Integral Action of FB58 starting in negative values

Hello everyone,

     I've implemented an FB58 to control the temperature in a spray dryer, and due to thermal shock danger in the heating duct, I am making it very slow, to accomodate the client's needs, thus not using the derivative action. Now, I see that from the beginning—ambient temperature—the integral action is taking negative values. I really don't see how this can happen from the formula presented in the manual. It seems to always reach 0 and start growing positive for a specific error value, but that value doesn't really mean anything. I tried initializing it in 0, but then I would have to write code to reset that bit so it can grow according to the past errors. Why does this happen?

Thanks in advance.

hdhosseini

Hello  karkas,

For a negative integral part ,when machine in ambient temperature, check the error if it is negative.If so, for negative errors (PV>SP), you need a negative gain.But ,when the integral part is getting an oscillatory response, you need to be sure:

1. Ti should be long enough(TIME) to ensure that even higher negative error not to  cause rapid negative integral part.
   
   
2. Improper selection of CYCLE and Ti

For more accurate comments, please share your FB58 program section and some online values.You need to let the PID  accumulate enough value of integral part not to run out of it so quickly in case of process fluctuations.

I hope it helps,

hdhosseini

Hello hdhosseini,

     Thanks for your reply. I will go to the factory tomorrow and take pictures of the online values. But for now I can tell you for sure that the error is never negative. The SP is always 250°C and I start the controller—Manual to Automatic—from 30°C to 50°C of PV, and actually in this part, when the error is greatest in the positive direction, is when the integral part is greatest in the negative direction. 

     The client needs the system to heat to 250°C in about 1 hour, not more and not less, because as they say—I need to check this information, but even when it's not, it's the client's requirement—the heating duct could suffer thermal shock if heated too fast. I know it could, but I don't know how fast it could actually be heated.  For this reason I'm not using the D action, it's not necessary, and even when used, it's always 0. I've made the FB58 get the PV to 250°C in about an hour with the integral action doing what I described. But I really don't know why at that same point of the error for the same Ti, it always reaches 0. I can plot more accurately in MATLAB the three values from the curve recorder, but don't really know how to generate another .csv with the error and the integral action so you can see it. I don't have the developer's license for the SCADA to generate that file there either. Do you know how to do this?

     I'll come back with those pictures. I think I will stay with that behavior, because if I get the integral action to start from 0, I'll need to test it again several times because it'll reach the SP in less than 1 hour, thus heating the duct too fast. But I want to know why this happens. The integral action is not oscillating, it's just starting from about -14 for an error of about 200 - 220.

Thank you.

PD: Is there a way to post more than 1 image here? I haven't been able to find how.

The purpose of PID controller is to reject distrubance, not to set Kp,Ti  such that rise time of output value has a desired ramp time. Such controlled system will perform very poor on disturbance.
You should implement a temperature profile vs. time and use it as setpoint value. Further if you know the steady state characteristics of your sytem: controller output vs. temperature, you can make a look up table and feedforward the value after PID (DISV value). 

As for negative integrator value there is a my explanation (not sure):
In manual mode the regulator outputs a certain set value,. If the transition from manual to auto has to be bumpless, then u=Kp*E + int(E)dt = u_manual this implies that integral part I has to be I=u_manual-P, minimal u_manual, large error will give you negative value.

Hi,

Cant we increase setpoint gradually to reach 250*C in required time ? (instead of giving 250 directly as a set point)

hdhosseini

Adding new points:

1. If error is positive for a long time, integral part will positive with small rate of accumulation depends on Ti.
   
   
2. You can adjust the time you want to reach to SP by changing gain and Ti.
   
   
3. Try again to record data by opening the regarding DB and online options available there.
   
   
4. You can make a .zip or .rar archive of a collections of miscellaneous files and upload it.If memory is limited (3MB), break it into smaller parts.

Hi hdhosseini,

     Attached is a picture of the values when the integral action is negative, it's in Spanish, but the names look very similar. I changed the temperature to be controlled when the system is started in order to provide it with a better protection against thermal shock.  I probably haven't read enough, but how can I create that file?

Marko Bursic

The purpose of PID controller is to reject distrubance, not to set Kp,Ti  such that rise time of output value has a desired ramp time. Such controlled system will perform very poor on disturbance.
You should implement a temperature profile vs. time and use it as setpoint value. Further if you know the steady state characteristics of your sytem: controller output vs. temperature, you can make a look up table and feedforward the value after PID (DISV value). 

As for negative integrator value there is a my explanation (not sure):
In manual mode the regulator outputs a certain set value,. If the transition from manual to auto has to be bumpless, then u=Kp*E + int(E)dt = u_manual this implies that integral part I has to be I=u_manual-P, minimal u_manual, large error will give you negative value.

There was a question regarding controlled heating prosess, like ramping the temperature SP.

https://support.industry.siemens.com/tf/ww/en/posts/150850/

For feedforward action you would need a LUT of the control value at steady state. For example you heat to the 250 deg, and you write the control value at steady state, then you heat at 150, 100, 50 and you have a output values vs temperature (all at steady state!), or for simplicity you take just one and make linear function. If you forward this value at DISV input, no matter what PID output is you have settled the system at rough value, the PID will add/subtract from this point.
It has to be clear that you may use only temperature setpoint value (SP) for LUT, never the PV. As the feedforward comes in the closed control loop, as open loop addition, it won't have any influence on the stability of the closed loop. But with a constraint: the PV may never be used for LUT input, because you can create one additional closed loop, then.
So, this (open loop-always!) added value acts like a distrubance to the regulation loop, and so the PID input is called DISV. Since the regulation loop has to reject all kinds of disturb frequencies that can put the entire system in the instability, the same applies to DISV value. Finnaly we can say that adding a DISV value has no influence on the stability of the closed loop, as it rejects all kinds of unwanted frequncies - the Kp, Ti, Td have to be choosen such.

If you don't want a "help" of feedforward you can ommit it, but your first point to achevie is to tune the controller tight at desired SP, then to implement ramping the SP toward the final SP. 

EDIT:

The same rule applies to the ramp generator. It has to be an open-loop generator, you can however take the initial start point from PV, but then you continue to ramp only with ramp time, no PV should be used therafter, to not incidentaly create a second closed loop. 

A depiction worth more than words, right: