(302a) Control System Design for External Insulin Injection: a Bmi Optimization Approach Conference: AIChE Annual MeetingYear: 2006Proceeding: 2006 AIChE Annual MeetingGroup: Computing and Systems Technology DivisionSession: Poster Session: Recent Developments in Systems and Process Control Time: Tuesday, November 14, 2006 - 3:15pm-5:45pm Authors: Chen, C. L., Department of Chemical Engineering, National Taiwan University Tsai, H., Department of Chemical Engineering, National Taiwan University The diabetes affects almost 170 million people worldwide now. The insulin by dysfunction of beta cells in pancreas is the central abnormality of this disease. It is a disordered state that insulin stimulates glucose transport in skeletal muscle and fat, and inadequately suppresses hepatic glucose production. The normal blood glucose concentration level in humans is in a narrow range (70?110 mg/dl). Both in-vivo and in-vitro experiments have revealed that the human insulin is secreted from the pancreas in oscillatory manners in two time scales. It is widely believed that the rapid pulsatile oscillation is caused by the insulin secretory bursts from the Langerhans islets in hundreds of beta-cells in pancreas at a period of 5-15 minutes. The slower ultradian oscillation refers to the plasma glucose concentration oscillation with period in the range of 50-150 minutes. A mathematical dynamic model with two delay time parameters of the glucose-insulin regulation is proposed to demonstrate the oscillatory behavior. The blood glucose regulation for a utility function involving a medical treatment of diabetes mellitus can be seen as a control problem. The problem such as multi-objective control and simultaneous stabilization with structured controller is BMI optimization of being non-convex and NP-hard in complexity. The purpose of the study is using the model of the feedback regulation to generate a better external insulin injection algorithm via BMI optimization approach. In this paper, we performs optimization over a suitably defined non-convex function at each iteration with the property of guaranteed convergence to the optimum, while at the same time is computationally faster and numerically reliable. The simulation results have shown the improvement from ordinary doses and proposed automatic regulatory in overall glucose control.