PID bumping issue

[Priyadarsini]

Hi all,

I am working on an engine testing system. This uses a PID to control the dynamometer. The performance of this PID is a little confusing. When the setpoint is reduced, the PID output decreases steeply until certain point. But after this point, the output increases a little and then again decreases gradually - causing a bump in the unloading phase. The same profile is followed by the Process Variable also. What might be the reason for this bump?

Setpoint

Process Variable

PID Output

Thanks in advance,

Priyadarsini

[srikrishnaNF]

Hi priyadarshini,

 Are you using manual tuning or autotuning?

Thanks and regards,

srikrishnaNF

[Priyadarsini]

auto tuning. Further if it might be useful, Linearity=1 & Beta=1. I tried changing the Beta but still the problem exists. Reducing the Ti(min) seemed to smooth down the bump in the unloading phase, but on the other hand, loading becomes oscillating much. Customer does not prefer gain scheduling (different PID parameters for different error ranges). So is there a way to define a single PID that works for both loading and unloading? But it is also important to know why this bump occurs.

[Angela M]

Hello,

First, the dynamics of the system appear to be significantly different loading and unloading.  The best way to handle this would be to have appropriate gains for each set of characteristics.  What is the customer's objection to using gain scheduling?  It is a very common and appropriate way to handle non-linear systems.

It is possible you may be able to find a set of gains that would handle both loading and unloading, but neither response will be ideal, since it will need to accomodate a wide range of responses.  I also notice that there appears to be a "bump" going up as well which makes me wonder if there is something going on with the system at that specific level.  Depending on your settings, the PID response may be very sensitive to noise, so if there is noise in the system, that may contribute to unexpected results.

Regards,

Angela M

[Priyadarsini]

The customer is not preferring gain scheduling because it makes the tuning more difficult. Their end customers, who are used to tuning with single parameters, would find it difficult to tune and find out 2 set of PID parameters. But from all your comments & our study it seems that the gain scheduling is a positive step towards solving this problem. We shall communicate the same to the customer and try this out with the actual system.

Thanks,

Priyadarsini