Autoprofit

Opdrachtgever: Delft University of Technology/Eindhoven University of Technology
Opdracht: Presentatie van onderzoeksresultaten Autoprofit (Engels)
Grafisch ontwerp: Marieke van de Graft, grafisch en illustratief ontwerp

Autoprofit Brochure

Autoprofit

Motivation & Philosophy

The European process-industry faces massive challenges. Just-in-time production of customer specified products is a prerequisite to survive in a global demand driven market. Product quality, pricing and delivery time put continuous pressure on the improvement of production processes. Furthermore, environmental regulations and scarcity of raw materials are forcing industry to transform to more sustainable production methods. Waste of basic material, energy-loss and CO2 emissions need to be minimized.

Model-based control and optimization systems such as Model Predictive Control (MPC) and Real-Time Optimization (RTO) are now commonly used in process-industry to optimize both the production process and the economic performance. Although these technologies offer substantial benefits on ensuring operational conditions and fulfilling product specifications, costs of deployment and maintenance are also significant.

AUTOPROFIT gains important reductions of these maintenance costs. It brings a high level of autonomy to these operation support technologies, hence minimizing human intervention.

AUTOPROFIT uses a three-step philosophy in its next-generation support strategy:

  1. Decreasing modeling costs.
  2. Automating just-in-time maintenance of the models and the model-based application (MPC, RTO).
  3. Using economic criteria for detecting and diagnosing in the automated maintenance.

AUTOPROFIT is a three year European research project. Four academic and three industrial partners, known for their expertise on model based technology, are working together in the development of ‘advanced autonomous model-based operation of industrial process systems’. The project is co-funded by the European FP7 program.

Approach

The ultimate goal of the AUTOPROFIT-project is autonomous maintenance of production processes. To achieve this autonomy, meaning minimization of human intervention by increasing the accuracy of model based controllers, the project focusses on technology development in main aspects of systems and control.

The basic principle lies in the continuous monitoring of key-indicators for economic performance. This economic performance can fluctuate under influence of changes in material costs or quality, energy prices or losses due to the violation of constraints.

The functioning of model-based control and operation systems declines over time. In case a significant performance drop is detected, an instant action from the operation system is desired.

Least costly experiment design is used to diagnose the actual cause of the economic malperformance with minimal impact on resources. The algorithm examines process variables without production-loss.

Subsequently, the most economically effective way to recover the system performance is determined. Hence, identification of a new model in closed loop or retuning of the controller is inevitable.

At this moment linear, time-invariant models are the standard in model based control systems. AUTOPROFIT aims at the development of new linear models in closed loop. A future goal is the development of a model structure that can have both static as dynamic behavior, enabling smooth transitions between linear and nonlinear dynamics.

AUTOPROFIT uses a selection of weighting matrices for controller auto-tuning, focussing on a good balance between system robustness and performance.

Results (overall)

The AUTOPROFIT project has already lead to promising results in the pre-field development phase. Close-cooperated research between the academic and industrial partners has resulted in the following outcomes:

  • The technology is done; the fundamentals are ready for testing in real, commercial situations.
  • The economic performance criteria are identified and are ready to be used in test signal design, performance-monitoring and diagnosis.
  • Novel model calibration and adaptation techniques are developed.
  • The basic principles for nonlinear modeling (LPV) are ready for implementation
  • Least costly test signal design in closed loop has been developed. This fundamental technique balances the test duration and impact on process operation against the accuracy of the new model.
  • The model based control application has evolved for automated (re)tuning.
  • A practical tool which integrates all of the technological developments into a single software package has been developed with an intuitive graphical user interface. GUI allows the tool to be used by the wider process control community.