Process and plantwide control

Industrial use of advanced process control increases rapidly, and candidates who combine process knowledge and control expertise are in high demand in industry. Control is an enabling technology, thus basic for any industry-based society. The use of advanced control is transforming industries previously regarded as "lowtech" into "high-tech". In process control (Skogestad, Preisig, Jäschke), the objective of the research is to develop simple yet rigorous tools to solve problems significant to industrial applications (of engineering significance).

Up to now, the design of the overall "plant-wide" control structure has been based on engineering experience and intuition, whilst the aim has been to develop rigorous techniques. The concept of "self-optimizing control" provides a basis for linking economic optimization and control (Skogestad). For example, for a marathon runner, the heart rate may be a good "self-optimizing" variable that may be kept constant in spite of uncertainty. Control is done in a hierarchical construct. At the bottom of the hierarchy, the main issue is to "stabilize" the operation and follow the setpoints provided by the layer above.

Further up in the hierarchy one finds optimising control co-ordinating the control of units and plants. A special case is sequential control, which is used to implement recipes in batch operations but also is the basics of handling start-up and shut-down as well as all fault and emergency handling. Another important concept is controllability, which links control and design. Here the main focus is on applications, which currently include reactor and recycle processes, distillation columns, gas processing plants, cooling cycles including liquefied natural gas (LNG) plants, low-temperature polymer fuel cells and anti-slug control.

Process modelling

The fourth generation of a high-level modelling tool is presently being developed (Preisig), which we aim to apply to large-scale plants, including the Mongstad refinery. It incorporates object-oriented tools for efficient thermodynamic modelling, which extend into the efficient computation of thermodynamic information. Rather than a traditional implementation of activity or fugacity coefficients, emphasis is put on the use of structured equation sets governed by thermodynamic consistency rules (Haug-Warberg). The thermodynamic models are implemented in symbolic form with automatic differentiation capabilities and serves as the basis of several industrial strength simulations (YASIM, CADAS) and energy accounting tools (HERE) in cooperation with Norsk Hydro and Yara. A primary aspect of thermodynamic (and other physics) modelling is the required consistency of physical units. We have a procedure to obtain self-consistent models, including automatic generation of gradients. This technique has so far been tested up to sixth order gradients, which are needed for higher-order critical point calculations.

AI-Powered Systems Engineering for Sustainable and People-Centered Solutions

This research (Nogueira) integrates artificial intelligence, advanced process control, systems optimisation, digitalisation, and automation to develop transformative solutions for industry and society. The approach, termed AiP2S2 (Artificial Intelligence-powered Products, Processes, Scales, and Systems), combines AI-driven methodologies with process systems engineering to enhance efficiency, robustness, and sustainability across various applications.

Research Themes and Applications:

1. AI for Intelligent Process Control & Optimization: Complex industrial processes operate under dynamic conditions, requiring advanced control strategies to ensure efficiency, stability, and adaptability. Research in this area focuses on integrating AI with control theory to develop hybrid approaches that combine control understanding with data-driven learning.
2. Bridging Scales - From Fundamental Phenomena to Process Operations: Industrial and natural systems often span multiple scales, from molecular interactions to large-scale processes, making their modelling and optimisation a significant challenge. The research integrates scientific machine learning with process systems engineering to address this, enabling a unified framework that connects fundamental physical principles with data-driven models.
3. AI for Product and Process Innovation: The design of new industrial processes and materials requires predictive tools that balance accuracy and computational efficiency. Research in this area focuses on AI-driven methodologies to enhance product and process development by integrating AI with PSE to accelerate innovation cycles, contributing to the development of sustainable and high-performance industrial solutions.
4. People-Centered Systems Engineering & Sustainability: The role of AI extends beyond industrial efficiency to broader societal and environmental applications. Research in this area focuses on people-centred AI-driven systems, ensuring that technological advancements align with cultural, educational, and sustainability goals. For example, AI-powered PSE solutions are applied to cultural heritage conservation, in education, and energy integration.

Biosystem engineering

The biosystems engineering research area (Skjøndal-Bar) applies dynamic modelling and control in order to understand and regulate processes in biology and chemistry, such as aerobic fermentation bioproceses, intercellular processes (molecular biology), enzymatic reactions, hormonal regulation and sensory-motion control. Part of the Microbial Feedback Control Laboratory (MFCL), our current research focuses on:

1. Real-time Model predictive control (MPC) of fermentation bioprocesses. We are not focusing only on simulations, but on real-time aspect of the feedback control, include input-output (I/O) information exchange between the user (in Matlab) and the microbial cell factories in the fermentation vessels.
2. Advanced bioprocess modeling, including microbial monocultures, multicultures (microbial consortia), Deep-learning hybrid models (mechanistic combined with neural networks)
3. Advanced process control structures, including multi-layer MPC, reinforcement learning (machine learning scheme for control), PID cascades, adaptive models and robust multi-variable control.
4. Estimators and Digital Twins, including Real-time Extended and Uncented Kalman filters (with many variations), Real-time moving Horizon Estimator for Fed-batch and Continous microbial fermentation proceses.
5. The affect of noise and its reduction on bio-processes, including low-pass filters, Fourier transform representations, and other strategies.
6. Sensory-motion control models. Dynamic models of sensory processing and flight motion in bats, is with cooperation with labs in Israel and USA, where we investigate bats motion and navigation in the air and the manner bats process its sensory information to fly, navigate and find food. This projects include field works (mexico, Israel, US) and laboratory work, in addition to dynamic modeling and parameter estimation.

More information about the Microbial feedback control laboratory