Periodic Reporting for period 1 - GuEst (New Directions in Guaranteed Estimation of Nonlinear Dynamic Systems and Their Applications toChemical Engineering Problems)
Reporting period: 2018-06-01 to 2020-05-31
The technique of guaranteed estimation promises a revolutionary step to how industrial process managers build, handle and adapt the prediction mathematical models. These models are used to monitor the equipment operation, to train the operating personnel, and are also exploited to steer the plants' behavior towards the most profitable or the most resource-efficient regimes. Improved monitoring and control is key to energy and resource efficiency of the assets and it enables a shift towards autonomous production and sustainability in chemical industry. Advantages of guaranteed estimation come from the fact that no unnecessary assumptions must be made regarding the quality and measurement-error distribution of the sensed data, which establishes an increased reliability of the obtained estimation results. The work on this project developed the essential parts of guaranteed estimation techniques for real-world exploitation. We focused on the estimation of parameters of the nonlinear dynamic models while combining the estimation with model validation principles and while creating a hybrid estimation technique that enjoys the advantages of both guaranteed estimation and conventional approaches. In order to drive the operation of the plant (here we focus on the plants of chemical industry) to an efficient working regime, the technology of optimal and robust control is required. Our project builds upon the developments of robust control and develops novel optimal robust control techniques that incorporate the information on guaranteed estimates into the actions, i.e. manipulations of the plants' degrees of freedom. As a result, a safe, reliable and resource-efficient operation is established. The theoretical developments of the project were implemented into software tools and were released as an open-source project such that the collaboration with academia and industrial stakeholders is fostered. A demonstration on a pilot plant was done in the project to showcase the benefits of developed techniques in the real-world environment.
The work plan of the project GuEst followed the structure with three workpackages (WPs): WP 1 - Efficient guaranteed estimation, WP 2 - Online monitoring and optimal control based on guaranteed estimation, and WP 3 - Tool development and demonstration of the results.
Technical work performed:
1. Development of approaches to moving-horizon (state) guaranteed estimation
2. Creation of framework for guaranteed estimation using low-order models
3. Development of optimal experiment design techniques based on guaranteed estimation
4. Development of approaches to dual model-based optimizing control
5. Development of approaches and software for efficient guaranteed estimation
Dissemination activities carried out:
1. Presentation of the project developments at ten conferences and three co-organized workshops
2. Preparation and publication of fourteen full-text conference contributions and seven journal papers
3. Preparation of three posters promoting the project developments
Conducted outreach and other activities to increase impact:
1. Meetings with industry (seven face-to-face meetings, participation at an industrial event)
2. Presentation at European Researchers' Night 2019 in Bratislava; visits to high schools
3. Preparation of EU project proposals
The main results entail novel algorithms for efficient guaranteed parameter estimation. One of the methodologies employs a principle of moving-horizon estimation. The approximation of the guaranteed parameter estimation set is reached via consideration of only those measurements that bring significant improvement to precision of the obtained estimates. Another way to reduce the computational burden was investigated by a novel set arithmetic, a so-called interval superposition arithmetic, and by sampling techniques adapted from Bayesian statistics. Another breakthrough was reached in the guaranteed estimation of (low-order) data-based models. The decision-making paradigm is created to construct a model of optimal order given the variability and error in the data. Optimal experiment design methods have been developed to tackle uncertainty issues in the experiment design and to bridge between classical (linearization-based) experiment design and between fully nonlinear experiment design. The method enables to design experiments with improved precision. An approach was created to model-based optimizing control by tracking the necessary conditions of optimality. The crucial aspect here is the development of algorithmic techniques for propagation of parameter uncertainty into the optimality loss under requirement of robust constraints satisfaction. Ellipsoidal calculus was exploited in creating a new state estimation technique. The method uses a predictor-corrector approach with set intersection and the earlier developed moving-horizon-like estimation approach.
