Modelling and optimisation of blood glucose control for type 1 diabetes using multi-parametric programming and model-based predictive control (mp-MPC) / Associate Professor Dr Ayub Md. Som

Md. Som, Ayub (2014) Modelling and optimisation of blood glucose control for type 1 diabetes using multi-parametric programming and model-based predictive control (mp-MPC) / Associate Professor Dr Ayub Md. Som. Sabatical Report. Faculty of Chemical Engineering. (Unpublished)

Abstract

This research attempts to develop a new control algorithm to regulate the blood glucose level (BGL) for Type 1 Diabetes. In doing so, Multi-Parametric Programming technique is used to develop the computer algorithm; whereas Model-Based Predictive Control (MPC) is adopted for the design of the controller. Non-Linear Bergman Minimal Model is used to represent the three compartments; plasma glucose, plasma insulin and effective insulin compartments. AU simulation and optimisation works are carried out using gPROMS™. Two types of observations are made so as to study the performance of the proposed control algorithm mainly, the control of BGL values without meal disturbance and the control of BGL values following meals. For the control of BGL without meal disturbance, it is found that the BGL values increase substantially at first and fluctuate around 80 mg/dL to 130 mg/dL. They then tend to level off at 120 mg/dL for sometimes before dropping drastically to 60 mg/dL. However, the BGL values remain at 80 mg/dL prior to reaching its steady state condition at the end of the simulation work. For the control of BGL values following meals, it is found that there are three peaks occurred, which obviously indicate a sudden change in the BGL values in conjunction with the introduction of Fisher meal effect into the glucose-insulin dynamic system. Three simulation works are carried out using three different algorithms so as to refine the performance of the controller. For all three cases, theBGL profiles are almost the same in which they tend to fluctuate initially around 65 mg/dL to 120 mg/dL prior to levelling off at 80 mg/dL throughout the remaining periods. These results match with the works carried out by the previous workers. The only major difference is that the value of exogenous insulin infusion rate, u is on the higher side. This could be due to different diabetic models used and inconsistency in choosing the units for the different parameters. However, it can be concluded that with the proposed control algorithm, both hypoglycemia and hyperglycemia are avoided and it is hoped that the algorithm can be easily installed in the form of microchip for the benefit of the diabetic patient in the near future.

Metadata

Item Type: Monograph (Sabatical Report)
Creators:
Creators
Email / ID Num.
Md. Som, Ayub
UNSPECIFIED
Subjects: Q Science > Q Science (General) > Back propagation (Artificial intelligence)
R Medicine > RC Internal Medicine
Divisions: Universiti Teknologi MARA, Shah Alam > Faculty of Chemical Engineering
Keywords: Artifical Intelligence, Multi-Parametric Programming, Model-Based Predictive Contro
Date: 2014
URI: https://ir.uitm.edu.my/id/eprint/70412
Edit Item
Edit Item

Download

[thumbnail of 70412.pdf] Text
70412.pdf

Download (24MB)

ID Number

70412

Indexing

Statistic

Statistic details