(7gy) Novel Strategies for Quantification of Model Uncertainty and Real-Time Optimization of Batch Operations

In the last few decades, there has been a steady growth in application of optimization techniques to the solution of problems of industrial importance, e.g. supply-chain management, operational planning/scheduling and optimal process control/dynamic optimization. This trend has been motivated by both increasing global competition and tighter environmental regulations, which are driving industry towards reducing costs while also minimizing environmental impacts.

Although extensive process systems engineering research has been reported so far, most of it has addressed continuous processes and only limited attention has been devoted to batch operations, which are widely used to produce high value-added products (specialty chemicals, pharmaceuticals, cosmetics, etc.). Given the opportunities that this application domain provides, I have focused my research on methodologies/applications belonging to four main areas:

• Operational planning and scheduling of batch processes;
• Deterministic and stochastic model predictive control and/or dynamic optimization of batch processes;
• Static and dynamic estimation of model uncertainty from experimental measurements (model uncertainty is represented as the probability distribution of an appropriate set of uncertain parameters);
• Dynamic identification of the optimal uncertainty set in stochastic dynamic optimization and model-based control problems (the optimal uncertainty set is simply intended as the ensemble of the most relevant uncertain parameters of the process model).

In the past, I have also addressed supply-chain problems¹, and continue to have interests in that enterprise-level research area as well.

More recently, I have focused my research efforts on developing methods for the optimal, real-time management of both single batch operations and multi-unit/multi-step batch processes. In particular, I have formulated and implemented two dynamic optimization and/or model-based control algorithms applicable to single batch operations (BSMBO&C ^2,3 and RBSMBO&C ^4,5) as well as a novel strategy aiming at the integration of scheduling, dynamic optimization and model-based control of multi-unit batch processes (MUBSMBO&C ⁶). The BSMBO&C and MUBSMBO&C algorithms assume that model mismatch is negligible while RBSMBO&C explicitly accounts for model uncertainty, through a novel dynamic scenario selection procedure. These approaches are very flexible, and will accommodate virtually any batch system and/or performance metrics.

My latest research work involves the estimation of the probability distributions (PDFs) of the uncertain parameters of ODE/DAE models as well as the real-time selection of the optimal uncertainty set in stochastic online optimization and control problems. I have already developed a methodology for estimating the PDF of the uncertain parameters of an ODE/DAE model, which relies on a novel back-projection concept (PDFE&U ⁷), and provides satisfactory accuracy at a low computational cost. In fact, PDFE&U appears to be more efficient that many existing state-of-the-art random sampling-based approaches, such as MCMC. I am currently working on a strategy for dynamic estimation of the optimal set of uncertain parameters to be used in a stochastic real-time optimization and/or model-based control problem, which exploits and combines PDF estimation techniques, e.g. PDFE&U, multi-point sensitivity analysis, and novel raking indices. Both of these frameworks show interesting synergies with RBSMBO&C.

In the near future, I am planning to apply these algorithms both to solve drug delivery problems (individualized drug dosing) and to improve the efficiency and reliability of food processing technologies, active pharmaceutical ingredients (APIs) production and specialty chemicals synthesis. Moreover, I am also interested in augmenting MUBSMBO&C to make it capable of explicitly handling model uncertainty, as well as in further improving the computational efficiency of all of the aforementioned strategies. Further algorithmic improvements may also involve incorporation of goal programming techniques into some of the aforementioned strategies in order to solve multi-objective scheduling, dynamic optimization and model-based problems under uncertainty.

Teaching Interests:

My research work and graduate studies have provided me with deep knowledge of control theory, numerical methods, dynamic and steady-state modelling, coding/parallel computing (especially in C++), and statistics. Therefore, I would be interested in teaching any course broadly related to these subjects, e.g. applied statistics, process control, numerical methods, dynamic and steady-state modelling and process optimization. I would also like to introduce and teach two new classes, whose topics lie at the boundaries of Chemical Engineering, Statistics and Computer Science:

• Applied statistics for process optimization/control under uncertainty;
• Parallel programming techniques applied to process simulation and optimization.

Since optimization strategies have become an essential tool for the solution of many problems of industrial relevance in the last few years, I strongly believe these interdisciplinary courses would provide new chemical engineers with useful and valuable knowledge.

References:

1. Rossi F, Manenti F, Reklaitis G. A general modular framework for the integrated optimal management of an industrial gases supply-chain and its production systems. Computers and Chemical Engineering. 2015;82:84-104.
2. Rossi F, Manenti F, Buzzi-Ferraris G. A novel all-in-one real-time optimization and optimal control method for batch systems: Algorithm description, implementation issues, and comparison with the existing methodologies. Industrial and Engineering Chemistry Research. 2014;53:15639-15655.
3. Rossi F, Copelli S, Colombo A, Pirola C, Manenti F. Online model-based optimization and control for the combined optimal operation and runaway prediction and prevention in (fed-) batch systems. Chemical Engineering Science. 2015;138:760-771.
4. Rossi F, Reklaitis G, Manenti F, Buzzi-Ferraris G. Multi-scenario robust online optimization and control of fed-batch systems via dynamic model-based scenario selection. AIChE Journal. 2016, DOI: 10.1002/aic.15346.
5. Rossi F, Manenti F, Pirola C, Mujtaba I. A robust sustainable optimization & control strategy (RSOCS) for (fed-)batch processes towards the low-cost reduction of utilities consumption. Journal of Cleaner Production. 2016;111:181-192.
6. Rossi F, Casas-Orozco D, Reklaitis G, Manenti F, Buzzi-Ferraris G. A computational framework for integrating campaign scheduling, dynamic optimization and optimal control in multi-unit batch processes. Computers and Chemical Engineering. 2017;DOI: http://dx.doi.org/10.1016/ j.compchemeng.2017.05.024.
7. Rossi F, Manenti F, Buzzi-Ferraris G, Reklaitis G. A novel back-projection approach to the estimation or update of the probability distribution of the uncertain parameters of an ODE/DAE model. In preparation. Estimated submission at the end of 2017.