PID anti-windup

Do we have a possibility for monitoring anti-windup?

No this is not logged. I’ll see if that’s possible.

I set the parameters you suggested. Yesterday I ran a step up to 68 deg. It didn’t seem to want to come any closer than 67.19 deg. I would expect the integrators to pull it up. A high Ti value should mean that I would have to wait longer.

Before I start fiddling with the parameters, it would be nice to have a sanity check of the system. I you would, what data should I send you?

Incidentally, do you see a possibility for anti-windup monitoring?

Export your blocks from the spark page, top right menu, export blocks.
For both PIDs, share the graph of the block so we can see what the internal values of the PID are during the run.

What I am thinking of is adding a field to the PID that shows the integrator increase per second, which is P - anti-windup.

Run on Wednesday 27.10.2020

The MT I value has some steps that I do not understand.

graph-HERMS Graph-spark-one-downsample_1m (8).csv (19.4 KB) graph-HERMS HLT PID-spark-one-downsample_1m.csv (26.9 KB) graph-HERMS MT PID-spark-one-downsample_1m.csv (26.1 KB)

Run today 29.10.2020

The process is still running, until 14:30

brewblox-blocks-spark-one.json (20.4 KB) graph-HERMS Graph-spark-one-downsample_1m (8).csv (11.3 KB) graph-HERMS HLT PID-spark-one-downsample_1m.csv (20.1 KB) graph-HERMS MT PID-spark-one-downsample_1m.csv (19.3 KB)

Screenshots of the graph are a lot easier for me. You can also export the graph as image.

Do these do the trick? They are from today 29.10.2020

Yes, that’s helpful. I see that the setpoint for the HLT is swinging too much.

MT PID
You can see that the swings in the output of the MT PID have about the same shape as the D part.
The D part is too big. Try setting Td to 2 or 3 minutes instead. Or try with Td at zero at first, then if you have overshoot, increase it to 2 minutes.
Try making the integrator react a bit slower too, with Ti at 10m.

For the HLT, Td at 10 seconds results in a bit of a ‘grassy’ value, because of the bitflips.
Here you can also set it to 0, or set it to 1 min for a bit more filtering.

You can also set some filtering in both setpoints.

Your system seems to respond quite differently compared to mine. Do you have some photos of the setup? Do you recirculate the HLT with a pump to stir it continuously?