Sharad Bhartiya
Core Faculty
Professor
311, Chemical Engineering
Core Faculty
Professor
Amar Dye Chem Award 2004
Optimal operation of SMBC involves determining the internal flow rates and switch time that yield a cyclic steady state corresponding to an optimal performance metric such as feed throughput or productivity. Our work envisages using dynamic optimization using SMBC models to determining these best operations. The work also covers optimal startup, recovering from feed upsets, transitions between optimal operating modes, and flexible ways of operation such as multi-period operation. An in-house fabricated SMBC for separation of a glucose-fructose system and capable of computer controlled operation is being tested.
MPC of fast dynamic systems has always been a challenging task, the main hurdle being the time required to calculate the constrained optimization problem at each sample. e-MPC has emerged as one way of arranging these computations. However, a number of issues need to be overcome before this technology can mature. Our work is progressing on two fronts: use of multiple models in e-MPC and developing new ways of solving the point-location problem (solved online in e-MPC). The multiple model approach envisages nonlinear compensation via use of multiple linear models with some form of convex aggregation. Optimzation based on multi-parametric programming leads to an offline generation of numerous regions which must be searched online to determine the location of the current operating point. We are collaborating with Prof. Mani Bhushan to explore discriminant function approaches for this point location problem. The applications include fuel cell and gas turbine systems.
Hybrid dynamic systems involve interactions between continuous (such as in transport models) and discrete variables (such as in switching of dynamics). Constrained optimal control or MPC of such systems require solutions of MIQP/LP/NLP. Our work attempts to reformulate the control problems using set-theoretic methods ease online computing while guaranteeing stability. In particular, we proposed a dual terminal set approach that allows an efficient solution of the online optimization problem in MPC. The work is being extended to robust control using tube-based formulations as well as robust adaptive extensions using range inclusion functions.