-time optimization and control with inaccurate models” and contributed to the final panel discussion.

dyn @ ESCAPE30

Prof. Engell and five of his PhD students attended the 30th European Symposium on Computer Aided Process Engineering (ESCAPE30). Originally, everyone looked forward to visiting Milano, Italy in May 2020. However, due to the pandemic, the ESCAPE30 was postponed to August 31st til September 2nd and took place in a virtual format.

P. Azadi, S. Ahangari Minaabad, H. Bartusch, R. Klock, S. Engell: Nonlinear Prediction Model of Blast Furnace Operation Status
M. Cegla, S. Engell: Reliable Modelling of Twin-screw Extruders by Integrating the Backflow Cell Methodology into a Mechanistic Model
S. Gerlich, Y.-N. Misz, S. Engell: Online process monitoring in SMB processes
S. Kaiser, S. Engell: Integrating Superstructure Optimization under Uncertainty and Optimal Experimental Design in early Stage Process Development
C. Klanke, V. Yfantis, F. Corominas, S. Engell: Scheduling of a Large-scale Industrial Make-and-Pack Process with Finite Intermediate Buffer using Discrete-time and Precedence-based Models
P. Schmiermoch, B. Beisheim, K. Rahimi-Adli, S. Engell: A Methodology for Data Based Root-cause Analysis for Process Performance Deviations in Continuous Processes
S. Wenzel, F. Riedl, S. Engell: Market-like Distributed Coordination of Individually Constrained and Coupled Production Plants with Quadratic Approximation

dyn @ CEC 2020

From July 19th to 24th, Tim Janus represented the dyn group at the IEEE Congress on Evolutionary Computation (CEC). Due to the pandemic, the conference could not take place in Glasgow, United Kingdom, but only in a virtual format. T. Janus presented the following paper:

T. Janus, A. Lübbers, S. Engell: Neural Networks for Surrogate-assisted Evolutionary optimization of Chemical Processes

dyn @ IFAC World Congress 2020

From July 11th to July 14th, many current and former group members represented the dyn group at the 21st IFAC World Congress in Berlin, Germany. Due to the special circumstances in 2020, the event was held as the 1st Virtual IFAC World Congress. The followong contributions were presented at the conference:

Y. Abdelsalam, S. Subramanian, S. Engell: A Simplified Implementation of Tube-Enhanced Multi-Stage NMPC
T. Ebrahim, S. Engell: A Bi-level Approach to MPC for Switching Nonlinear Systems
A.R. Gottu Mukkula, P. Valiauga, M. Fikar, R. Paulen, S. Engell: Experimental Real Time Optimization of a Continuous Membrane Separation Plant
A.R. Gottu Mukkula, S. Kern, M. Salge, M. Holtkamp, S. Guhl, C. Fleicher, K. Meyer, M.P. Remelhe, M. Maiwald, S. Engell: An Application of Modifier Adaptation with Quadratic Approximation on a Pilot Scale Plant in Industrial Environment
J. D. Hernández, L. Onofri, S. Engell: Optimization of the electric efficiency of the electric steel making process
A. Tatulea-Codrean, J. Fischer, S. Engell: A Multi-stage Economic NMPC for the Tennessee Eastman Challenge Process
A. Tatulea-Codrean, T. Mariani, S. Engell: Model Predictive Control Approach to Autonomous Race Driving for the F1/10 Platform
S. Thangavel, R. Paulen, S. Engell: Dual multi-stage NMPC using sigma point principles
S. Thangavel, S. Engell: An efficient model-error model update strategy for multi-stage NMPC with model-error model
V. Yfantis, S. Büscher, C. Klanke, F. Corominas, S. Engell: A Two-stage Simulated Annealing-based Scheduling Algorithm for a Make-and-Pack Production Plant

dyn @ ECC 2020

Anwesh Reddy Gottu Mukkula and Sakti Thangavel represented the dyn group at the European Control Conference from May 12th to May 15th. Originally planned to take place in Saint Petersburg, Russia, the conference took place in virtual due to the Covid-19 pandemic. The following papers were presented:

A. R. Gottu Mukkula, S. Engell: Guaranteed Model Adequacy for Modifier Adaptation With Quadratic Approximation
T. Thangavel, R. Paulen, S. Engell: Adaptive multi-stage NMPC using sigma point principles

dyn @ ACODS 2020

From February 16th to 19th, Prof. Engell and his PhD student Sakthi Thangavel represened the dyn group at the Advances in Control & Optimization of Dynamical Systems (ACODS) conference in Chennai, India. Prof. Engell gave a plenary lecture on “Robust Performance Optimizing NMPC by Multistage Optimization” and was also invited as a member of the closing panel discussion. He also visited NEERI in Chennai to discuss the work on algae-based wastewater treatment within the LOTUS project, and the control group at IIT Madras. S. Thangavel presented the following paper:

T. Thangavel, R. Paulen, S. Engell: Multi-stage NMPC using sigma point principles

Left Joined

dyn Members 2020

We all would like to thank our partners, colleagues, students and alumni for their support and the fruitful collaboration all throughout 2020. We wish you a bright, happy and successful year 2021!

Farewell to our colleagues

Lisa Guckenberger

Lisa started working as Prof. Engell´s administrative Assistant in June 2017. Getting to know the administrative work for the various research projects of the dyn group and managing the finances for the chair, she was able to get a position at the Third-Party Funding Management of TU Dortmund. From this position, she is supporting the BCI department among others with the administration of research projects. She will keep in good memory the collegial work atmosphere.

Sandra Isabel Gröning

Isa started her work at the dyn group in October 2019 as one of Prof. Engell's secretaries replacing Lisa during her parental leave. When Lisa returned to the group, Isa joined the Department of Human Resources in July 2020.

Egidio Leo

Egidio started his research career with the dyn group in October 2016 as a Marie-Curie fellow. He focused on mathematical optimization under uncertainty with applications on planning and scheduling problems for the process industry. As a Marie-Curie Early Stage Researcher, Egidio was also a visiting researcher at the petrochemical company Ineos Cologne and at Carnegie Mellon University in Pittsburgh, USA. Since November 2020, Egidio is employed at the BP Gelsenkirchen in the production planning department.

Clemens Lindscheid

Clemens Lindscheid started in August 2015 at the dyn group with research on the application of model based optimizing control algorithms within the EU project MOBOCON. His research covered a whole range of activities from providing reliable basic automation as a basis for advanced process control, which he also realized during the EU-project ADREM and an industrial project, to the integration of the operator during the operation of such algorithms. Clemens Lindscheid is currently taking time off from work to develop his own projects.

Lukas Samuel Maxeiner

After finishing his B.Sc. and M.Sc. degrees in Chemical Engineering at TU Dortmund University with a semester abroad at Carnegie Mellon University in Pittsburgh, USA, Lukas joined the dyn group in January 2015. During the time as research associate, his main research topic was dual-based distributed optimization for the use case of symbiotic optimal decision making between entities that do not necessarily trust each other and therefore only want to exchange limited information. This research was part of the EU funded research projects DYMASOS and COPRO. Other topics included the application of MILP in industrial practice, hybrid modelling, and, as a private side project, investigating possible applications of blockchain technology in the chemical industry. Since June 2020, Lukas is employed at the Evonik Technology & Infrastructure department. Currently, he is working on advancing digitalization in production plants and is based out of Marl. The time at the chair has shaped him professionally and personally, and connected him in spirit with the fellows that went the same way.

Yannik Misz

Yannik-Noel Misz did his B.Sc. in biochemical engineering at TU Dortmund University followed by an M.Sc. degree in Automation & Robotics. After his master thesis on state and parameter estimation in simulated moving bed processes, he joined the dyn group as a research associate after being a student assistant for two years. He continued his work on the ending CoPro project and joined the LOTUS project as well. From December 2020, he will be working as a process automation engineer in industry on projects covering a broad range from waste water treatment to ship engines.

Alexandru Tatulea-Codrean

Alexandru Tatulea-Codrean first started to work with the dyn group in 2012 as a student assistant. After completing his master's degree in Automation & Robotics, Alex joined the group as a research associate in May 2014. What followed were 6 rewarding years spent working on topics related to NMPC and optimization, in the company of highly motivated co-workers and students alike. The experience gained at the dyn group and the overall desire to continue developing software tools and APC topics brought him to Bayer AG in Leverkusen, where he started working in May 2020. The similarities between his new job and the old one are surprisingly and pleasantly numerous. However, Alex says that he misses teaching the A&R courses and the liveliness of student projects. He is looking forward to staying in touch with the group in the future!

Sakthi Thangavel

Sakthi did his Bachelor Degree in Mechatronics Engineering at Anna University, Chennai, India. After his Bachelors, he worked in Infosys Ltd as a Systems Engineer for 2 years. He came to Germany in 2012 to start his M.Sc. degree in Automation & Robotics at TU Dortmund which he finished with a master thesis in the area of Dual control. Afterwards, he joined the dyn group in January 2015 as a research associate. Sakthi was very active in both teaching and research projects. His research interests are in dual control, model predictive control, optimal control and optimal resource allocation. During his time at the dyn group, he was involved in two successful EU projects: CONSENS and MOBOCON. Sakthi will continue his professional journey as an APC Engineer at the INEOS Phenol, Gladbeck, starting in January 2021. Sakthi will always cherish the time he spent at the dyn group, which was very welcoming and helped him to develop both personally and professionally. He will always consider the dyn group as his extended family.

New dyn Members

Mohamed Elsheikh

Mohamed Elsheikh received his B.Sc. in Mechatronics engineering from the German University in Cairo in 2017. He wrote his Bachelor thesis with the title “Determining Stabilizing Parameter Spaces for Time Delay Systems” at the institute of Automatic Control (IRT) at RWTH Aachen University. In 2017, he started his M.Sc. degree in Process Automation and Robotics at TU Dortmund University. During his master studies, he did an internship at Bayer AG where is also did his master thesis with the title “Multi-Rate Moving Horizon Estimation for Bioprocesses” in cooperation with the dyn group. Mohamed joined the dyn group as a research associate in August 2020.

Jesus David Hernandez Ortiz

Jesus holds a bachelor’s degree in Electronics Engineering from Universidad Industrial de Santander in Colombia and a master’s degree in Energy Engineering from Politecnico di Milano. He has held various positions as a Process Control engineer in the oil and gas and the metal processing industries and joined the dyn group as a research associate in January 2020. Since 2017, he has cooperated with the chair as a member of the Marie Curie H2020 PRONTO project.

Filippo Tamagnini

Filippo Tamagnini studied Chemical and Process Engineering at Alma Mater Studiorum, Bologna. During his master studies he spent one semester at TU Dortmund University preparing his master thesis on the topic of state estimation applied to a laboratory case study. He joined the dyn group in May, 2020 and is currently involved in the EU Horizon 2020 Project SIMPLIFY (Sonication and Microwave Processing of Material Feedstock).

