A novel tuning method for implementing dynamic matrix control on a water tank system

Model Predictive Control Schemes are now being widely used for plant-wide control to satisfy inequality constraints on plant inputs and outputs. Dynamic Matrix Control is a model predictive control scheme where step response coefficients are used to represent the plant model. The author’s work over...

Full description

Saved in:
Bibliographic Details
Main Author: Sudarshan, Srinivasa Ramanujam.
Other Authors: Cai Wenjian
Format: Final Year Project
Language:English
Published: 2013
Subjects:
Online Access:http://hdl.handle.net/10356/54275
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Model Predictive Control Schemes are now being widely used for plant-wide control to satisfy inequality constraints on plant inputs and outputs. Dynamic Matrix Control is a model predictive control scheme where step response coefficients are used to represent the plant model. The author’s work over the past year has been in learning and implementing the Unconstrained Dynamic Matrix Control Algorithm in a Single Input-Single Output (SISO) water tank system and comparing a newly developed DMC tuning strategy (NlStar) with the exsiting Shridhar-Cooper Tuning strategy. MATLAB was interafced with an on Board DSP chip which connects to the sensors and pumps in the tank. A linear approximation of the Water tank system model was used to derive the optimal controller using both tuning techniques. The cost function taken into consideration had both error to set point and control action as parameters and were minimized. The two tuning strategies were compared on the basis of system performance, control action (change in voltage required) and robustness on occasions of plant model mismatch. Both Simulations and tests on a practical water tank system were used for comparisons. Tuning DMC controllers are considered difficult especially by plant operators as the relationship between parameters are not explicit. The new tuning technique overcomes this difficulty by introducing one unified variable called ’k’ which varies from 0 to 1. Greater the value of ’k’, the system becomes slower and more stable. Lower the value of ’k’, system performance increases and becomes faster. The novel NlStar tuning technique was found to provide the same level of performance as the Shridhar-Cooper tuning method with greater ease to tune manually.