Study On The Control Method Of Piezoelectric Driven Insulin Pump

Piezoelectric Insulin Micro Pump is a kind of continuous microdosing device which aims at assisting diabetics in automatic insulin injection.However,due to the high order nonlinear process of the insulin-glucose metabolic dynamic model,the dynamic lag associated with insulin pharmacokinetics and the delay of gluocse concentration between subcutaneous tissue and blood,insulin overdelivery is inevitable,thus giving rise to hypoglycemia.In addition,the low precision of micro injection of piezoelectric pump also restricts its rapid development.To reduce the probability of hyperglycemia and hypoglycemia and improve the safety of insulin pump,the calculation method which could avoid excessive insulin input and the control method which could improve the accuracy of the insulin injection rate of piezoelectric pump are studied.The specific contents are as follows,First,the working principle of piezoelectric insulin pump is analyzed,the mathematical model of the output flow of piezoelectric pump is constructed.On this basis,the mapping mode between the insulin injection rate and the control signal of the pump is designed.To solve the problems of poor blood control caused by the lack of flexibility in the existing insulin injection rate setting methods and the low output flow accuracy caused by the inherent properties of piezoelectric pump.The overall control block diagram which aims at improving insulin pump performance is put forward.First of all,calculating insulin injection rate by using blood control algorithm.And correcting the final injection rate with glucose prediction modular which has high accuracy.Then setting pump control signal on the basis of the actual output flow characteristics of the pump.Finally,adopting closed-loop control method that has strong adaptability to control the output flow of the pump.Second,in considering of the non-linear and time-varying of glucose-insulin regulation model,PID control is used to calculate the insulin dosage.To improve the anti-interference ability of the controller,BP neural network is used to adjust the controller parameters.To solve the problem of hypoglycemia which casused by the delay that lies in insulin transportation and glucose data acquiring,a glucose prediction model based on GRU network was constructed.Once the occurrence of hypoglycemia is predicted,the pump would be shut down to avoid excessive insulin input.To verify the prediction accuracy of the training model,the average absolute percentage error and Clark grid analysis are used.In addition,several time series prediction methods are used as contrasts.The Simulink model of “BP-PID+GRU” is constructed to verify the glucose control effects.Results shows that the incidence of hypoglycemia could be reduced to some extent.Third,the fuzzy PID control method of actual injection rate based on particle swarm optimization algorithm is proposed.The conversion method from theoretical injection rate to the driving signal of piezoelectric pump is given first.To improve the accuracy of the actual injection rate,the control method of insulin injection rate under single pulse signal is studied.To improve the robustness of the control process,the fuzzy PID control method is used.In consideration of the initial parameters of fuzzy PID having a great impact on the the controller performance,particle swarm optimization algorithm is used to optimize the initial parameters,Simulink simulation results show that the controller has both high accuracy and small overshoot.