Homeroasters - Home Roasting Coffee Community » BUILDING A ROASTER » Drum Roasters
Who is here? 1 guest(s)
Page 2 of 2: 12
 Print Thread
Is automated Artisan PID control of a drum roaster a "fool's errand"?
Yasu
Top  #26 Print Post
Posted on 05/29/2025 1:55 AM
Controlling ET is easy with PID settings.
First, warm up the machine sufficiently without adding beans until the exhaust temperature reaches 200°C.
Then, set the SV value to 180°C and perform auto-tuning.
Record that value.
Next, perform step-by-step tuning at 190°C, 200°C, 210°C, 220°C, 230°C, 230°C, and 240°C, and record the PID values.
These values are entered into the PID palette for each PV temperature band and implemented.
Based on this, setting the upper and lower limits for the MV value completes the temperature setting.

Next, use the ramp soak setting to arbitrarily determine the ET value for each time interval, and adjust the BT result to approach the ideal profile.

The omron controller has a mode that automatically switches target values and performs auto-tuning in each temperature zone to determine the PID value, and these settings can be completed in about 30 minutes,
Edited by Yasu on 05/29/2025 3:36 PM
 
Hoejmark
Top  #27 Print Post
Posted on 05/29/2025 12:03 PM

Quote

renatoa wrote:

In control theory (and practice) everything starts with the step response test.
https://en.wikipe...p_response
This shows a lot about the process behavior, and could provide the answer to your question.
Short story, I would preheat the machine to a stable state, as you do for normal charge, let it settle for at least a minute, then turn the gas up with a step at least 10% of your normal gas range.
Record the temperature evolution until it settle again, and post the log file. As data file, not image, please.

An alternative would be to ask the Artisan author for a profile following feature based on power input, not temperature. Would be much simpler to implement than actual PID profile following, I did it in my TC4 fork. Such approach works well for same beans in same day, or in same environment conditions as the master roast was been done.



Hi renatoa

I could do this with my roaster if you are interested?
 
renatoa
Top  #28 Print Post
Posted on 05/29/2025 2:30 PM
In fact, you will do it... and not just once, but many times smile
Because it is part of the process recommended above by Yasu... and it needs to be done at several points, for several temperatures.
I say this because the step response test is part of the PID tuning process, read more on the internet about the Ziegler-Nichols open loop tuning method.
 
Yasu
Top  #29 Print Post
Posted on 05/29/2025 3:14 PM
It's just my own theory.

Successful automatic control using PID in a drum type is

The control is done using the exhaust temperature.
Because even without beans, the control delay of the device can be measured with high accuracy by auto-tuning,,

If control is applied from BT, it is quite difficult to measure accurate values because the beans keep changing during the measurement of control delay.
Auto-tuning is also difficult.

It is also necessary to deal with the drum temperature difference change.

Once I could control it well by continuously measuring the drum temperature and automatically switching the PID value in the temperature range, but since I used artisan's alarm function, I changed to the current method because it is too much trouble to manage.

If it is controlled by exhaust temperature, the drum temperature difference is self-responsively canceled, so..,


Of course, it is necessary to continuously measure the progress of roasting, the result is the bean temperature, and whether to take action (manually or automatically) from ΔBT or to evaluate what has been done and reflect it in the next roasting.

It will only be in the realm of playful fine-tuning, though,
Edited by Yasu on 05/29/2025 3:24 PM
 
renatoa
Top  #30 Print Post
Posted on 05/30/2025 2:46 AM
You could be right for drum machines where roasting is done mainly via convection.
There are exotic machines though, which looks as a typical drum, but internal they are using a significant radiant heat quota for roasting, as is the FZ94.
For those machines ET has a very misleading evolution... at the end of the roast, about last third, air is cooler than beans ! hard to elaborate a rule for such cases...
 
Yasu
Top  #31 Print Post
Posted on 05/30/2025 4:46 PM
Yes, I think it depends on how the hot air is channeled through the drum type.
I think it is popular to use a hot-air type where the combustion part is separated from the drum and the hot-air is applied to the beans from the perforated metal part from the back, instead of applying it to the drum periphery.

I think the double drum is of that type.

My roasting machine is a hybrid type that combines 10KW of gas combustion and 3KW of electric heater.

I wanted to use a baking method that prioritizes radiation and heat transfer depending on the beans, so I set the hot-air generator to vary from 0 to 3 kW and directly apply it to the beans, changing it to 0 kW to give priority to radiation and heat transfer, or 3 kW to give priority to hot air, depending on the beans.

I think I have PID automatic roasting by exhaust temperature in both patterns so far.
I think there is a difference because I switch the palette of PID, but is it a coincidence that the balance is good?
Lucky me.
Edited by Yasu on 05/30/2025 4:53 PM
 
renatoa
Top  #32 Print Post
Posted on 06/01/2025 2:04 PM
Some like it hot... Grin

https://github.co...d_hardware
 
Yasu
Top  #33 Print Post
Posted on 06/02/2025 4:22 AM
Even if we aim for automatic roasting, I think the definition of automatic varies from person to person.
I have considered various offerings.
1. It would control the heat, exhaust, and drum speed for each hour of roasting, tracing them at any value, or it would automatically adjust to regenerate the values of previous results.

