We would like to give a warm welcome to our new lab member Carlos Jose Gonzalez Rojas . He is going to start his Phd in the PRINTYOURFOOD project.

We are going to present two papers at ADCHEM

Modeling of milk acidification


Local parameter identifiability of large-scale nonlinear models based on the output sensitivity covariance matrix

The AIDA project performed fundamental and applied research towards predictive capabilities and the integration of digital technologies to different market-sectors. We have studied  how Artificial Intelligence solutions can be extended and improved with the inclusion of dynamic (transient) process information hence will lead to better predictions, and therefore better applicability.

The AIDA project focused on monitoring the energy consumption in buildings with the aim at predicting the required energy and get a more efficient management with a reduction of cost and lower environmental impact.

The work performed in this project has allowed to set and validate the proof of concept of the hybrid approach (AI + dynamic models) which has demonstrated to be promising for improving the assessment of single AI frameworks.

Partners: ABB, SASOL, Boliden, TUDelft, AWTH, TU/e, KTH

Funding: EU

It is widely recognized that the life-time performance of current model-based operation support systems, like Model Predictive Control (MPC), Real-Time Optimization (RTO) and soft-sensors for large-scale complex dynamic processes is rather limited, particularly due to the fact that the underlying dynamic models need to be adapted/calibrated regularly, requiring dedicated measurement campaigns executed by highly specialized engineers.  In this project, we aimed to develop technology in order to bring the current model-based operation support technology  to a higher level of autonomy. Such a technology requires developments in experiment design, performance monitoring and diagnosis, closed loop identification, autotuning so that it can optimize plant performance under varying operational conditions and adapting to changing circumstances.

Figure 1: Autoprofit Decision Tree


Partners: TU/e DSM , FrieslandCampina, Corbion , ISPT

Rigorous models (first principle models) are an accepted technology in process industries and development of such models are common practice.  There are various modeling environments available with complex reaction kinetics formulations and rigorous thermodynamics information. The models used have become more and more dynamic and are used for offline studies such as process design and optimization, dynamical analysis, control structure selection and operator training. These models are generally nonlinear, large scale and result in differential algebraic representations. They describe a wider range of process operation compared to data driven models. Despite their extensive offline use, the process knowledge within these rigorous models is still not utilized extensively in the production/operation environment for real time online model based applications.  The goal of this project is to integrate rigorous in daily operation of chemical processes and make a step further in closing the gap between offline and online use of rigorous models in these applications.

Two research directions have been identified

  1. Alignment of rigorous models with actual process behavior (online calibration) with a predefined level of accuracy.
  2. Development of technology for the optimal design, dynamic operation, control and decision making for processes under uncertainty.


Partners: TU/e, Corbion, DSM, Huntsman, FrieslandCampina

This project aims to integrate chemometrics and model-based control to increase the controllability and therefore the potential profit for advanced industrial processes.

Based on a number of industrial cases the benefits of this approach will be shown in the areas of quality or efficiency improvements

  1. Developing chemometrics for on-line integration of PAT data with process measurements.
  2. Integrating first-principle information and empirical observations in Advanced Process Control
  3. Using chemometric information from PAT to enhance the information used in Advanced Process Control.