Title: Extensions to experiment design and identification algorithms for large-scale and stochastic processes
Authors: Kaushik Subramanian, Siddhartha Kumar, Sachin C. Patwardhan, Vinay Prasad
Addresses: Department of Chemical and Biological Engineering, University of Wisconsin, Madison, WI 53706, USA. ' Department of Chemical and Materials Engineering, University of Alberta, Edmonton, AB T6G 2V4, Canada. ' Department of Chemical Engineering, Indian Institute of Technology Bombay, Powai, Mumbai 400076, India. ' Department of Chemical and Materials Engineering, University of Alberta, Edmonton, AB T6G 2V4, Canada
Abstract: We explore extensions of optimal experiment design algorithms for large-scale deterministic catalytic kinetic systems, and of system identification and control techniques for stochastic thin film deposition systems. In the case of optimal experiment design, we suggest the use of principal component analysis and clustering to identify similarities in parameters and their sensitivities. This enables the grouping of similar parameters together, thus reducing the dimensionality before applying a D-optimal experimental design, and ensuring identifiability of the selected parameters and better conditioning of the Fischer information matrix. In the case of thin film deposition, since the stochastic deposition process is simulated using kinetic Monte Carlo models, closed form models that can be used for control are not available. We present modifications to proper orthogonal decomposition methods for non-ergodic systems, and methods for reparameterisation of autoregressive integrated moving average models for identification of compact closed form models and model predictive control.
Keywords: system identification; proper orthogonal decomposition; POD; D-optimal design; model predictive control; catalysis; kinetics; thin film deposition; stochastic processes; experimental design; Monte Carlo simulation; modelling.
International Journal of Advanced Mechatronic Systems, 2011 Vol.3 No.1, pp.3 - 13
Published online: 18 Mar 2015 *Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article