Current insulin therapy for patients with type 1 diabetes often results in high variability in blood glucose concentration and may cause hyperglycemia (high glucose level>10 mmol/L) and hypoglycemic (low blood glucose level< 3.8 mmol/L) episodes. Closing the glucose control loop with a fully automated control system improves the quality of life for insulin-dependent type-1 diabetic patients. This paper presents a closed loop control system that is based on minimization of the risk of the future hypo or hyperglycemicepisodes. The blood glucose level is predicted after 60 minutes using recurrent neural network (RNN), and the fuzzy logic controller (FLC) calculates the insulin dose according to previous setting for patient. The controller tunes the insulin infusion rate to minimize the predicted risk of hypo- or hypoglycemica. The system is tested and evaluated using a simulated diabetic patient model with three meal challenges, and direct performance measures are measured for the resulted controlled blood glucose (BG). Our results indicated that, using our controller can control the blood glucose level without any recorded hyperglycemia even if the scheduled meals are increased by 10%. A mild hyperglycemic episode was recorded when a meal is 35% more thanthe scheduled meal,
but it continues for short time (around 2 Hours, where it is dangerous to be continued for few days or weeks) then the glucose returned to the target range. All the direct performance measures for the controlled blood glucose with our controller is within the standard levels that are mentioned in literature.