An extended Kalman filter formulation for systematic transfer of information from batch to batch

Cited 2 time in webofscience Cited 0 time in scopus
  • Hit : 247
  • Download : 0
In this paper, a modified extended Kalman filter (EKF) formulation is developed that allows for a systematic batch-to-batch transfer of the parameter estimates and covariance information. A simple and tunable stochastic description for the batch-to-batch behavior of the model parameters and initial conditions is proposed. Then, a method for incorporating such a model into a general ODE model describing the batch process is presented. An extended Kalman filter formulated based on this augmented model is capable of systematically transferring the parameter estimates and the covariance information at the end of a batch run to initialize the filter for the next batch. This provides an advantage over traditional EKF formulations, which can suffer due to insufficient measurements imposed by the finite batch duration. A simple tuning parameter is introduced in the stochastic parameter model that weighs the relative importance of the variance components for the batchwise-correlated and independent portions. This would allow the filter performance to be tuned to the particular situation at hand. An application to a simple chemical reactor and a nylon 6,6 autoclave is discussed to demonstrate the usage and the benefits of the proposed formulation.
Publisher
AMER CHEMICAL SOC
Issue Date
2003-04
Language
English
Article Type
Article
Keywords

SEMIBATCH POLYMERIZATION REACTORS; STATE ESTIMATION; CORRECTION POLICIES; QUALITY

Citation

INDUSTRIAL ENGINEERING CHEMISTRY RESEARCH, v.42, no.8, pp.1753 - 1760

ISSN
0888-5885
DOI
10.1021/ie0204181
URI
http://hdl.handle.net/10203/81109
Appears in Collection
CBE-Journal Papers(저널논문)
Files in This Item
There are no files associated with this item.
This item is cited by other documents in WoS
⊙ Detail Information in WoSⓡ Click to see webofscience_button
⊙ Cited 2 items in WoS Click to see citing articles in records_button

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0