Joschka Winz

Joschka Winz studied Chemical Engineering in Dortmund from 2014 to 2020 completing both the Bachelor and the Master program. His Bachelor thesis is titled “Adaptive sequential sampling for surrogate modelling of fugacity coefficients”. He conducted his Master thesis at BASF SE in Ludwigshafen am Rhein on the topic “Parameter estimation using a dynamic simulation of batch distillation experiments”. During his studies he worked as a student assistant at the dyn group studied abroad at Carnegie Mellon University in Pittsburgh, USA, for one semester. In June 2020, he joined the group as a research associate. Currently, his research focusses on the application of artificial intelligence to process industry by means of hybrid modeling with emphasis on process optimization.

People Conferences Publications

Introduction by Prof. Sergio Lucia

Dear Colleagues and Friends,

In October 2020, the Laboratory of Process Automation Systems was established at the Department of Chemical and Biochemical Engineering of the TU Dortmund. After finishing my PhD in 2014 at the DYN Chair, and several intermediate stops in Magdeburg, Boston, and Berlin, I am very happy to come back to the BCI to establish my own research group.

The group starts with three great PhD Students: Benjamin, Felix and Marco. I am very happy and grateful that they decided to leave Berlin to come with me to Dortmund, even in the middle of the current difficult situation.

Together with them, we aim to establish a larger group with a research focus on the interface between automatic control, numerical optimization and machine learning. The increasing availability of computing power, large amounts of data, and digital twins should not be ignored. But, what is the best way to leverage all these new possibilities to obtain the largest possible benefit? How can this be done in a systematic fashion? These are some of the questions that our group will study in the coming years.

We plan to apply the methodological advances of the group to different fields, focusing on the advanced operation of chemical or biotechnological processes that can contribute to more efficient and sustainable industries. We are looking forward to interesting collaborations with colleagues both in the department and in industry!

I wish all of you a great holiday period, even if probably very different than other years, and a successful and happy year 2021!

Sergio Lucia

PAS Members

Felix Fiedler, Benjamin Karg and Marco Molnar joined the newly founded Laboratory for Process Automation Systems (PAS) at TU Dortmund on the 1st of October this year. Previously, they were working with Prof. Sergio Lucia at TU Berlin at the chair of Internet of Things for Smart Buildings.

Felix Fiedler

Felix Fiedler started his PhD in October 2018. In his research, Felix Fiedler works on the edge of machine learning and predictive control, with publications such as “Economic nonlinear predictive control of water distribution networks based on surrogate modeling and automatic clustering” (IFAC WC 2020). He is looking forward to his new life in Dortmund and appreciates the warm welcome of the people from TU Dortmund, especially during challenging times. Stay healthy everyone!

Benjamin Karg

Benjamin Karg started his PhD in July 2017. His research interests are control engineering, artificial intelligence and edge computing. By exploiting the expressive capabilities of deep neural networks, complex control and decision-making algorithms can be approximated, which in return allows the deployment of said algorithms on computationally limited hardware. Currently he is using optimization-based and probabilistic methods for the verification of learned controllers to obtain guarantees on performance and safety with respect to operational requirements and for modifying the controllers such that they satisfy the performance criteria.

Marco Molnar

Marco received his B.Sc. and M.Sc. in Electrical Engineering from TU Berlin. During his studies, he spent two quarters at the University of California Santa Barbara where he had the chance to do research at the Dynamic Robotics Laboratory of Prof. Katie Byl. His master thesis at TU Berlin was focused on training data aggregation algorithms for the modeling of explicit MPC controllers by deep neural networks. In July 2020 he started his Ph.D. and is working on extending robust MPC to high dimensional systems in a collaboration with the Systems Control and Optimization Laboratory of the University of Freiburg.

Conferences

PAS @ ECC 2020

From 13.05.2020 to 15.05.2020 PhD student Felix Fiedler represented the chair at European Control Conference (ECC). The conference was supposed to be hold in St. Petersburg, Russia but had to be reorganized as an online event, due to the Corona pandemic. ECC 2020 was a success regardless, with over 400 participants from more than 45 countries. Felix Fiedler presented his work “A Probabilistic Moving Horizon Estimation Framework Applied to the Visual-Inertial Sensor Fusion Problem”. Many thanks go to the co-authors Dirk Baumbach, Anko Börner and Sergio Lucia for the great collaboration.

PAS @ IFAC 2020

The IFAC world-congress occurs only every three years and is one of the largest gatherings of control researchers in the world. This year marked no exception with over 3000 presented papers in more than 250 sessions. Unfortunately, the control community cannot control global pandemics and the event, originally planned to take place in Berlin, Germany, was also reorganized as an online event. Felix Fiedler participated with a presentation on the project “Economic nonlinear predictive control of water distribution networks based on surrogate modeling and automatic clustering”. WaterGate, as the authors, Felix Fiedler, Andrea Cominola and Sergio Lucia liked to call the project, was received with great interest.

PAS @ CCTA 2020

Montreal, Canada would have been the location for the 4th conference on control technology and applications (CCTA) from 24.-26.08.2020 but unfortunately Covid also impaired this event and forced the organizers to move to a virtual concept. Felix Fiedler represented the chair at this conference with a contribution on “PredicTor: Predictive Congestion Control for the Tor Network”, a project that was conducted together with Christoph Döpmann, Florian Tscholz and Sergio Lucia.

PAS @ CDC 2020

PhD student Benjamin Karg represented the Chair of Process Automation Systems at the 59th Conference on Decision and Control, which would have taken place from December 14th to 18th on Jeju Island, South Korea, but was held fully virtually due to given circumstances. The presented contribution proposed a method which allows to obtain a deep neural network controller guaranteeing feasibility and asymptotically stabilizing behavior. Thanks for the fruitful discussions with the co-author Sergio Lucia.

Publications

Journal Articles 2020

Ghasemi, Lucia, Lucia:

Computing in the blink of an eye: Current possibilities for edge computing and hardware-agnostic programming

IEEE Access 8, 41626-41636, Full Paper

With the rapid advancements of the internet of things, systems including sensing, communication, and computation become ubiquitous. The systems that are built with these technologies are increasingly complex and therefore require more automation and intelligent decision-making, while often including contact with humans. It is thus critical that such interactions run smoothly in real time, and that the automation strategies do not introduce important delays, usually not larger than 100 milliseconds, as the blink of a human eye. Pushing the deployment of the algorithms on embedded devices closer to where data is collected to avoid delays is one of the main motivations of edge computing. Further advantages of edge computing include improved reliability and data privacy management. This work showcases the possibilities of different embedded platforms that are often used as edge computing nodes: embedded microcontrollers, embedded microprocessors, FPGAs and embedded GPUs. The embedded solutions are compared with respect to their cost, complexity, energy consumption and computing speed establishing valuable guidelines for designers of complex systems that need to make use of edge computing. Furthermore, this paper shows the possibilities of hardware-agnostic programming using OpenCL, illustrating the price to pay in efficiency when software can be easily deployed on different hardware platforms.

Karg, Lucia:

Efficient representation and approximation of model predictive control laws via deep learning

IEEE Transactions on Cybernetics, vol. 50, no. 9, pp. 3866-3878, 2020, Full Paper

We show that artificial neural networks with rectifier units as activation functions can exactly represent the piecewise affine function that results from the formulation of model predictive control of linear time-invariant systems. The choice of deep neural networks is particularly interesting as they can represent exponentially many more affine regions compared to networks with only one hidden layer. We provide theoretical bounds on the minimum number of hidden layers and neurons per layer that a neural network should have to exactly represent a given model predictive control law. The proposed approach has a strong potential as an approximation method of predictive control laws, leading to better approximation quality and significantly smaller memory requirements than previous approaches, as we illustrate via simulation examples. We also suggest different alternatives to correct or quantify the approximation error. Since the online evaluation of neural networks is extremely simple, the approximated con-trollers can be deployed on low-power embedded devices with small storage capacity, enabling the implementation of advanced decision-making strategies for complex cyber-physical systems with limited computing capabilities.

Lucia, Subramanian, Limon, Engell:

Stability properties of multi-stage nonlinear model predictive control

Systems & Control Letters 143, 104743

This paper discusses the stability properties of a robust nonlinear model predictive control (NMPC) scheme that is based on a multi-stage optimization formulation. The use of a scenario tree to represent the uncertainty makes it possible to formulate a closed-loop robust approach with recourse which improves the open-loop approach in terms of performance and domain of attraction. We show that a straightforward formulation of a multi-stage NMPC scheme does not guarantee Input-to-State stability (ISS) in a deterministic setting, in contrast to the results that one gets using stochastic stability concepts. Since for many applications deterministic stability guarantees are desired, we provide an alternative formulation to achieve deterministic ISS and recursive feasibility guarantees for the case of discrete values of the uncertainty. The design and the performance of the proposed schemes are illustrated by simulations for a nonlinear reactor.

Conference Articles 2020

Braun, Albrecht, Lucia:

A hierarchical attack identification method for nonlinear systems

59th IEEE Conference on Decision and Control, Jeju Island, South Korea, 2020

Many autonomous control systems are frequently exposed to attacks, so methods for attack identification are crucial for a safe operation. To preserve the privacy of the subsystems and achieve scalability in large-scale systems, identification algorithms should not require global model knowledge. We analyze a previously presented method for hierarchical attack identification, that is embedded in a distributed control setup for systems of systems with coupled nonlinear dynamics. It is based on the exchange of local sensitivity information and ideas from sparse signal recovery. In this paper, we prove sufficient conditions under which the method is guaranteed to identify all components affected by some unknown attack. Even though a general class of nonlinear dynamic systems is considered, our rigorous theoretical guarantees are applicable to practically relevant examples, which is underlined by numerical experiments with the IEEE 30 bus power system.

Braun, Albrecht, Lucia:

Identifying attacks on nonlinear cyber-physical systems in a robust model predictive control setup

European Control Conference, St. Petersburg, Russia, 2020

The design of resilient control strategies has become a crucial security issue for autonomous control systems which are exposed to the threat of attacks to both their physical and cyber components. In this paper, we present a setup consisting of two interlacing approaches towards secure control of multi-agent nonlinear dynamic systems under attack. First, we combine aspects from Pareto optimality and robust model predictive control (MPC) to maintain the system in a feasible state even if attacks with large impact occur. Second, we propose an attack identification method based on ideas from signal recovery, considering optimization problems with penalty terms for the violation of the Karush-Kuhn-Tucker conditions. We compare our approach to Nash- and Pareto-based non-robust MPC and illustrate the solution procedure with a nonlinear numerical example.

Braun, Albrecht, Lucia:

Hierarchical attack identification for distributed robust nonlinear control

21st IFAC World Congress, Berlin, Germany, 2020

Developing tools for attack identification in large-scale networked control systems is a research area of increasing significance for the secure and reliable operation of autonomous control systems. Due to scalability limits and privacy issues of individual subsystems, attack identification methods should not rely on global model knowledge. We address systems of interconnected nonlinear subsystems with coupled dynamics or constraints in a distributed control setup. The local controllers share information about the coupling variables of the subsystems and are designed to be robust towards attacks and uncertain influences through neighboring subsystems. We present a scalable hierarchical attack identification method which monitors the evolution of the coupling variables after an attack occurred in some unknown subsystem. Based on the mutual exchange of local sensitivity information among the subsystems, the propagation of the attack through the network is approximated. The propagation equations are used to formulate a quadratic program whose solution determines the attack signal that explains the observed network evolution best. The developed approach is applied to the IEEE 30 bus system to illustrate attack identification in power systems with faulty buses.