2. It will automatically bake BT, ROR, etc. under certain conditions (if the bean brand, weight, moisture content, etc. are the same) with any profile
(infinite number of programs, files or manual intervention required)

3. Even if the conditions change (bean brand, weight, moisture content, temperature, season), it is possible to automatically bake BT, ROR, etc. with any profile at
.
(Basically, only the number of profiles (roasting method) is required, and the roasting machine automatically corrects for any differences in conditions.)

What is your goal of automatic roasting?
Or, please let us know at
that you are getting to this level.
Edited by Yasu on 06/02/2025 4:26 AM
 
renatoa
Top  #34 Print Post
Posted on 06/02/2025 6:58 AM
For me the profile means a target dry time, a target FC moment, and a development according to origin and desired roast level.
How to get there, to that dry end and that first crack is exclusively a matter of machine natural curve, i.e. applying as little corrections as possible.
How do you find that curve ? By measuring and analyzing the step response during the preheat phase.
For those unfamiliar with these concepts, the short story follows: every system in this world has a response when applying a stimulus.
Think to the step ahead when you get a kick in the ass Grin
The typical response of a system output (beans temperature) to a input (heat) change looks like in the attached image.
That curve parameters are all a matter of machine build, environment conditions, and beans characteristics.
There is no way to force a system behave different than such natural response curve.
The key is to match your desired profile with a sequence of such response slopes, and try having the least possible heat changes, for the least beans stress.
The ideal would be to apply power changes as close as possible around the inflection point.
Increasing power too early will result in a steeper start of RoR curve, that led too early touching the summit where the power capacity of the machine is running out, thus following an abrupt RoR crash. Just like an athlete who didn't know how to dose his effort collapses.
Increasing power too late will result in a RoR curve dip, if too many dips the RoR curve lost its smoothness.
The second image shows such case... the red/orange line is the heat, the power increase was too late, thus the dip in the yellow RoR curve... power should be increased earlier, where the arrow is pointing to the start of the dip.

To assist me taking the decisions how to handle the power, my roast controller is computing the RoR trending, to signal me the inflection point of the step response curve.
During dry phase I am applying successive power adjustments in those inflection points, until reaching the desired RoR level for my desired dry phase.
Permanently the roast controller is computing a dry end estimator, precise enough to reach DE in a 10 seconds ballpark.
During browning phase the logic is the same but reversed, now I am aiming for a target RoR at a desired time of FC... and go there under an ideal RoR decrease slope.
The power decreasing steps are more tricky than for drying, the step response for cooling is not so simple as for heating.
Here I still need manual input, this part can't be automated as easier as the drying phase.
Especially when you have an IR heating significant quota in the energy mix, where beans color enters in the scene as a supplementary variable to heat and airflow.
renatoa attached the following images:
image_2025-06-02_145302877.png image_2025-06-02_145235366.png
Edited by renatoa on 06/10/2025 6:01 AM
 
Yasu
Top  #35 Print Post
Posted on 06/02/2025 11:23 PM
Is it correct to understand that the parameters of the curve depend on each bean (weight and other conditions), and that it is necessary to change the parameters for each condition?

2. It will automatically bake BT, ROR, etc. under certain conditions (if the bean brand, weight, moisture content, etc. are the same) with any profile
(infinite number of programs, files or manual intervention required)

Does renatoa roast using the type of automation described in 2?
 
renatoa
Top  #36 Print Post
Posted on 06/03/2025 3:39 AM
Yes, I am roasting as described above, but manual, not automatic.
Described the steps of the process and the controller screen in the RoastAssistant thread.
Where is the pleasure to push the button and be a spectator ? Grin

The curve parameters are stable enough on a known machine, for a fixed load and ambient conditions.
Beans origin influence is too small. The most influential are the processing type and moisture. But not more than 2-3% of the power control range.
 
Yasu
Top  #37 Print Post
Posted on 06/03/2025 4:58 AM
Automatic roasting is enjoyed with the success, accuracy, reproducibility, and purpose of automation to enjoy in this thing itself.

Now, even if the data such as brand, weight, moisture content, etc. are different, we can roast the same profile no-touch just by changing the residual heat conditions.

Of course, I also enjoy switching to manual operation with analog dials for all operations such as combustion temperature, heater output, etc., depending on my mood.
 
renatoa
Top  #38 Print Post
Posted on 06/03/2025 6:25 AM
Well... there are machines and machines... not every ET is the same for all.
I can't apply your routine in my 100% radiant roaster, where the air is cooler than the beans !
It tops about 160C after dry end, then start to decrease slowly, as the fan % is increased.
The air is a cooling factor all the roast time, no moment is involved in heat transfer.

Meanwhile... BT is still BT... and obeys to the step response rule.
 
Page 2 of 2: 12
Jump to Forum: