Artisan, Fuji PXG4 and the Coffee-Tech FZ94: auto-tuning with beans

1kg of beans rolling inside the FZ-94

Last weekend we added a Fuji PXG-4 PID to the FZ-94 roaster, Marko Luther provided the configuration of Artisan to facilitate this and we allowed the Fuji to autotune with an empty drum, finding the P, I and D variables to manage the heating elements in order to keep a perfectly stable temperature as measured in the upper Bean Temperature probe.

Aborted
Next, I gave it a try with a 1.2kg batch of beans but had to abort because the beans got quite hot while the Fuji was calmly analysing the temperatures and I wanted to keep the process within a safe margin. Beans were dropped at a Tonino #106 value so they were not really dark yet.

1kg load, 40% flow
Tonight I prepared a load of 1kg of beans and I set the airflow to 40% capacity with a drum speed of 60%. Below is a screen snapshot of the full Artisan control panel with all sliders, LCDs and pre-roast buttons. It's all pretty basic for anyone who has roasted beans before.

Full screen with explanation

Full screen of Artisan up to the start of auto-tuning. Everything is very simple and one glance is sufficient to understand it all, but just in case: Fan slider on 30% of the slider range which corresponds to 40% capacity. Drum speed at 60%, Drum Temp maximum set to 95% of slider range corresponding to about 180ºC. On top, ON/OFF toggle to start / stop Artisan communications, START/STOP toggle for roasting, CONTROL buttom to open the menu for PID commands, like entering RAMP/SOAK sequences, initiating auto-tune, initiating an automatic background-following roast et cetera. Top right LCD is time after CHARGE of bean mass. Below that, on the right, top to bottom: ET=temperature of airflow out of drum towards chaff collector, BT= the high Bean Temp probe, delta BT= Rate of Rise, BTlow= Bean Temp of lower probe in default location, DT= Drum Temp, DTmax= limit of DT before elements shut off, FAN= fan speed% of max capacity, Drum= drum rotation speed % of maximum, SVbean= target Bean Temp, DUTY= % of max power sent into the heating elements. First row of buttons below the graph to be pressed at the matching events during a roast. Bottom row: FAN on/off switch, DRUM on/off switch, two optional hysteresis buttons in case one wants to set the margin for the PID to worry about. Buttons to indicate if one, two or all three elements are manually switched on/off, and finally two PID variable settings, PID1 for warmup and PID2 for roasting.

I warmed up the machine to a BT of 100ºC and with the recently auto-tuned values this went very smoothly.

Zooming in on the warmup:

You can see how the recently auto-tuned PID settings for an empty drum work well. Without intervention of the operator the FZ-94 is cruised to a stable BT of 100ºC. Fan speed has ben set to 40 early on and from that point on no modifications were made. Orange line is the target BT value, reached at about t minus 2 on the right of the graph. DTmax is way up above the graph (green line). The highest brown line is BT and the lower light brown line depicts the BT probe measured in the lower position where it is usually less hot (closer to the room temp air inlet). Black and blue lines are DT and ET which over time tend to grow close to one another. Red line corresponds to the power applied to all three heating elements at the same time.

Blink
Next, the beans are charged into the drum, the 'roast event starts' and once more I send the "start auto-tune" command to the PXG-4 using the CONTROL menu in the Artisan screen. The bottom right LED of the PXG-4 front starts blinking, to the right of the target BT value of 100.0ºC:

.

PID variables found
Fot the next half hour, the PXG-4 analysed the response time of the bean mass when 100% of heating power is applied to raise the BT, when heating is ZERO as the bean mass cools, then 100% On again, and OFF, until at about the 32 minutes mark, the best PID values have been established to be P=10, I=449 and D=86.

The "Do Not Disturb" blinking LED stopped.

On target
I allowed the PXG-4 to continue controlling the BT and indeed, the bean mass was cruised to the 100ºC target and locked on to that temperature succesfully.

Zooming in on the auto tune phase:

Fuji PXG-4 autotuning. While warming up, the top BT probe indicates a significantly higher temperature than the default BT location probe. Towards the bottom of temperature slopes, DT and ET tend to overlap. The two 'hills' on the bottom are Rate of Rise trends for BT.

Zooming in on the phase after auto-tune has been completed:

Auto-tune completed, BT cruised to target and locked on. The lower BT probe value stable on a cooler level.

Next
So next time I will charge a load of 1kg and see if I can already use the Background Roast function of Artisan to 'auto-pilot' this roast.

Profile design
This will probably not work right away because currently the designed background will not be 100% realistic. I cannot copy a previously successful roast from the same roaster because in this new setup the dual probe in the high position is communicating with the Fuji PXG4 and as the roast progresses it sends out a higher BT signal than before.

Higher BT readings
During the earlier (aborted) auto-tune attempt at a higher temperature, FC seemed to occur around 180ºC whereas before it would be at 160ºC-170ºC so the end temperature might be about 202ºC instead of recent roasts ending at 184ºC.

When Tije first prepared the BT probe device for the Trier Insert, I was hoping to get a higher reading and this new setup seems to provide me with an even more pronounced 'spread' of the Bean Temperature which would make details of the process more controllable during the second half of the  roast.

Conclusion
Roasting PID values have been established and saved.
Getting a realistic roast profile will be next and then I can really see if the PID and the auto-tuned settings will help guide the BT along the planned roast profile.