Eckhoff, Kok, Lucia, Seel:

Sparse Magnetometer-free Inertial Motion Tracking--A Condition for Observability in Double Hinge Joint Systems

21st IFAC World Congress, Berlin, Germany, 2020

Inertial measurement units are commonly used in a growing number of application fields to track or capture motions of kinematic chains, such as human limbs, exoskeletons or robotic actuators. A major challenge is the presence of magnetic disturbances that result in unreliable magnetometer readings. Recent research revealed that this problem can be overcome by exploitation of kinematic constraints. While typically each segment of the kinematic chain is equipped with an IMU, a novel approach called sparse inertial motion tracking aims at infering the complete motion states from measurements of a reduced set of sensors. In the present contribution, we combine the magnetometer-free and the sparse approach for real-time motion tracking of double-hinge joint systems with non-parallel joint axes. Analyzing the observability of the system, we find a condition which assures that the relative orientations between all segments are uniquely determined by a kinematic constraint, which contains only the gyroscope readings. Furthermore, we propose a moving-horizon estimator and validate it in a simulation study of three movements with different degrees of excitation. The results of this study confirm all theoretical conjectures and demonstrate that magnetometer-free sparse inertial real-time motion tracking is feasible under precise and simple excitation conditions.

Fiedler, Baumbach, Börner, Lucia:

A Probabilistic Moving Horizon Estimation Framework Applied to the Visual-Inertial Sensor Fusion Problem

European Control Conference, St. Petersburg, Russia, 2020

We propose a novel method to compute the arrival cost for the moving horizon estimator. The choice of the arrival cost is an important challenge and is known to have significant influence on the performance of the estimator. Most common approaches are based on implementing a complementary extended Kalman filter to propagate an approximate measure of the uncertainty. Our approach is based on the probabilistic interpretation of the moving horizon estimator and its analogy to the maximum a posteriori estimator. We derive a method to directly obtain the required uncertainties from the Hessian of the moving horizon estimation objective function. We showcase our novel approach with the challenging visual-inertial sensor fusion problem that commonly arises in visual navigation systems. The estimation performance is significantly better compared to our previous results based on the extended Kalman filter. Additionally, the proposed algorithm calibrates the inertial sensor online and is immediately ready for operation.

Fiedler, Cominola, Lucia:

Economic nonlinear predictive control of water distribution networks based on surrogate modeling and automatic

21st IFAC World Congress, Berlin, Germany, 2020

The operation of large-scale water distribution networks (WDNs) is a complex control task due to the size of the problem, the need to consider key operational, quality and safety-related constraints as well as because of the presence of uncertainties. An efficient operation of WDNs can lead to considerable reduction in the energy used to distribute the required amounts of water, leading to significant economic savings. Many model predictive control (MPC) schemes have been proposed in the literature to tackle this control problem. However, finding a control-oriented model that can be used in an optimization framework, which captures nonlinear behavior of the water network and is of a manageable size is a very important challenge faced in practice. We propose the use of a data-based automatic clustering method that clusters similar nodes of the network to reduce the model size and then learn a deep-learning based model of the clustered network. The learned model is used within an economic nonlinear MPC framework. The proposed method leads to a flexible scheme for economic robust nonlinear MPC of large WDNs that can be solved in real time, leads to significant energy savings and is robust to uncertain water demands. The potential of the proposed approach is illustrated by simulation results of a benchmark WDN model.

Fiedler, Döpmann, Tschorsch, Lucia:

PredicTor: Predictive Congestion Control for the Tor Network

4th IEEE Conference on Control Technology and Application, Montreal, Canada, 2020, Full Paper

In the Tor network, anonymity is achieved through a multi-layered architecture, which comes at the cost of a complex network. Scheduling data in this network is a challenging task and the current approach shows to be incapable of avoiding network congestion and allocating fair data rates. We propose PredicTor, a distributed model predictive control approach, to tackle these challenges. PredicTor is designed to schedule incoming and outgoing data rates on individual nodes of the Tor architecture, leading to a scalable approach. We successfully avoid congestion through exchanging information of predicted behavior with adjacent nodes. Furthermore, we formulate PredicTor with a focus on fair allocation of resources, for which we present and prove a novel optimization-based fairness approach. Our proposed controller is evaluated with the popular network simulator ns-3, where we compare it with the current Tor scheduler as well as with another recently proposed enhancement. PredicTor shows significant improvements over previous approaches, especially with respect to latency.

Karg, Lucia:

Stability and feasibility of neural network-based controllers via output-range analysis

59th IEEE Conference on Decision and Control, Jeju Island, South Korea, 2020

Neural networks can be used as approximations of several complex control schemes such as model predictive control. We show in this paper which properties deep neural networks with rectifier linear units as activation functions need to satisfy to guarantee constraint satisfaction and asymptotic stability of the closed-loop system. To do so, we introduce a parametric description of the neural network controller and use a mixed-integer linear programming formulation to perform output range analysis of neural networks. We also propose a novel method to modify a neural network controller such that it performs optimally in the LQR sense in a region surrounding the equilibrium. The proposed method enables the analysis and design of neural network controllers with formal safety guarantees as we illustrate with simulation results.

Mammarella, Alamo, Lucia, Dabbene:

A probabilistic validation approach for penalty function design in stochastic model predictive control

21st IFAC World Congress, Berlin, Germany, 2020

In this paper, we consider a stochastic Model Predictive Control able to account for effects of additive stochastic disturbance with unbounded support, and requiring no restrictive assumption on either independence nor Gaussianity. We revisit the rather classical approach based on penalty functions, with the aim of designing a control scheme that meets some given probabilistic specifications. The main difference with previous approaches is that we do not recur to the notion of probabilistic recursive feasibility, and hence we do not consider separately the unfeasible case. In particular, two probabilistic design problems are envisioned. The first randomization problem aims to design offline the constraint set tightening, following an approach inherited from tube-based MPC. For the second probabilistic scheme, a specific probabilistic validation approach is exploited for tuning the penalty parameter, to be selected offline among a finite-family of possible values. The simple algorithm here proposed allows designing a single controller, always guaranteeing feasibility of the online optimization problem. The proposed method is shown to be more computationally tractable than previous schemes. This is due to the fact that the sample complexity for both probabilistic design problems depends on the prediction horizon in a logarithmic way, unlike scenario-based approaches which exhibit linear dependence. The efficacy of the proposed approach is demonstrated with a numerical example.

Journal Articles Conference Papers Master Theses Bachelor Theses

Journal Articles 2020

Beisheim, Krämer, Engell:

Hierarchical aggregation of energy performance indicators in continuous production processes

Applied Energy, 264, 114709, , Full Paper

The mitigation of the climate change requires a significant reduction of the fossil energy consumption in all industrial sectors. The implementation of formalized management systems supports the industry to continuously improve the energy performance which is measured using so called “Energy Performance Indicators”. One essential requirement for the evaluation is the correction of these indicators and the corresponding baselines by the influences of external static or dynamic factors, e.g. the ambient conditions, the product spectrum or the plant load. This is in particular difficult for large integrated production sites as e.g. in the chemical industry. In this contribution, an aggregation method is proposed to exploit the analysis of the factors on lower hierarchical layers for the evaluation of the performance of an aggregated domain. Thereby, the resulting aggregated indicator and the corresponding baseline consider all the identified factors from the lower layers, which facilitates the analysis and allocation of possible savings potentials. The concept is applied exemplarily to an integrated chemical production site and the contributions of each plant to the deviation of the energy performance of the site are analyzed. The method facilitates the verification of the improvement of the energy performance as required by ISO 50001:2018 and helps decision makers to prioritize investments in energy efficiency projects. The results can be used for discussions with policy-makers, certification bodies and other stakeholders on energy efficiency targets.

Bouaswaig, Rahimi-Adli, Roth, Hosseini, Vale, Engell, Birk:

Application of a grey-box modelling approach for the online monitoring of batch production in the chemical industry

At-Automatisierungstechnik, 68 (7), pp. 582-598

Model-based solutions for monitoring and control of chemical batch processes have been of interest in research for many decades. However, unlike in continuous processes, in which model-based tools such as Model Predictive Control (MPC) have become a standard in the industry, the reported use of models for batch processes, either for monitoring or control, is rather scarce. This limited use is attributed partly to the inherent complexity of the batch processes (e. g., dynamic, nonlinear, multipurpose) and partly to the lack of appropriate commercial tools in the past. In recent years, algorithms and commercial tools for model-based monitoring and control of batch processes have become more mature and in the era of Industry 4.0 and digitalization they are slowly but steadily gaining more interest in real-word batch applications. This contribution provides a practical example in this application field. Specifically, the use of a grey-box modeling approach, in which a multiway Projection to Latent Structure (PLS) model is combined with a first-principles model, to monitor the evolution of a batch polymerization process and predict in real-time the final batch quality is reported. The modeling approach is described, and the experimental results obtained from an industrial batch laboratory reactor are presented.

Castro, Dalle Ave, Engell, Grossmann, Harjunkoski:

Industrial Demand Side Management of a Steel Plant Considering Alternative Power Modes and Electrode Replacement

Industrial and Engineering Chemistry Research, 59 (30), pp. 13642-13656

As a major energy consumer, steel plants can help stabilize the power grid by shifting production from periods with high demand. Electric arc furnaces can be operated at different power levels, affecting the energy efficiency, the duration of melting tasks, and the rate of electrode degradation, which has previously been neglected. We thus propose a new mixed-integer linear programming (MILP) formulation for optimal scheduling under time-of-use electricity pricing that captures the tradeoffs involved. It relies on the resource-task network (RTN) for modeling processing tasks with variable electrode mass depletion and replacement tasks that regenerate the mass. Results for an industrial case study show that the high-power mode, which allows for faster execution and to fit more tasks in low-price periods but is the least energy-efficient and consumes the largest mass of electrode, is mostly avoided. It indicates that electrode replacement plays an important role in total cost minimization.

Hebing, Neymann, Engell:

Application of dynamic metabolic flux analysis for process modeling: Robust flux estimation with regularization, confidence bounds, and selection of elementary modes

Biotechnology and Bioengineering, 117 (7), pp. 2058-2073, Full Paper

In macroscopic dynamic models of fermentation processes, elementary modes (EM) derived from metabolic networks are often used to describe the reaction stoichiometry in a simplified manner and to build predictive models by parameterizing kinetic rate equations for the EM. In this procedure, the selection of a set of EM is a key step which is followed by an estimation of their reaction rates and of the associated confidence bounds. In this paper, we present a method for the computation of reaction rates of cellular reactions and EM as well as an algorithm for the selection of EM for process modeling. The method is based on the dynamic metabolic flux analysis (DMFA) proposed by Leighty and Antoniewicz (2011, Metab Eng, 13(6), 745–755) with additional constraints, regularization and analysis of uncertainty. Instead of using estimated uptake or secretion rates, concentration measurements are used directly to avoid an amplification of measurement errors by numerical differentiation. It is shown that the regularized DMFA for EM method is significantly more robust against measurement noise than methods using estimated rates. The confidence intervals for the estimated reaction rates are obtained by bootstrapping. For the selection of a set of EM for a given st oichiometric model, the DMFA for EM method is combined with a multiobjective genetic algorithm. The method is applied to real data from a CHO fed‐batch process. From measurements of six fed‐batch experiments, 10 EM were identified as the smallest subset of EM based upon which the data can be described sufficiently accurately by a dynamic model. The estimated EM reaction rates and their confidence intervals at different process conditions provide useful information for the kinetic modeling and subsequent process optimization.

