(272g) Generalized S.V.D. Observers for Linear Time Invariant Systems

ABSTRACT: The use of observer models for improvement of process control is important for achieving accuracy and processing efficiency, especially for large multivariate process systems. The reduced order observer is particularly advantageous for reducing computational complexity of estimating state variables.

The problem is modeled as a linear time-invariant continuous system with stochastic uncertainties of white gaussian noise accounting for both process disturbances, and sensor inaccuracies respectively. Noise contributions of multiple sensors are characterized and scaled before being filtered by a soft sensor. The soft sensor design was based on a best linear unbiased estimation of output measurements using generalized singular value decomposition (GSVD) of coefficient matrices of measured variables and noise respectively. The resulting state estimates was conducted through a reduced-order Kalman-Bucy observer.

This method of observation was tested on a simple sample case of a biochemical CSTR around both its stable and unstable steady states. And the GSVD modified observer is shown to outperform the ordinary Kalman filter with real measurements of deviations from Gauss-Markov white noise model and around unstable steady state operation.