Statistical Approaches to Robust Identification of Multi-modal Processes

Yaojie LuUniversity 2014 · DOI

Abstract

In this thesis, time-varying behaviour, nonlinearity and switching dynamics are generally treated as multi-modal behaviour. Two multi-model modelling techniques, i.e., the linear parameter varying (LPV) technique and the switched modelling technique, are investigated to model the multi-modal processes. The robustness of pro-posed algorithms is enhanced by modelling the noise as t distributions. Two statistical approaches, i.e., EM algorithm and variational Bayesian algorithm, are used for process identification. The proposed algorithms are verified by simulations and experiments. Finally, soft sensors based on proposed algorithms are designed to effectively estimate the steam-quality for the once-through steam generators used in the oil sands industry.
BibTeX
@inproceedings{lu2014statistical,
  author = {Yaojie Lu},
  title = {{Statistical Approaches to Robust Identification of Multi-modal Processes}},
  booktitle = {University},
  year = {2014},
  doi = {10.7939/r3r95p},
}

Loading related content...

Figure