Hebing, Tran, Brandt, Engell:

Robust Optimizing Control of Fermentation Processes Based on a Set of Structurally Different Process Models

Industrial and Engineering Chemistry Research, 59 (6), pp. 2566-2580

The performance of most bioprocesses can be improved significantly by the application of model-based methods from advanced process control (APC). However, due to the complexity of the processes and the limited knowledge of them, plant–model mismatch is unavoidable. A variety of different modeling strategies (each with individual advantages and deficiencies) can be applied, but still, the confidence in a single process model is often low; therefore, the application of classical APC is difficult. In order to operate under possible plant–model mismatch, a robust closed-loop optimizing control strategy was developed in which the mismatch is counteracted by an adaptive model correction and the parallel usage and evaluation of structurally different models. Robust multistage nonlinear model predictive control is used for the online optimization of the process trajectories in order to maximize the performance. The adapted, structurally different models are used herein as weighted scenarios for the prediction of the process, which account for structural uncertainties. It is shown in simulation studies of a CHO cultivation process that the usage of multiple, adapted models as scenarios improves (1) the accuracy of the state estimation and (2) the overall process performance.

Hernández, Onofri, Engell:

Numerical Estimation of the Geometry and Temperature of An Alternating Current Steelmaking Electric Arc

Steel Research International, 2000386, Full Paper

A channel arc model (CAM) that predicts the temperature and the geometry of an electric arc from its voltage and impedance set-points is presented. The core of the model is a nonlinear programming (NLP) formulation that minimizes the entropy production of a plasma column, the physical and electrical properties of which satisfy the Elenbaas–Heller equation and Ohm's law. The radiative properties of the plasma are approximated utilizing the net emission coefficient (NEC), and the NLP is solved using a global numerical solver. The effects of the voltage and impedance set-points on the length of the electric arc are studied, and a linear formula that estimates the length of the arc in terms of its electrical set-points is deducted. The length of various electric arcs is measured in a fully operative electric arc furnace (EAF), and the results are used to validate the proposed models. The errors in the predictions of the models are 0.5 and 0.4 cm. In comparison, the existing empirical and Bowman formulae estimate the length of the experimental arcs with errors of 2.1 and 2.6 cm. A simplified formula to estimate the temperature of an electric arc in terms of its electrical set-points is also presented.

Klessova, Thomas, Engell:

Structuring inter-organizational R&D projects: Towards a better understanding of the project architecture as an interplay between activity coordination and knowledge integration

International Journal of Project Management, 38 (5), pp. 291-306, Full Paper

The architects of inter-organizational R&D projects organize collaboration by structuring the activities and the knowledge base of the project. How do these two dimensions interplay and what are the implications on the project execution? The paper aims at developing new perspectives on inter-organizational multi-actor R&D projects using an exploratory inductive multi-case study of projects funded by the European Union's Research and Innovation Programmes. The projects have been studied simultaneously in terms of activity coordination and knowledge integration as well as the implications of their interplay on collaboration, project resilience and project management. The paper provides empirical evidence about six patterns of project architecture. The workflow-integrated architecture disintegrates the knowledge base, provides a lower collaboration potential and may require high management efforts, while a workflow-decomposed architecture makes project management easy but provides little added value from the inter-organizational setting. Nearly decomposable architectures offer the highest collaboration potential under contingent conditions.

Lucia, Subramanian, Limon, Engell:

Stability properties of multi-stage nonlinear model predictive control

Systems & Control Letters, 143, 104743

Maxeiner, Engell:

An accelerated dual method based on analytical extrapolation for distributed quadratic optimization of large-scale production complexes

Computers & Chemical Engineering, 135, 106728, Full Paper

Chemical production sites usually consist of plants that are owned by different companies or business units but are tightly connected by streams of materials and carriers of energy. Distributed optimization, where each entity optimizes its objective and the transfer prices of energy and materials are adapted by a coordinator, is a promising approach to this kind of problems, as confidentiality of internal data can be preserved. In this contribution, we propose an extension of the widely used subgradient methods for inequality constrained distributed QPs, which we call analytical extrapolation (AE). Therein, the analytical structure of the dual function is exploited to speed up convergence. Two strategies for handling changing sets of active constraints are presented. We investigate the performance of our algorithm on test problems, where different problem parameters are varied, and show that the performance of our algorithm is in most cases significantly better than that of other methods.

Maxeiner, Engell:

Comparison of dual based optimization methods for distributed trajectory optimization of coupled semi-batch processes

Optimization and Engineering, 21 (3), pp. 761-802, Full Paper

The physical and virtual connectivity of systems via flows of energy, material, information, etc., steadily increases. This paper deals with systems of sub-systems that are connected by networks of shared resources that have to be balanced. For the optimal operation of the overall system, the couplings between the sub-systems must be taken into account, and the overall optimum will usually deviate from the local optima of the sub-systems. However, for reasons, such as problem size, confidentiality, resilience to breakdowns, or generally when dealing with autonomous systems, monolithic optimization is often infeasible. In this contribution, iterative distributed optimization methods based on dual decomposition where the values of the objective functions of the different sub-systems do not have to be shared are investigated. We consider connected dynamic systems that share resources. This situation arises for continuous processes in transient conditions between different steady states and in inherently discontinuous processes, such as batch production processes. This problem is challenging since small changes during the iterations towards the satisfaction of the overarching constraints can lead to significant changes in the arc structures of the optimal solutions for the sub-systems. Moreover, meeting endpoint constraints at free final times complicates the problem. We propose a solution strategy for coupled semi-batch processes and compare different numerical approaches, the sub-gradient method, ADMM, and ALADIN, and show that convexification of the sub-systems around feasible points increases the speed of convergence while using second-order information does not necessarily do so. Since sharing of resources has an influence on whether trajectory dependent terminal constraints can be satisfied, we propose a heuristic for the computation of free final times of the sub-systems that allows the dynamic sub-processes to meet the constraints. For the example of several semi-batch reactors which are coupled via a bound on the total feed flow rate, we demonstrate that the distributed methods converge to (local) optima and highlight the strengths and the weaknesses of the different distributed optimization methods.

Thangavel, Paulen, Engell:

Robust Multi-Stage Nonlinear Model Predictive Control Using Sigma Points

Processes 2020, 8, 851, Full Paper

We address the question of how to reduce the inevitable loss of performance that is incurred by robust multi-stage NMPC due to the lack of knowledge compared to the case where the exact plant model (no uncertainty) is available. Multi-stage NMPC in the usual setting over-approximates a continuous parametric uncertainty set by a box and includes the corners of the box and the center point into the scenario tree. If the uncertainty set is not a box, this augments the uncertainty set and results in a performance loss. In this paper, we propose to mitigate this problem by two different approaches where the scenario tree of the multi-stage NMPC is built using sigma points. The chosen sigma points help to capture the true mean and covariance of the uncertainty set more precisely. The first method computes a box over-approximation of the reachable set of the system states whereas the second method computes a box over-approximation of the reachable set of the constraint function using the unscented transformation. The advantages of the proposed schemes over the traditional multi-stage NMPC are demonstrated using simulation studies of a simple semi-batch reactor and a more complex industrial semi-batch polymerization reactor benchmark example.

Wenzel, Riedl, Engell:

An efficient hierarchical market-like coordination algorithm for coupled production systems based on quadratic approximation

Computers & Chemical Engineering, 134, 106704, Full Paper

In an increasingly digitized, integrated, and connected production environment, the coordination of complex coupled systems of systems is essential to ensure a resource efficient and optimal operation. Large production sites or industrial clusters are such systems, where the coordination is challenging. One of the challenges is the physical coupling of the individual production plants via networks of shared resources such as steam, intermediates, wastewater or heating gas. If only a restricted amount of business relevant data can be exchanged within the system, distributed market-like coordination methods can be used to establish an auction in a micro market to optimally balance the shared resource networks. The drawback of these methods is their slow rate of convergence. Thus, we propose an efficient hierarchical market-like coordination algorithm that approximates the responses of the subsystems to find the optimal solution within fewer iterations. We demonstrate the efficiency of approach for two numerical case studies.

Conference Articles 2020

Abdelsalam, Subramanian, Engell:

A Simplified Implementation of Tube-Enhanced Multi-Stage NMPC

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

In a previous work, multi-stage NMPC and tube-based NMPC schemes were combined into a single framework called tube-enhanced multi-stage NMPC with the goal of achieving an improved trade-off between simplicity and performance. In tube-enhanced multistage NMPC, the large uncertainties are handled using a multi-stage primary controller and the small uncertainties are handled using a multi-stage ancillary controller that tracks the predictions of the primary controller. In this work, we propose the replacement of the multistage ancillary controller by a single scenario NMPC that tracks the predicted trajectories of one of the scenarios of the multi-stage primary controller. The scenario that will be tracked by the ancillary controller as well as the ancillary controller model are time varying and are adapted to the current plant dynamics. The benefits of the new formulation are demonstrated on the benchmark Williams-Otto Continuous Stirred Tank Reactor (CSTR) example.

Abdelsalam, Subramanian, Engell:

Asymptotically Stabilizing Multi-Stage Model Predictive Control

59th Conference on Decision and Control - Jeju Island, Republic of Korea (CDC2020)

In this paper, we propose an asymptotically stabilizing formulation of multi-stage nonlinear model predictive control (NMPC) for plants with state and input dependent uncertainties. We derive time-varying Lyapunov-type sufficient conditions for asymptotic stability. We then propose a novel multi-stage NMPC formulation with time-varying terminal constraints, which guarantees asymptotic stability of the origin based on the derived stability conditions. The time-varying nature of the terminal constraints renders the control law and consequently the controlled system dynamics time-variant. For this situation, the derived time-varying stability conditions provide a suitable framework for stability analysis. We demonstrate the advantages of the proposed scheme over previous formulations of multi-stage NMPC on a cart simulation study.

