Closed loop boost control

[Tim D]

Please could someone comment on my closed loop boost control dilemma on my Subaru Impreza STI with GTX3076r turbo.

I have tuned the various parameters including PID setting to give excellent control, the MAP and Target boost agree to around 5 kpa and remain extremely stable.  This is all carried out on the road and pretty repeatable!  A dyno session gave similarly good results.

However, during a track day, after only a few laps, the boost control requires attention to prevent overboost and oscillation.  

Has anyone else experienced this and what approach should I take to achieve a setup that's good for both road and track?
  
Note, I can't upload attachments, but I'm thinking it's a fairly generic query?

[Adamw]

Will need to see a log to help.  You can share it using Onedrive/Google drive/dropbox etc.

[Tim D]

Thanks Adam, here’s a tune and log that works well on the road…

788.pclr  https://drive.google.com/open?id=1pjCW7otZjlRO3j8JqtsJj-2yhvleHtnw

788 (On Road).llg https://drive.google.com/open?id=1zBqW645IBygLo8KAn92o901FGIpFjJYn

 

But not so good on the track…

788 (Brands Hatch).llg  https://drive.google.com/open?id=1LQvNR99bBTaqN-AVhe5ytYShSjByVNMp

 

A few major tweaks to bring it (sort of) under control…

793.pclr  https://drive.google.com/open?id=1e86d4wXrepp3DSImjCoY2RQhGgPd0zeO

793.llg   https://drive.google.com/open?id=1WyUPMAdXe0I20CgvfD8J26by-ut3XzDE

Note, if you search for the Percentage FCut (%) this is where boost exceeded my engine protection limit.

 

Comments welcome (any part of the tune)… I think fuelling is good, but have limited experience on what’s good and what’s bad!

Thanks

Tim

[Adamw]

I think there are quite a few factors all having some influence.  I will list a few in no particular order:

1. In about the first 3/4 of the 788 brandshatch log it looked like you were either learning the track or learning the car (or maybe it was wet or something) as throttle input was fairly erratic and rarely above 60 or 80%. In the road test you were at a stable 100%TP for a decent period.  This has two consequences; 1, your boost target is based on throttle positon - in closed loop control theory this is called your "set point".  A relatively slow control loop needs to be very well tuned to cope with an erratic or rapidly changing set point.  2, I suspect that the wastegate DC table probably didint have much testing or tuning done in the lower boost target/TP cells.  In the later log, as well as tune improvements your throttle is also fully open much more and far less erratic so some of the improvement was due to the driving style also.
2.  The wastegate DC table still looks like it needs a little work.  In all the places I looked it seemed to need about 6-8% less DC than what is in that table.
3. In the 788 brandshatch log most of the oscillation was coming from too much Integral gain.  I see you lowered it in the later tune but it still looks like it has too much.  Try dropping it to 0.1.
4. I think you could try increasing both Prop and derivative. Try adding 50% as a quick test.
5. The over boost cut is due to the TP being only at about 65% for most of the spool (lower boost target) and boost control had already entered stage 3 (integral is switched on).  Then you open throttle more (increasing boost target) so the integral winds up while it is trying to reach that target and you are also pushing in way more DC from the wastgate DC table which causes the big overshoot.  I think a combination of all the suggestions above will help this.  Maybe flatten the boost target table a bit, get the numbers in the wastegate DC table closer to correct and reduce integral.  You can try changing the base DC mode to stage 2 only but I would only do that if the rest of it doesnt fix it.

 

[Tim D]

Thanks very much for the detailed response.

1. Your assumption was correct, I was learning the track.  Good point about the wastegate DC table, I had previously spent a lot of time tuning it, but more recently, had to adjust the WG actuator as I think the spring had relaxed over time.  My original method of tuning WG table was to set the AP/TP table to limit TP to 70%, 80%, 90% etc to precisely target the relevant rows in WG table (does that make sense?)

2. I will review WG table as you suggest.

All your other suggestions make great sense - I will do some more tuning.....

Thank you

[K3rm1tTh3Fr0g]

Bumping for question about G4X sorry - is ~1psi over target boost considered overboosting when using a 38mm EWG and 3 port mac on a G4X using closed loop boost control? EJ20 WRX.

Thanks!

[Adamw]

I guess it depends on the context.  Control loop tuning is always a balance between response time, stability and overshoot.  If you want no overshoot then you need a slower approach.  A short 1psi overshoot would be acceptable to most users and pretty typical in my experience, however I also know of club-level racing classes over here where you have something like a 18psi limit and you are only allowed to exceed that for a total of 2 secs per race from memory, so in that situation, 1 psi overshoot may not be the ideal compromise.

[K3rm1tTh3Fr0g]

Thank you for the reply. Now that I recall/looked at logs it appears it overboosts by around ~1-2 PSI. I'll post a log/tune in a bit after I do some more pulls.

Another question, this car is flex tuned for E85 and I am noticing the WG DC 1 and WG DC 2 tables are identical and boost target is what differentiates boost levels. Do the WG 1 and WG 2 tables represent WGDC on the different fuel bleds? Or is it referring to something different all together?

[Adamw]

I can only guess without seeing the config, but I would say the 2 wastegate tables are probably set up in interpolate mode so the DC tables and the target tables  are blended based on ethanol content.  But if those DC tables already have boost target on one axis then there is no blending needed, both tables could be the same. 

[koracing]

Most flex fuel cars I tend to do a single WG duty table and use ethanol content as one axis on the boost target table.