Several industrial processes fall under the category of distributed parameter systems (DPSs) which are modelled as coupled partial differential equations (PDEs). The conventional approach is to employ early-lumping technique whereby the PDEs are spatially discretized to arrive at a reduced-dimension model consisting of differential algebraic equations (DAEs) and then use it for controller synthesis. However, the conventional controller synthesis approach doesn’t utilize complete information on the behaviour of state variables throughout the spatial domain of the system for design and implementation. In this work, the reduced-dimension DAE models will be constructed using interpolating polynomials by applying orthogonal collocation (OC) method. The key idea is to utilize spatial state profiles obtained using the interpolating polynomials to develop nonlinear model predictive control (NMPC) schemes for distributed parameter systems. In particular, special properties of certain class of polynomials, such as Bernstein polynomials will be exploited to develop efficient schemes for constraint management. In addition, automated model based fault diagnosis will be integrated with the NMPC scheme to intelligently handle sensor failures and unmeasured disturbance /model parameter variations. The efficacy of the proposed NMPC approach will be demonstrated by conducting simulation studies on benchmark problems in the literature.
Name of Faculty