Azadi, Ahangari Minaabad, Bartusch, Klock, Engell:

Nonlinear Prediction Model of Blast Furnace Operation Status

30th European Symposium on Computer Aided Process Engineering, VIRTUAL SYMPOSIUM, 2020

The operation status of a process in the steel industry is mainly defined by three aspects, efficiency, productivity and safety. It provides guidance for the operators to make decisions on their future actions. The abrasive process environment inside a blast furnace (BF) makes it demanding to analyse the operation status by direct internal measurements. The blast furnace gas utilization factor (ETACO) is an essential indicator of the process efficiency. Besides efficiency, productivity and safety can, to some extent, be derived from the pressure drop (DP) and the top gas temperature (TG). This paper presents a nonlinear autoregressive network with exogenous inputs (NARX) model for the simultaneous multistep ahead prediction of ETACO, DP and TG, based upon a new set of fast and slow dynamic input attributes. Validation results using real industrial plant measurements show that this approach not only enables monitoring of the current operation status but also provides prediction capability by including the slow dynamics of the blast furnace into the model.

Cegla, Engell:

Reliable modelling of twin-screw extruders by integrating the backflow cell methodology into a mechanistic model

30th European Symposium on Computer Aided Process Engineering, VIRTUAL SYMPOSIUM, 2020

Process modelling for twin-screw extruders is important for the optimal design, control and understanding of these machines. Existing models are often describing the residence time distribution (RTD) of the melt based on experimental data without the usage of further process knowledge. These completely data driven methods are unreliable for more advanced extrusion processes as a strong coupling between many internal states exists, which may not be reflected in the measurements. Therefore the use of a mechanistic model is beneficial to be able to address all important effects simultaneously. The standard mechanistic model describes the RTD as a series of continuous stirred tank reactors. However, this approximation is not capable of describing tailing effects that can occur when elements that promote distributive mixing are present. These effects can be described by the backflow cell model (BCM). Within the BCM the unidirectional flow is divided into an upstream flow and downstream flow with a fixed flow ratio for a series of tanks. This model can be in cooperated into the CSTR model, exploiting the similarities of the structure of the two models. In this work, the combination of the two methods is presented and applied to different screw geometries.

Ebrahim, Engell:

A Bi-level Approach to MPC for Switching Nonlinear Systems

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

In this paper, a nonlinear model predictive control scheme for switching dynamical systems is presented. The controller comprises of two layers of optimization. The upper layer is based on the embedding transformation technique, hence it does not require prior knowledge of the switching sequence. In particular, it provides the optimal relaxed switching sequence together with the optimal regulating inputs and the corresponding trajectories of the states. Within the lower layer, the integrality constraints are restored and a switching solution is recomputed to minimize the error with respect to the trajectories given from the upper-level optimization. The scheme is presented and the bounds of the integer approximation errors are evaluated together with brief recursive feasibility analysis. Simulation results of a tracking and an economics optimizing nonlinear MPC for a supermarket refrigerator system show the applicability and efficiency of the proposed approach.

Gerlich, Misz, Engell:

Online Process Monitoring in SMB Processes

30th European Symposium on Computer Aided Process Engineering, VIRTUAL SYMPOSIUM, 2020

Conventionally, preparative chromatographic separation processes are operated in batch mode. For more efficient separation, the simulated moving bed (SMB) process has been introduced. Due to its hybrid dynamics, optimal operation of the SMB process is challenging. For increased process efficiency, model-based optimizing control schemes can be applied. These schemes require online information about the states and the parameters of the plant. The online process monitoring strategy presented here is based on the transport dispersive model of the SMB process and simultaneously estimates the states and the parameters of the individual columns by exploiting the switching nature of the SMB process. The scheme can be activated before the process reaches its cyclic steady state (CSS). The strategy is demonstrated for the separation of two amino acids.

Gottu Mukkula, Engell:

Guaranteed Model Adequacy for Modifier Adaptation With Quadratic Approximation

2020 European Control Conference (ECC), Saint Petersburg, Russia, 2020

Iterative real-time optimization methods like modifier adaptation based techniques are used to identify the real process optimum in the presence of structural and/or parametric plant-model mismatch. For modifier adaptation based techniques, as a prerequisite, the a priori process model has to satisfy the model adequacy conditions, i.e. positive definiteness of the Hessian as only the gradients are corrected based on the available measurement information. Modifier adaptation with quadratic approximation (MAWQA) is a powerful real-time optimization tool in which either the surrogate quadratic approximation model or the a priori known process model with gradient correction is chosen for optimization based on how well they represent the process measurements. This can lead to the selection of a quadratic model which does not satisfy the model adequacy conditions and can cause undesired oscillations. It is therefore necessary to make sure that the chosen model is adequate. In this paper, we propose to use convex quadratic approximation and thereby to ensure satisfaction of the model adequacy conditions and the convergence to the real process optimum. We demonstrate the performance of the proposed MAWQA with guaranteed model adequacy (GMA) scheme using a chemical engineering case study.

Gottu Mukkula, Kern, Salge, Holtkamp, Guhl, Fleicher, Meyer, Remelhe, Maiwald, Engell:

An Application of Modifier Adaptation with Quadratic Approximation on a Pilot Scale Plant in Industrial Environment

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

The goal of this work is to identify the optimal operating input for a lithiation reaction that is performed in a highly innovative pilot scale continuous flow chemical plant in an industrial environment, taking into account the process and safety constraints. The main challenge is to identify the optimum operation in the absence of information about the reaction mechanism and the reaction kinetics. We employ an iterative real-time optimization scheme called modifier adaptation with quadratic approximation (MAWQA) to identify the plant optimum in the presence of plant-model mismatch and measurement noise. A novel NMR PAT-sensor is used to measure the concentration of the reactants and of the product at the reactor outlet. The experiment results demonstrate the capabilities of the iterative optimization using the MAWQA algorithm in driving a complex real plant to an economically optimal operating point in the presence of plant-model mismatch and of process and measurement uncertainties.

Gottu Mukkula, Valiauga, Fikar, Paulen, Engell:

Experimental Real Time Optimization of a Continuous Membrane Separation Plant

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

This paper deals with the optimal operation of a continuously operated laboratory membrane separation plant. The goal is to find an economically optimal regime of operation using the transmembrane pressure (TMP) and the operating temperature as adjustable set- points for the low-level controllers. The main challenge is to identify the optimum in the absence of an accurate process model. We employ an iterative real-time optimization scheme, modifier adaptation with quadratic approximation (MAWQA), to identify the plant optimum in the presence of the plant-model mismatch and measurement noise. Two experiments are performed; one with and one without a productivity constraint. The experimental results show the capabilities of the MAWQA scheme to identify the process optimum in real-world scenarios. The optimum identified by the MAWQA scheme coincides with the optimum of a surrogate model that was built using a larger data set.

Gottu Mukkula, Engell:

Iterative real-time optimization of the continuous production of amines in a thermomorphic solvent system

10. ProcessNet-Jahrestagung und 34. DECHEMA-Jahrestagung der Biotechnologen 2020

Thermomorphic solvent systems (TMSs)enable the realization of homogeneously catalyzed reactions with high selectivities and highly efficient catalyst recycling. The real-time optimization of the operation of a miniplant-scale continuous process for the production of amines using a TMS system is described. Modifier adaptation with quadratic approximation (MAWQA) is an iterative real-time optimization scheme, which drives the process to its optimum, despite model deficiencies by updating the optimization problem iteratively. MAWQA with guaranteed model adequacy (GMA) further improves the MAWQA scheme. It is applied for the real-time optimization of the amine production processin a TMS miniplant at TU Dortmund University.

Hernández, Onofri, Engell:

Optimization of the electric efficiency of the electric steel making process

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

This work reports numerical and practical results of an open-loop optimal control formulation that reduces the power consumption of the electric arc furnace (EAF) steel production process. A control vector parametrization technique is used to optimize the batch trajectory with the goal to minimize the energy losses of the process. First principles models are utilized to describe the dynamics, as well as the influence of the voltage and impedance set-points on the process. The results of the dynamic optimization provided a sequence of set-points (called a melting profile) that aligns well with intuition: the profile utilizes high power levels during the high efficiency stages of the process, and low power levels as the batch moves towards a more energy inefficient state. The benefits of the proposed optimized mode of operation are demonstrated by an experimental study case. An optimal melting profile was calculated and implemented in a fully operative ultra-high power EAF. For a series of 19 test batches, the energy consumption and the batch time of the process were reduced by 4.5% and 4.6% for one type of steel.

Janus, Lübbers, Engell:

Neural Networks for Surrogate-assisted Evolutionary optimization of Chemical Processes

2020 IEEE Congress on Evolutionary Computation (CEC), Glasgow, United Kingdom, 2020

In the chemical industry commercial process simulators are widely used for process design due to their extensive library of models of plant equipment and thermodynamic properties and their ease of use. Most of these simulators compute the steady-states of complex flowsheets, but their models are inaccessible and derivatives with respect to their model parameters are not available. Evolutionary algorithms are a suitable approach for the global optimization of such black-box models, but they require the evaluation of many individuals. Applications to industrial-size case-studies suffer from high computational times where the numerical simulations consume the majority of the time. This contribution proposes the use of neural networks as surrogate models to guide the evolutionary search. These models are trained multiple times during the evolutionary search and are used to exclude nonpromising individuals and to generate candidate solutions. We demonstrate the performance improvement due to the use of the surrogate models for a medium-size case-study of a chemical plant consisting of a reactor, a liquid-liquid separation and a distillation column. The results show that the required number of simulations can be reduced by 50%.

Kaiser, Engell:

Integrating Superstructure Optimization under Uncertainty and Optimal Experimental Design in early Stage Process Development

30th European Symposium on Computer Aided Process Engineering, VIRTUAL SYMPOSIUM, 2020

We present an iterative methodology that combines superstructure optimization, sensitivity analysis, and optimal design of experiments. In the early design phase, usually no accurate models for use in superstructure optimization are available, and the uncertainties of the models can influence the structure of the optimal design. The accuracy of the models is gradually increased by experimental investigations. In order to reduce the time and effort needed for process development, the experiments should focus on the most influencing parameters with respect to the design decisions. After one or few process structures have been fixed, further experiments will then lead to quantitatively accurate predictions. The methodology is applied to the case study of the hydroaminomethylation of decene.

Klanke, Yfantis, Corominas, Engell:

Scheduling of a Large-scale Industrial Make-and-Pack Process with Finite Intermediate Buffer using Discrete-time and Precedence-based Models

30th European Symposium on Computer Aided Process Engineering, VIRTUAL SYMPOSIUM, 2020

We address the short-term scheduling of a two-stage continuous make-and-pack process with finite intermediate buffer and sequence-dependent changeovers from the consumer goods industry. In the current layout of the plant under consideration, the two stages, product formulation and packing, are directly coupled, i.e. the products of the formulation stage go directly to the packing stage. As for different products either one of both stages can be the bottleneck, a gain in productivity can be obtained if the two stages are decoupled by a buffer so that the formulation lines and the packing lines can both run at full capacity. The disadvantage of this setup is an increased complexity of the scheduling problem, so that support for the schedulers must to be provided. We employ a mixed-integer programming problem formulation for this purpose. The problem at hand is characterized by a large number of products in several product families, product specific order quantities and deadlines, product dependent production times, sequence-dependent changeover times, and a finite intermediate buffer. As the problem turned out to be intractable for the planning horizons of interest, a solution approach that employs a discrete-time scheduling model, a precedence-based presorting model and a decomposition strategy that is enhanced by several heuristics was developed.