Technical work performed:
1. Development of approaches to moving-horizon (state) guaranteed estimation
2. Creation of framework for guaranteed estimation using low-order models
3. Development of optimal experiment design techniques based on guaranteed estimation
4. Development of approaches to dual model-based optimizing control
5. Development of approaches and software for efficient guaranteed estimation
Dissemination activities carried out:
1. Presentation of the project developments at ten conferences and three co-organized workshops
2. Preparation and publication of fourteen full-text conference contributions and seven journal papers
3. Preparation of three posters promoting the project developments
Conducted outreach and other activities to increase impact:
1. Meetings with industry (seven face-to-face meetings, participation at an industrial event)
2. Presentation at European Researchers' Night 2019 in Bratislava; visits to high schools
3. Preparation of EU project proposals
The main results entail novel algorithms for efficient guaranteed parameter estimation. One of the methodologies employs a principle of moving-horizon estimation. The approximation of the guaranteed parameter estimation set is reached via consideration of only those measurements that bring significant improvement to precision of the obtained estimates. Another way to reduce the computational burden was investigated by a novel set arithmetic, a so-called interval superposition arithmetic, and by sampling techniques adapted from Bayesian statistics. Another breakthrough was reached in the guaranteed estimation of (low-order) data-based models. The decision-making paradigm is created to construct a model of optimal order given the variability and error in the data. Optimal experiment design methods have been developed to tackle uncertainty issues in the experiment design and to bridge between classical (linearization-based) experiment design and between fully nonlinear experiment design. The method enables to design experiments with improved precision. An approach was created to model-based optimizing control by tracking the necessary conditions of optimality. The crucial aspect here is the development of algorithmic techniques for propagation of parameter uncertainty into the optimality loss under requirement of robust constraints satisfaction. Ellipsoidal calculus was exploited in creating a new state estimation technique. The method uses a predictor-corrector approach with set intersection and the earlier developed moving-horizon-like estimation approach.
The project encompassed several beyond-the-start-of-the-art technologies, which enable the use of guaranteed parameter estimation in real-life (industrial) conditions and which enable a reliable decision-making based on the improved estimation. The project delivered novel theoretical and algorithmic developments, open-access data, publications and software tools, and demonstrations of the developed technologies for estimation and control of chemical processes.
Since substantial advances in the applicability of the guaranteed estimation technology were shown within the project, the project results will potentially foster the innovation in industry. European chemical industry employs 1.2 million people in the EU and contributes €551 billion to the EU economy. Unlike few other industrial sectors the chemical industry in EU has still not managed to reach production rates of 2009 (pre-crisis) levels. It is thus among the most important targets of the current movements towards increased energy efficiency and sustainable development. Another set of challenges arises from the demographical changes in EU, e.g. aging of the skilled personnel and striving for quality and safety of working and living environments. This results in a tendency towards reduction of direct involvement of humans at high-risk production sites. Not only technical innovations, new plants and new technologies are needed to achieve the highly ambitious sustainability milestones that are set by the EU, such as 40% cut in greenhouse gas emissions compared to 1990 levels by 2030, and to address the aforementioned demographic and global economic challenges. It is also the exploitation of the opportunities in existing and in future plants. Improved monitoring and control is key to energy and resource efficiency of the assets and it enables a shift towards autonomous production and sustainability in chemical industry.
Since substantial advances in the applicability of the guaranteed estimation technology were shown within the project, the project results will potentially foster the innovation in industry. European chemical industry employs 1.2 million people in the EU and contributes €551 billion to the EU economy. Unlike few other industrial sectors the chemical industry in EU has still not managed to reach production rates of 2009 (pre-crisis) levels. It is thus among the most important targets of the current movements towards increased energy efficiency and sustainable development. Another set of challenges arises from the demographical changes in EU, e.g. aging of the skilled personnel and striving for quality and safety of working and living environments. This results in a tendency towards reduction of direct involvement of humans at high-risk production sites. Not only technical innovations, new plants and new technologies are needed to achieve the highly ambitious sustainability milestones that are set by the EU, such as 40% cut in greenhouse gas emissions compared to 1990 levels by 2030, and to address the aforementioned demographic and global economic challenges. It is also the exploitation of the opportunities in existing and in future plants. Improved monitoring and control is key to energy and resource efficiency of the assets and it enables a shift towards autonomous production and sustainability in chemical industry.