Research On Closed-loop Insulin Pump Controller Based On Model Parameter Estimation

Diabetes is a hyperglycemia disease caused by the damage of pancreas and the lack of insulin. Insulin injection in vitro is now the commonly treatment in clinical. The most advanced treatment equipment is the dynamic insulin pump, but it needs the medical staff or patients to set the amount of insulin dose, which is easy to cause the imprecise dose and undulation in glucose. In order to realize the automation closed-loop of insulin pump and improve the precision of insulin dose, we propose the research on the closed-loop insulin pump controller based on parameter estimation: the model parameter estimation algorithm can reduce the mismatch between controlled object and glucose model, then the closed-loop insulin pump can automatically optimize the best insulin dose according to the glucose concentration. In the process of treatment, glucose concentration must be controlled in the reasonable scope without hyperglycemia and hypoglycemia, which means the efficacy and safety of the closed-loop insulin pump controller are highly valued.The paper first analyzes the basic principle of the closed-loop insulin pump control system, demonstrates the importance of establishing the glucose regulation model from the aspects of the whole control system simulation research and the glucose control algorithm, completes the comparison of several kinds of glucose regulation model, and finally selectes the model proposed by the Hovorka team. The form of SIMULINK model and M file model are established based on differential equations, and the meal subsystem model is improved. Finally through the open loop simulation, the rationality of model is validated.Then the paper expounds the basic principle of the model parameter estimation algorithm and dynamic matrix control algorithm.The paper proposes the idea of appling the dynamic matrix control to glucose regulation, completes the optimal output formula derivation and algorithms program. Because of the differences in human and interference factors, making the changless glucose regulation model as the forecast model in control algorithm is easy to cause the model mismatch and inaccuracy. Thus on the basis of the blood glucose regulation mechanism, the paper analyses and selects six model parameters causing differences in glucose regulation. The method of least square is adopted to realize the model parameter estimation in different controlled objects.The paper also establishes experimental platforms respectively in T1DM simulator and STM32microcontroller. MATLAB language and C#language are used in programming of dynamic matrix controller with parameter estimation, completing the glucose control simulation in ten controlled objects. As a result, glucose concentration is reduced in normal scope from about150mg/dl without hypoglycemia in2h~3h.