Zhenhua Tian
PhD Candidate (curriculum vitae)
A State-Shared Modeling Approach to Transition Control
Year Started:  Fall 2000
Anticipated Graduation:  Dec 2003
Advisor:  Dr. Karlene A. Hoo

B.S. Applied Mathematics, 1991, Beijing University of Chemical Technology.
M.S. ChE, 1994, Beijing University of Chemical Technology.

This research addresses transition control of multi-product continuous processes. The framework that is proposed consists of multiple models in a model reference adaptive control structure. (MMRAC)

Non-adaptive and adaptive models are used in the MMRAC. Additionally, two types of adaptive models are used. One that re-initializes its parameters from any non-adaptive model whose identification error is judged as the smallest one, and one that starts its adaptation from its previous parameter values. The controllers’ parameters are adapted based the control error between the process outputs and the desired setpoints and on the identification errors. The models have the same functional form (can be relaxed) and each model has an identified controller such that if that model produced the best estimate of the process then its controller would be placed in feedback with the process.

  This multiple model structure also requires a performance measure to select the best model and a switching scheme to decide when to place the best model’s controller in feedback with the process. Most importantly, the scheme avoids placing each controller in feedback with the process to determine which model/controller pair is suitable to regulate or track the desired outputs. The MMRAC can be used for both regulation and tracking. However, to address constraints, interactions and time delays, a multiple model-based predictive control framework is also being investigated.

Progress to Date: Demonstrated the use of model-reference adaptive control (MRAC) on a continuous-stirred tank reactor with first order irreversible exothermic reaction. This single-input single-output process undergoes composition and temperature disturbances at each operating point and transitions between high, medium, and low conversion states. The adaptive models used are linear transfer function models identified using Matlab’s system identification toolbox.  A multiple-input multiple-output nonlinear interactive two-phase reactor system is being investigated next.  

Refereed Manuscripts: For reprints contact K. A. Hoo (khoo@coe.ttu.edu)

  1. Z. Tian and K. A. Hoo, Transition Control Using State-Shared Model Approach , Compt. & Chem. Engng., 27(11), 1641-1656, 2003.
  2. Z. Tian and K. A. Hoo, State Shared Model for Multi-input Multi-output Systems, submitted to IEEE Transactions on Control Systems Technology, 2003.
  3. Z. Tian and K. A. Hoo, Global Linearizing Control of Calcium Dynamics in Cardiac Myocytes, submitted to J. Process Control, 2003.

Proceedings

  1. Z. Tian and K. A. Hoo, Transition Control Using Multiple Adaptive Models and an H-infinty Controller Design, 2002 American Automatic Control Conference, paper TM11-6, Anchorage, AK, May 2002.
  2. Z. Tian and K. A. Hoo, A State-Shared Model Predictive Control Approach To Transition Control, AIChE National Conference, Indianpolis, IN, Nov., 2002.
  3. Z. Tian and K. A. Hoo, A State-shared Modeling Approach to Transition Control, IFAC ADCHEM 2004, Hong Kong, PRC.
  4. Z. Tian and K. A. Hoo, A State-shared Model Predictive Control Approach to Transition Control of a Multi-input Multi-output  System, 2003 American Automatic Control Conference, Denver, CO, June 2003.
  5. Z. Tian and K. A. Hoo, Global Linearization and Control of a Calcium Dynamics in Cardiac Myocytes, AIChE National Conference, San Fransciso, CA, Nov., 2003.