PID controller tuning for insulin regulation system

Hello there,

I am working on a project related to BioSystems, I have the attached model for the system, I am wondering what is the best way to tune the PID controllers in order to get the desired glucose level.

How should I select the P I D to get the best results.

I have attacehd the model in earlier version of Matlab aswell 2015b.

Thanks.

[Mathieu NOE]

Hi

pls share also the script to initiaze the data inside the simulink file

it the simulink model up to date - it is a bit different from the screen shot (I saw the display of blood glucose with the two limits lines are not present)

tx

[Abdulrahman Odhah]

Hello Mathieu,

Actually I have added the levels as just an indicators [constants] no more no less. everything else is same and up to date.

The prupose of the regulatore is to keep the glucose level between the 50-140. and I want to tune the PIDs to do so.

[Mathieu NOE]

ok

got this

but I stll miss the script that initialize all constants in your simulink file

[Mathieu NOE]

okso I tried to tune the PID

Integral term is not needed, mostly P and a lot of D (10x more D than P) to keep things stable;

but it seems the "process" after the PID is unable to apply fast enough the corrections so that the bumps are still there

I went up to P = -100 and D = -800 , and even after removing the saturation block, still no visible effect on the error.

[Abdulrahman Odhah]

You are totally right for the second PID, but you if you noticed the second PID will be activated only at the saturation and won't affect anything in other cases, actually this is the way that anti windup mechanism has been implemented and yeah the I part will be zero in this case for the sack of the antiwindup.

about the results it is not about stability here in this system it is about forcing or let's say controlling the doses of the insulin injected into the patient as per the following scenario:

1- sensor is reading the current glucose level in the blood, which is supposed to be normal ( 90-120)

2- patient eat snacks or meals that contains sugar, this action increases teh level of the sugar in the blood and since the body can't produce enough insulin to process this amount of sugar (diabetes) , the sugar sensor will trigger the insulin pump to work.

3- the insulin motivates the cells to process the sugar which means decrease the sugar level again to the normal range or close to it.

So back to our topic , tuning the PID, the good result can be obtained in the following conditison:

1- assumption that all the existing parameters in the model are OK and just the PID parameters shall be manuplated.

2- the sugar level shouln't go below 50 at any case.

3- the suger level can resonate between 60 -120 +/- 10.

4- the faster degrade of the sugar level is better.

5- a spark of sugar level is expected after a meal.

6- the a ttached image shows an acceptable operation of the system with the curvers shows the following:

- curve in black (thin) is the response in the body without the system.

- curve in black (thik line) is the desirable response.

- RED is good.

- green is not good because the level drops below the safety line.

Answer 1

[Abhiroop Rastogi]

Hi Abdulrahman,

I am not sure if you have gotten your answer to this question yet, but just in case it did not happen, I want to mention that there are PID tuner tools available in Simulink that can be utilized for the purpose you are looking for.

You can follow the section on PID Controller tuning in Simulink which will let you adjust the type of responce you need in terms of "Response Time" or "Transient Behaviour" also.