Leo, Engell:

A Novel Multi-stage Stochastic Formulation with Decision-dependent Probabilities for Condition-based Maintenance Optimization

30th European Symposium on Computer Aided Process Engineering, VIRTUAL SYMPOSIUM, 2020

The challenge addressed in this work is the integrated production planning and condition-based maintenance optimization for a process plant. We take into account uncertain information of the predicted equipment degradation adopting a stochastic programming formulation. To adjust the likelihood of the failure scenarios, we embed a prognosis model, the Cox model, into the optimization problem. We propose here a novel endogenous uncertainty formulation where the decisions at one point in time have an impact on the probability of the uncertainty. We provide computational results implementing a custom branching within the global solver BARON and decomposing the problem via the Benders algorithm.

Schmiermoch, Beisheim, Rahimi-Adli, Engell:

A Methodology for Data Based Root-cause Analysis for Process Performance Deviations in Continuous Processe

30th European Symposium on Computer Aided Process Engineering, VIRTUAL SYMPOSIUM, 2020

The surge of computational power and the increasing availability of data in the process industry result in a growing interest in data based methods for process modelling and control. In this contribution a concept is described that uses statistical methods to analyse the root-causes for deviations from baselines that are used for the monitoring of the resource efficiency of the production in dashboards. This is done by comparing historical data during resource efficient operation under similar process conditions with data during inefficient operation. Statistically significant deviations are identified, sorted by the likelihood of causing the performance deviation. The concept is applied to a reference model called Best Demonstrated Practice which is in use at INEOS in Cologne. It represents the most resource efficient process performance under given conditions. Deviations from efficient plant performance are analysed using the described concept and the results are given to the operators as a decision support tool, including reference values for the degrees of freedom under the control of the operators. This concept is already used in a root-cause analysis tool at INEOS in Cologne and detected energy savings of over 20% for specific cases.

Tatulea-Codrean, Fischer, Engell:

A Multi-stage Economic NMPC for the Tennessee Eastman Challenge Process

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

This paper addresses the design and implementation of a robust nonlinear model predictive control (NMPC) scheme for a benchmark plant-wide control problem. The focus of our research is on the performance of direct optimizing control for a complex large-scale process which is subject to plant-model mismatch and external disturbances. As a benchmark case for control and monitoring applications, the Tennessee Eastman Challenge (TEC) process has been widely employed in many publications. We present a first NMPC implementation for this where only economics criteria are used for the control of the process. The results obtained demonstrate the viability of plant-wide economics optimizing NMPC. We also address the issue of robustness against model uncertainties and employ multi-stage NMPC to tackle these. Different possible multi-stage NMPC implementations are discussed and the trade-offs between economic performance and robustness are highlighted.

Tatulea-Codrean, Mariani, Engell:

Model Predictive Control Approach to Autonomous Race Driving for the F1/10 Platform

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

This paper addresses the challenges of developing an embedded non-linear model predictive control (NMPC) solution for the optimal driving of miniature scale autonomous vehicles (AVs). The NMPC approach lends itself perfectly to driving applications, provided that a system for localization and tracking of the vehicle is available. An important challenge in the implementation results from the need to accurately steer the vehicle at high speeds, which requires fast actuation. In this paper we present a solution to this problem, which employs an artificial neural network (ANN) controller trained with rigorous NMPC input-output data. We discuss the development process, from modelling until the realization of the ANN controller within the operating system of the AV. The procedure is demonstrated within the virtual environment of the popular F1/10 race car, an AV platform widely used in AI and autonomous driving challenges. The results contain both NMPC and ANN-based simulations for different race tracks and for different driving strategies. The main focus of this work lies in the formulation of the optimal driving control problem and the training method of the ANN. Our approach uses a standardization of the driving problem, which enables us to abstractize optimal driving and to simplify it for the learning process. We show how driving patterns can be learned accurately on a reduced set of training data and that they can subsequently be extended to new and more challenging driving situations.

Thangavel, Paulen, Engell:

Multi-stage NMPC using sigma point principles

2020 Advances in Control & Optimization of Dynamical Systems, Chennai, India, 2020, Full Paper

A novel non-conservative robust nonlinear model predictive control scheme (NMPC) based on the multi-stage formulation is introduced for the case of an ellipsoidal uncertainty set. Multi-stage NMPC models uncertainty by a tree of discrete scenarios. In the case of a continuous-valued uncertainty, the scenario tree is usually built for all combinations of the minimum, nominal and maximum values of the uncertainty. If the uncertainty set is ellipsoidal, the standard multi-stage NMPC augments the uncertainty set which results in a performance loss while using the robust NMPC approaches. We propose to mitigate this problem by tightly over-approximating the uncertainty set using the so-called sigma points. An ellipsoidal over-approximation of the reachable set of the system is predicted along the prediction horizon using the unscented transformation. The advantages of the proposed scheme over the traditional multi-stage NMPC are demonstrated for a benchmark semi-batch reactor case study.

Thangavel, Paulen, Engell:

Adaptive multi-stage NMPC using sigma point principles

2020 European Control Conference (ECC), Saint Petersburg, Russia, 2020

A novel robust nonlinear model predictive control scheme (NMPC) based on the multi-stage formulation is introduced in this paper. The scenario tree of Multi-stage NMPC is often built by assuming parametric uncertainty and by considering the minimum and maximum values of the parameters. This can augment the uncertainty set and can result in a performance loss, if more information is available, e.g. in a form of a parametric confidence region. We propose to mitigate this problem by tightly over-approximating the uncertainty set using the so-called sigma-points parameterization. In addition, the plant measurements are used to reduce the size of the uncertainty set and the scenario tree is updated online. The advantages of the proposed scheme over the traditional multistage NMPC are demonstrated for a benchmark case study.

Thangavel, Paulen, Engell:

Dual multi-stage NMPC using sigma point principles

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

Dual control is a technique that addresses the trade-off between probing (excitation signals) and control actions, which results in a better estimation of the unknown parameters and therefore in a better (tracking or economic) performance. Multi-stage NMPC is a robust-control scheme that represents the uncertainty using a scenario tree that is often built by assuming parametric uncertainty and by taking into account the minimum, nominal and maximum values of the uncertain parameters. If the uncertainty set is not a box, this procedure augments the uncertainty set and results in a loss of performance. Here, we mitigate this problem by tightly approximating the uncertainty set using the so-called sigma points and computing an ellipsoidal over-approximation of the reachable set of the system using the unscented transformation. We also improve the performance by considering the future reduction of the ranges of the uncertainties due to control actions and measurements thereby achieving implicit dual control actions. The advantages of the proposed approach over the standard multi-stage NMPC scheme are demonstrated for a linear and a nonlinear (semi-batch reactor) simulation case study.

Thangavel, Engell:

An efficient model-error model update strategy for multi-stage NMPC with model-error model

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

Multi-stage NMPC with model-error model (MS-MEM) handles structural plant-model mismatch present in the nominal model of the plant in a non-conservative fashion. A model-error model (MEM) that consists of a stable linear time-invariant dynamics and a static time-variant nonlinear mapping is built using the past data such that it captures the unmodeled dynamics of the plant. The scenario tree is built for the nominal and for the extreme realizations of the plant obtained using the nominal model and the model-error model, and a multi-stage decision problem is formulated. In this paper, we propose an efficient strategy to update the model-error model present in the MS-MEM approach if new measurements invalidate the model-error model. The advantages of the proposed scheme over the previous approach where only the gain of the linear model is updated are demonstrated for a continuous stirred tank reactor (CSTR) benchmark example.

Wenzel, Riedl, Engell:

Market-like Distributed Coordination of Individually Constrained and Coupled Production Plants with Quadratic Approximation

30th European Symposium on Computer Aided Process Engineering, VIRTUAL SYMPOSIUM, 2020

The optimal operation of large chemical or petrochemical production sites is challenging, because streams of shared resources such as steam or intermediates physically couple the individual production plants. For a feasible operation, it is essential to coordinate the production of the plants to balance the networks for the shared resources. Often, finding a central solution to this site-wide optimization problem is not possible due to various barriers such as the lack of models, too large complexity, or the management structure that hinders the exchange of relevant information. Market-like distributed optimization tackles some of these barriers, especially if only limited data exchange is possible between the constituent production plants. However, distributed optimization algorithms typically need many iterations, which hinders their practical application. In this contribution, we compare an algorithm for distributed market-like coordination that we formulated in previous work for individually constrained coupled systems with ADMM for a novel case study.

Yfantis, Büscher, Klanke, Corominas, Engell:

A Two-stage Simulated Annealing-based Scheduling Algorithm for a Make-and-Pack Production Plant

21st IFAC World Congress 2020 – Berlin, Germany (IFAC2020)

tba

Master's Theses 2020

Mohammed Akani

Economic operation of a COBR for reactive Zeolite crystallization using NMPC

supervised by: Robin Semrau

Zeolites are crystalline alumino-silicates materials. Due to possessing large surface area and pore volumes, zeolites can be used as catalysts, adsorbents and ion exchangers. Zeolites are conventionally produced by hydrothermal treatments which is usually carried out in batch reactors due to long synthesis time of zeolite crystallization. In this work, the production of zeolites in a continuous oscillatory baffed reactor (COBR) is considered. Continuous production of zeolite offers the advantage of lower variation in product quality, while decreases exibility of the plant. In order to enhance the exibility of the continuous reactor, the usage of nonlinear model predictive controller (NMPC) is investigated. As the first step, the performance of different discretization schemes is compared to each other. Subsequently, in order to control the crystallinity of the product, a NMPC scheme is developed and validated in simulation studies. Regarding the continuous processes, improvement of operational effciency is inevitable. With this purpose, the optimal steady state operation of the process is explored. Moreover, three different economic NMPC formulations, including tracking the optimal set points, economics optimizing control, and economics oriented tracking control, are implemented and tested in simulation studies. Finally, the economic performance of the controllers are compared to each other by considering a condition with frequent disturbances in the energy price. This work indicates that only tracking the optimal set points cannot provide an effcient performance, while considering the economic function directly in the controller objective, significantly improves the economic behavior of the system.

Hind Arab:

Bringing together estimation and optimizing control in simulated moving bed (SMB) processes

supervised by: Stefanie Gerlich

Preparative chromatography is important in the chemical industry for separation applications involving sensitive and physically similar components. The need to apply this chromatography in large-scale industries has driven the introduction of continuous chromatography. The simulated moving bed (SMB) process is a continuous chromatography having a counter-current flow between the mobile and the solid phase. The counter-current flow is approximated by periodically switching the valves between the columns which result in steep and nonlinear concentration profiles. However, these characteristics make the process modeling and optimization of the SMB process challenging. In order to optimize the economics of this challenging process, the do-mpc framework is adopted containing four modules: NMPC module, optimization-based estimation module, SMB process model module using the transport dispersive model (TDM) and pilot plant interface module. The goal of this work is to apply this scheme to real process data. First, preliminary experimental steps are conducted and afterwards an experimental application of the estimation scheme is conducted using real process data of an SMB experiment. Analyzing the effect of the initial values used in the estimation, it is concluded that initial values that are closer to the real values improve the speed of the estimation and hence the ability to perform more estimations. The second run of an estimation using the updates of the first run reduces the computational time by around 60 %. The effect of the parameter regularization term in the objective function is also studied finding that using small values increases the penalty on the difference between the update and the initial value. This results in slowing down the estimation when the difference is expected to be high.

Amitha Asokkumar:

Model Predictive Control of Molten Iron Quality Indices in Blast Furnace Ironmaking using a Data-Driven Model

supervised by: Pourya Azadi

The stable operation and control of a blast furnace have been a long-standing challenge in the iron and steel industry. Decreasing ore quality, increasing environmental regulations, and most importantly, the need for consistent hot metal quality demand advanced control solutions for sustainability and operational profitability of the blast furnace. Existing strategies are either inefficient due to limited online sensors, difficulty in direct measurements, irregular hot metal quality data, and/or inadequate to capture varying process response time and operating conditions. Through this work, a predictive neural network model using Nonlinear AutoRegressive with eXogenous inputs (NARX) architecture has been developed to be implemented in a model predictive control (MPC) scheme. The model focuses on accurately predicting the hot metal silicon content, a critical indicator of blast furnace thermal state, for an adequately long prediction horizon of 6 to 8 hours, by incorporating the multiscale dynamics of the process. The validation results of the model on plant measurements from thyssenkrupp Steel Europe showed the reliability of the approach. A preliminary framework of MPC using the model for tracking the desired set point of silicon content in hot metal has also been presented.

Dominik Bleidorn:

Constraint Programming and Evolutionary Algorithm approaches to the scheduling of a make-and-pack process with finite intermediate storage

supervised by: Christian Klanke

In this thesis the practical application of Constraint Programming model to a realistic scheduling problem is evaluated. A model is derived for a two-stage make-and-pack plant with optional storage units and tested on a small and a large scale instance of the case study. Beyond the basic Constraint Programming model multiple extensions to cope with the combinatorial complexity are developed. First experiments revealed that a monolithic approach, in which all decisions are made by the Constraint Programming model, can solve the smaller case study to optimality. However, as the monolithic Constraint Programming approach exhibited computationally intractable solution times for the large case study, different decomposition and hybridization approaches were investigated. One such extension is a Hybrid Metaheuristic, in the form of an Evolutionary Algorithm, coupled with the Constraint Programming model. The results show that this method is not suitable for the large scale instance, either. Finally, a Temporal Decomposition approach, in which partial schedules are concatenated, accomplishes the goal of providing decent solutions for the large scale instance and provides a method, which is further scalable. The results of the small case study are compared to the results of the work of Yfantis [1], from which the case study originates. When comparing the MILP approach from Yfantis, with the solution developed in this thesis, it became evident that the Constraint Programming-based approach yields results of the same quality, yet in a reduced amount of time.

Robin Ehrhardt:

Efficient Scheduling of Batch Production Plants with Systematic Use of Scheduling Heuristics

supervised by: Christian Klanke

Optimal production scheduling is a decision problem, for example to improve the productivity of manufacturing processes. This thesis addresses the scheduling problem of a lab-scale pipeless plant that resembles the classical job shop problem, that has shown to be a difficult optimization problem. The proposed approach uses an Evolutionary Algorithm (EA) based on a hyperheuristic. Hyperheuristics abstract the search space by trying to find combinations of low-level heuristics instead of optimizing the degrees of freedom in the optimization problem directly. In the proposed implementation, a priority rule-based genome is utilized to decide the sequence of operations to be executed instead of encoding the sequences directly. To generate schedules from the encoding, the case study is modeled in the simulation software INOSIM in the form of discrete event simulation (DES). The results will be quantified by comparing with an exact timed-automata based approach provided by Schoppmeyer (2015) and schedules generated under default behavior of INOSIM. The approach will be evaluated according to reproducibility of results and sensitivity towards individual elements of the encoding.

Zheng Li:

Genetic Algorithm-based Scheduling for Consumer Goods Industry

supervised by: Christian Klanke

Nowadays, consumers’ demands for more product varieties and shorter delivery times impose a big challenge for the manufacturers in the fast-moving consumer goods industry. The organization of production activities has to be renovated to respond to consumers’ demands more efficiently and precisely. In a typical make-and-pack production process, sequence-dependent changeovers and storage constraints play a major role in short-term scheduling. This thesis deals with the production scheduling of a two-stage formulation and packing plant. The goal is to minimize the makespan and the total changeover time in the shop floor schedules. A genetic algorithm is proposed in this thesis to tackle the scheduling problem with limited intermediate storage capacity. A repair algorithm and a local search strategy are implemented. Constraint handling is performed by a corresponding scheduler. The results show that near-optimal schedules can be obtained within reasonable computational time. This thesis has proven that the GA can be applied to solve such kind of scheduling problem when there is limited computational time and only near-optimal solutions are required.

Henrik Minten:

Data-driven modelling to calculate energy efficient indicators of batch processes.

supervised by: Tim Janus, Maximilian Cegla, Marc Kalliski, and Rubin Hille

This master thesis deals with two challenges regarding energy efficiency in production processes. Firstly, the basis for optimising the energy efficiency of a plant is outlined. Two methods for determining the improvement potential and for mapping indicators are combined in a novel way to create a plant-wide indicator that describes the optimisation potential by the plant operator. To illustrate the proposed approach, the methods are subsequently applied to an example process. Complications resulting from inconsistent process data are addressed by means of suitable procedures for determining missing measurements and adjusting the mass balance. The indicators that result from applying the proposed approach are used to determine the efficiency of the process and to evaluate the impact of procedural changes. Furthermore, a digital tool is developed to analyse and track the efficiency of individual batches, enabling the user to perform a root-cause analysis for suboptimal batches. This requires a carefully designed method to aggregate and propagate the performance indicators. The second challenge being addressed is the estimation of unmeasured energy consumptions. The distribution of energy consumption across subsystems plays an important role in identifying the potential for improving overall efficiency. In the proposed estimation approach, a linear optimisation is solved based on three formulations of the regression model with different levels of detail: (1) using only the information if the units are idling or operating, (2) adding information on the batch size processed, and (3) dividing the operating mode in the individual step numbers. Overall, the challenges considered have been successfully resolved. The combination of the known methods enabled a more precise analysis of the potential for improvement. In addition, the procedure for determining unmeasured consumption was demonstrated and validated using a process example.

Filippo Tamagnini:

EKF based State Estimation in a CFI Copolymerization Reactor including Polymer Quality Information

supervised by: Robin Semrau

State estimation is an integral part of modern control techniques, as it allows to characterize the state information of complex plants based on a limited number of measurements and the knowledge of the process model. The benefit is twofold: on one hand it has the potential to rationalize the number of measurements required to monitor the plant, thus reducing costs, on the other hand it enables to extract information about variables that have an effect on the system but would otherwise be inaccessible to direct measurement. The scope of this thesis is to design a state estimator for a tubular copolymerization reactor, with the aim to provide the full state information of the plant and to characterize the quality of the product. Due to the fact that, with the existing set of measurements, only a small number of state variables can be observed, a new differential pressure sensor is installed in the plant to provide the missing information, and a model for the pressure measurement is developed. Following, the state estimation problem is approached rigorously and a comprehensive method for analyzing, tuning and implementing the state estimator is assembled from scientific literature, using a variety of tools from graph theory, linear observability theory and matrix algebra. Data reduction and visualization techniques are also employed to make sense of high dimensional information.The proposed method is then tested in simulations to assess the effect of the tuning parameters and measured set on the estimator performance during initialization and in case of estimation with plant-model mismatch. Finally, the state estimator is tested with plant data.

Bachelor's Theses 2020

Malte Buchholz:

Online parameter estimation in simulated moving bed (SMB) processes

supervised by: Stefanie Gerlich

As an improvement to batch chromatography, the continuous simulated moving bed (SMB) process can be employed to reach a better productivity. To ensure minimum operating cost and the fulfillment of the product requirements, modelbased optimizing control schemes can be applied to the SMB process. A special challenge regarding this process is the scarce amount of available measurements. Since model-based optimizing control requires current and accurate model parameters as well as information about process states, state and parameter estimation schemes can be applied. Suitable state and parameter estimations that allow for estimating each column’s parameters individually have been developed and tested in simulations studies in previous works. The aim of this thesis is to evaluate the results of the state and parameter estimation scheme for a long SMB experiment. The challenge in application of the scheme to a pilot plant process is the need to consider effects from extra column equipment. The results of the state and parameter have shown that the estimation scheme does not converge. The handling of the extra column equipment has been identified as the most probable cause. Modeling a PFR that includes all effects is not detailed enough, especially the assumption that all influences of the extra column equipment are accumulated at the outlet ports can lead to deviations from the true behaviour of the plant, as the delay between individual columns possibly differs. When the effects are not considered detailed enough, the process model becomes inaccurate, which leads to the failure to converge of the state and parameter estimation scheme.

Tabea Menzel:

Comparison of local and global sensitivity analysis as a tool for model based process Development

supervised by: Stefanie Kaiser

Good chemical process models are a crucial part of chemical engineering. They can help choosing the dimensions of different equipment required in a chemical plant. A model can improve the understanding of the process. They can also replace expensive experiments and testing in a chemical plant. These advantages of a process model can only be used if the model itself is as close to reality as possible. A sensitivity analysis can contribute to the goal of improving the process model. It can be used to identify the most influencing parameters and can thus help to focus the experimental work on the important parameters. This can reduce the experimental effort that is needed to build a reliable process model. But improving a process model with a sensitivity analysis is only possible if the right sensitivity analysis is chosen. In this thesis the focus lies on two different sensitivity analysis methods: A local sensitivity analysis using linear regression and the Sobol method, which is a global variance-based sensitivity analysis method.

News New dyn doctors

CoPro Final Meeting

Geographically far away, digitally right there: participants from all over Europe and even beyond took part in the long-planned CoPro closing event via video conference, mostly from their home offices. Despite the Covid-19 outbreak, the EU research and innovation project celebrated its successful completion –and without any glitches or technical problems. More than 150 participants from 14 countries joined in on the internet over two days. To kick off the event, project coordinator Prof. Dr.-Ing. Sebastian Engell gave a positive assessment: “With the help of newly developed tools and software, the potential of industrial plants can now be better exploited, coordination can be optimized and the support for plant managers can be improved. CoPro has achieved its goals.”

In simultaneous poster sessions taking place on the first day of the event, the participants were able to learn something new, among other things, about dynamic planning of production processes, optimization of evaporators and heat exchangers and IT integration within the scope of the CoPro project. The CoPro sister projects COCOP, FUDIPO, MONSOON and MORSE were also presenting the research that is conducted throughout the EU in various constellations in order to optimize processes and create a more sustainable industry. All with the two major goals in mind: protect the environment and save resources.

The second day of the event was conducted under the headline of “Digitalisation for Sustainability” and led by Dr. Stefan Krämer, Bayer AG. Amongst others, insightful presentations by Karl-Uwe Bütof, Ministerialdirigent of the Ministry of Economy, Innovation, Digitalization and Energy of North-Rhine Westphalia, Dr. Klaus Schäfer, Chief Technology Officer of Covestro AG, and Jane Arnold, Head of Process Control Technology of Covestro AG were included. The meeting was concluded by presentations on the future of the sustainability of the european process industry, a glance at the SPIRE Rodmap and a vivid Panel Discussion on the contribution of digitalisation and optimization for more sustainability in the process industries.

For an overview of the scientific outcomes of the CoPro project, please visit the Research Gate profile of CoPro.

Minister Pinkwart übergibt Förderbescheid für OptiProd

Von rechts: Minister Pinkwart, Dr. Christian Sonntag, INOSIM GmbH, Manuel Remelhe, Bayer AG und Prof. Sebastian Engell

Prof. Dr. Andreas Pinkwart, Minister für Wirtschaft, Innovation, Digitalisierung und Energie des Landes NRW übergab am 15.1.2020 Förderbescheide an 11 Projekte, die ab Januar 2020 im Rahmen des Leitmarktwettbewerbs IKT.NRW gefördert werden. Hierunter ist auch das Projekt OptiProd NRW der Firmen INOSIM GmbH, Dortmund und Bayer AG, Leverkusen, und des Lehrstuhls für Systemdynamik und Prozessführung der Fakultät BCI der TU Dortmund. Ziel des Projekts ist die Optimierung von Produktionsplänen für Chargenprozesse in der chemischen und pharmazeutischen Industrie anhand detaillierter „digitaler Zwillinge“. Diese digitalen Zwillinge der Produktionsanlage sorgen dafür, dass Pläne direkt ohne händische Anpassungen in der Produktion nutzbar sind. Die 11 Projekte wurden aus 44 Anträgen ausgewählt.

NAMUR Award for Simon Wenzel

1 / 2

Simon Wenzel attending the virtual ceremony

2 / 2

Summary of his dissertation

❮ ❯

On the NAMUR Annual General Meeting 2020, DYN alumni, Dr.-Ing. Simon Wenzel has been awarded the NAMUR-Award 2020 for his dissertation entitled "Distributed optimization of coupled production systems via market-like coordination". The work leading to this dissertation was done in the context of the CoPro project coordinated by Prof. Engell from November 1st 2016, for the duration of 42 months. NAMUR is the international user association of automation technology in the process industries. The main goal of the association is to further develop the field of process automation by gathering expertise and proposing new ideas across companies. The Annual General Meeting of NAMUR is a congress for members and invited guests of NAMUR. It is attended by about 650 participants, predominantly from the industry. NAMUR awards one PhD student per year for his or her outstanding contribution that addresses intelligent process control and operational management or other elements of process automation. Due to the Corona pandemic, the award ceremony could only take place vritually. We congratulate our former colleague Simon Wenzel for winning this outstanding award!

Sergio Lucia new Professor for Process Automation Systems

On Oct. 1, 2020 Dr. Sergio Lucia started as the new professor for Process Automation Systems in the BCI department. Dr. Lucia got a Dr.-Ing degree from TU Dortmund in 2014 and thereafter was a PostDoc at Otto von Guericke-Uiversität Magdeburg before he was appointed as a Junior Professor at the Einstein Center Digital Future at TU Berlin in 2017. With him, three junior researchers will move from TU Berlin to TU Dortmund. In the coming semesters, we will share the teaching responsibilities in the field of system dynamics, control, optimization and process operations for the Bachelor and Master programs of the BCI department and the Master program in Automation and Robotics. A warm welcome to Professor Lucia!

Excursion to INEOS

Having successfully completed a group project titled “Projektierung einer Anlage zur gewinnbringenden Nutzung von Kohlenstoffdioxid aus den Rauchgasen eines Steamcrackers” (Design of a plant for the profitable utilisation of carbon dioxide from the flue gases of a steam cracker), supervised by Stefanie Kaiser and Stefanie Gerlich from the DYN group, the students visited their industrial partner INEOS in Cologne on January 14, 2020. To be more precise, the group was introduced to INEOS and visited their steam cracker. As a conclusion of the day and the group project, the students presented their results at the company followed by some interesting discussions.

PSE project group presentation at Merck

The group from the international study program "Process Systems Engineering" successfully completed the group project in the winter semester 2019-20, and had the chance to present and discuss the final results with the audience at Merck KGaA on February 21, 2020. The project on "Continuous Production of High-Quality Bis(tertiary-butyl-amino)silane in a Modular and Automated Plant" was proposed by Merck KGaA and supervised by Pourya Azadi and Anwesh Reddy Gottu Mukkula from the DYN group.

Workshop on "Deep Learning" – definition and application to process industry

In the scope of the PhD workshop “Deep Learning – What is deep learning and how can it be applied to the process industry”, several PhD students of the biochemical and chemical engineering department prepared talks to introduce their work with deep learning/machine learning/artificial intelligence methods. As a guest, Dr.-Ing. Fabian Bürger from the automotive corporation, Valeo attended the workshop. Dr.-Ing. Bürger received his PhD degree in the area of automatic optimization for machine learning methods. The talks led to several interdisciplinary discussions between visitors of the faculties statistics, finance, logistics, sports, computer science and biochemical and chemical engineering.
Talks:

Fabian Bürger - What is deep learning and its application to autonomous driving
Matthias Rodeck - Characterization of sprays by image recognition with neural networks
Janine Lins - Particle characterization during crystallization using CNNs
Corina Nentwich - When data is expensive: Adaptive Sampling for the training of surrogate models
Fabian Bürger - How to apply deep learning to other fields than image recognition
Tim Janus - Surrogate-assisted flowsheet optimization for steady-state models
Kai Kruber - Optimization of intensified distillation columns with machine learning methods
Pourya Azadi - Saving energy in a blast furnace by application of neural networks

A&R group project presentation

From left: Taher Ebrahim, Sowmiya Angamuthu, Monika Rajagopal Balamurugan, Kiran Nivrutti Borse, Anay Ghatpande, Khaled Elewa, Marina Rantanen Modeer, Egidio Leo (Source: private)
During the winter semester 2019/20 a group of MSc students from the Automation and Robotics program worked on their group project titled “Optimization and Scheduling of a Multi-robot Production Plant”. The group developed an optimal scheduling scheme for the operations of an experimental, pipe-less plant for chemical batch production under the supervision of Taher Ebrahim, Marina Rantanen Modeer and Egidio Leo. The project set out to optimize the plant operations carried out by a set of Automated Guided Vehicles transporting vessels between different production stations. The group successfully applied the Resource Task Network approach to create a schedule and further used a model-based approach for dynamic trajectory planning of the vehicles. The project was presented on Monday, 9 March 2020, for the DYN group.

Visit to DYN by Laurent Dewasme

Laurent Dewasme, M.Sc.Eng. and Ph.D., visited the DYN group as a guest researcher in August and September 2020. He is currently Senior Researcher and Substitute Part-Time Lecturer in the Systems, Estimation, Control and Optimization (SECO) Group of the University of Mons (UMons), Deputy Coordinator of the BioSyS Research Center and Secretary of the Institute for Biosciences of UMons. His research interests include modelling, parameter and state estimation, and optimizing control of (bio)processes. While at the dyn group, he worked on the application of real-time optimization (RTO) algorithms to chemical and biochemical process models, more particularly in the following topics:

Application of perturbation-based gradient estimation (Extremum Seeking Control), in the framework of necessary conditions of optimality (NCO) tracking, to chemical plants;
Modifier adaptation to bioprocess models with increasing complexity.

We wish him all the best in his career and look forward to his future delightful research outcomes.

Doctoral degree awarded to Simon Wenzel

On May 6th, 2020, Simon Wenzel finished the procedure for his doctoral degree with the oral examination at TU Dortmund University. The examinations took place under Corona conditions with only the examiners present in a lecture hall and the public attending via the internet. Thanks to the great preparation by Maximilian Cegla and Tim Janus, the web-based format worked out very well.

Simon Wenzel, from 2014 to 2019 a member of the DYN group at TU Dortmund and now with Evonik Technology & Infrastructure GmbH obtained the Dr.-Ing degree for his dissertation “Distributed optimization of coupled production systems via market-like coordination”. The main contribution of the thesis is a new algorithm for the adaptation of the prices when different coupled chemical plants within a site or chemical park are coordinated without sharing details of the individual plants and their loads, costs, etc. His dissertation results from the EU project CoPro that was coordinated by Prof. Engell. Congratulations to Simon Wenzel!

Doctoral degrees awarded to Benedikt Beisheim

1 / 2

Benedikt Beisheim defending his PhD thesis

2 / 2

Examination Committee of Benedikt Beisheim

❮ ❯

On May 8th, 2020, Benedikt Beisheim (INEOS in Köln) obtained the Dr.-Ing degree for his dissertation “Computation and Application of Resource Efficiency Indicators for Continuous Processes”. The main contributions of the thesis are an approach to compute the best-demonstrated practice baselines of chemical plants from observed data and a method for the aggregation of resource efficiency indicators and the analysis of the causes for their evolution over time. The oral exam took place on May 8, 2020. After the doctoral exam of Benedikt Beisheim, he and the Examination Committee enjoyed drinks from cans outside the building in perfect weather, obeying the rules for social distancing. His theses resulted from the EU project CoPro that was coordinated by Prof. Engell. Congratulations to Benedikt Beisheim!

Doctoral degree awarded to Lukas Hebing

Lukas Hebing sucessfully defended his dissertation titled "Modeling and Control of Fermantation Processes" on June 22, 2020. The main contribution of his thesis is a new approach for dynamic process models and robust online optimizing control for bioprocesses. Congratulations to Lukas Hebing!

Doctoral degree awarded to Lukas Samuel Maxeiner

Lukas Samuel Maxeiner, from 2014 to 2020 a member of the DYN group at TU Dortmund and now with Evonik Technology & Infrastructure GmbH obtained the Dr.-Ing degree for his dissertation “Dual-based methods for distributed optimization of interconnected systems”. After the doctoral exam, he and the Examination Committee celebrated outside the building, obeying the rules for social distancing. The thesis resulted from the EU project CoPro that was coordinated by Prof. Engell. Congratulations to Lukas Samuel Maxeiner!

Doctoral degree awarded to Sankaranarayanan Subramanian

Sankaranarayanan Subramanian sucessfully defended his dissertation titled "Tube-enhanced Multi-stage Model Predictive Control: Robust State and Output Feedback Control" on October 27, 2020. The principal theme of the thesis was addressing the challenge of the trade-off between optimality and complexity in the field of robust linear and nonlinear Model Predictive Control. Congratulations to Sankaranarayanan Subramanian!

30th Anniversary of the dyn Group

The DYN group was founded on August 1, 1990 and celebrated its 30th birthday in 2020. A special newsletter on the occasion of the 30th anniversary of the group foundation was published in August summarizing the activities of the dyn group in the past 30 years.

Please click here to view the Newsletter.