Who is here? 1 guest(s)
 Print Thread
Fluidbed Roaster project
wes
Hi All,

I'm new here but I have spent a few hours perusing the forums so please forgive me in advance if this topic has already been covered. I am wondering where everyone is placing their T/Cs in their fluid bed roasters. I have a 2lbs Sonofresco that I am wanting to hotrod and control with Artisan.

Thanks,
wes
 
allenb

Quote

wes wrote:

Hi All,

I'm new here but I have spent a few hours perusing the forums so please forgive me in advance if this topic has already been covered. I am wondering where everyone is placing their T/Cs in their fluid bed roasters. I have a 2lbs Sonofresco that I am wanting to hotrod and control with Artisan.

Thanks,
wes


Hi Wes and welcome to HRO! Please start a new thread within Fluidbed Roaster forum so as not to cause cdrake39's to take on a parallel topic within it. Once you do this, we will cut this post. Looking forward to hearing more about the Sonofresco sensors.
1/2 lb and 1 lb drum, Siemens Sirocco fluidbed, presspot, chemex, cajun biggin brewer from the backwoods of Louisiana
 
cdrake39
Hey folks, it's been a while since I've posted an update. Roasting has been going very well - usually roasting once a week to keep up. Batch sizes are around 200-250g depending on the beans being roasted. FC occurs around the 9min mark on most roasts, but the heater is often maxed out around the 6 min mark.
I'm just curious if I could possibly increase the batch size with some design modifications? Currently using an 1800w element - am I being too optimistic in assuming I can roast more than 250g with this?
Of course the easiest route might be to add a second heating element - but I was hoping for some suggestions on design geometry changes that could improve heat efficiency? For reference, pipe diameters are 1.5" tri clamp and the sight glass is a 4" tri clamp. I bored out a hole in the valve to which I secured a mesh screen instead of the traditional perf plate design. This allows for high airflow and better lofting (max of 325g when using the cyclone, or 425g with the top of the roaster 'open').
cdrake39 attached the following images:
screenshot_20211018-205438_messages.jpg 20211010_115652.jpg
 
CK
In the past, I've used navy beans to simulate green coffee... they're a cheap way to test how ET and BT will respond in prototype roasters, and they seemed close enough in size and density.
 
cdrake39

Quote

CK wrote:

In the past, I've used navy beans to simulate green coffee... they're a cheap way to test how ET and BT will respond in prototype roasters, and they seemed close enough in size and density.


Hi CK,

Sorry, not sure I follow - the roaster is up and running, I'm just curious if I can push the capacity of it further with some design changes. Not sure if the attached photos actually got posted. I have plenty of green coffee to test with. I'm curious as to whether or not I am already at my limit with the heating element (1800W) and batch size of 250g, or if there are tweaks I can make to the roaster design to push it closer to 300g+.
 
AgressivStreetLamp
@cdrake39,

Your build is basically exactly what I want to emulate. Really well done
A few questions, are you runnin everything off a single 110/15 circuit. Standard US socket. Not sure if Canada is different.

Also. Where did you source you’re tubes, Fittings, etc… Mcmastercarr? Can you link the items by chance?

I’ve got something similar in mind.

Edit,

I found most of the components on Amazon. I am compiling a part list here and will start a build post soon.

https://docs.goog...sp=sharing
Edited by AgressivStreetLamp on 11/24/2021 1:53 PM
 
cdrake39
Thanks! It is run off a 120V 20A circuit. Heater is 1740W and Blower is around 240W (total current draw around 16.5A when blower and heater are at max, which doesn't really happen during roasting - fan usually lofts 225g fine around the 75% mark).

Tri clamp fittings were sourced via Amazon (more expensive usually, but faster shipping) and Aliexpress (cheapest route).

Quote

AgressivStreetLamp wrote:

@cdrake39,

Your build is basically exactly what I want to emulate. Really well done
A few questions, are you runnin everything off a single 110/15 circuit. Standard US socket. Not sure if Canada is different.

Also. Where did you source you’re tubes, Fittings, etc… Mcmastercarr? Can you link the items by chance?

I’ve got something similar in mind.

Edit,

I found most of the components on Amazon. I am compiling a part list here and will start a build post soon.

https://docs.goog...sp=sharing
 
AgressivStreetLamp
Awesome, You got the 3" sight chamber as well?

Quote

cdrake39 wrote:

Thanks! It is run off a 120V 20A circuit. Heater is 1740W and Blower is around 240W (total current draw around 16.5A when blower and heater are at max, which doesn't really happen during roasting - fan usually lofts 225g fine around the 75% mark).

Tri clamp fittings were sourced via Amazon (more expensive usually, but faster shipping) and Aliexpress (cheapest route).

Quote

AgressivStreetLamp wrote:

@cdrake39,

Your build is basically exactly what I want to emulate. Really well done
A few questions, are you runnin everything off a single 110/15 circuit. Standard US socket. Not sure if Canada is different.

Also. Where did you source you’re tubes, Fittings, etc… Mcmastercarr? Can you link the items by chance?

I’ve got something similar in mind.

Edit,

I found most of the components on Amazon. I am compiling a part list here and will start a build post soon.

https://docs.goog...sp=sharing
 
cdrake39
I went with the 4", but 3" should be just fine. Not sure how the RC diameter relates to roasting heat efficiency, but there is more than enough space in a 4" sight glass w/ a conical reducer for 300g+. I imagine 3" would be perfect for the same batch size

Quote

AgressivStreetLamp wrote:

Awesome, You got the 3" sight chamber as well?

Quote

cdrake39 wrote:

Thanks! It is run off a 120V 20A circuit. Heater is 1740W and Blower is around 240W (total current draw around 16.5A when blower and heater are at max, which doesn't really happen during roasting - fan usually lofts 225g fine around the 75% mark).

Tri clamp fittings were sourced via Amazon (more expensive usually, but faster shipping) and Aliexpress (cheapest route).

Quote

AgressivStreetLamp wrote:

@cdrake39,

Your build is basically exactly what I want to emulate. Really well done
A few questions, are you runnin everything off a single 110/15 circuit. Standard US socket. Not sure if Canada is different.

Also. Where did you source you’re tubes, Fittings, etc… Mcmastercarr? Can you link the items by chance?

I’ve got something similar in mind.

Edit,

I found most of the components on Amazon. I am compiling a part list here and will start a build post soon.

https://docs.goog...sp=sharing
 
renatoa
3" is 76mm, slightly larger than a popper RC which maxes out at 150-200 grams.
Same for SR glass, which is 74mm inner diameter.
For 300 grams in such space the column will be higher than a (supposed) optimal, and needs more pressure for lift.
Nothing scientifically, just some thoughts based on figures from known machines.
 
cdrake39
Fair point. With the conical diffuser in my roaster (roughly 150mm tall) a 225g charge is below where the sight glass begins (about 1" or so from the top of the conical section). Realistically I could probably fit 500-600g in there without issue, but of course my capacity is only 225g due to heater limitations.


Quote

renatoa wrote:

3" is 76mm, slightly larger than a popper RC which maxes out at 150-200 grams.
Same for SR glass, which is 74mm inner diameter.
For 300 grams in such space the column will be higher than a (supposed) optimal, and needs more pressure for lift.
Nothing scientifically, just some thoughts based on figures from known machines.
 
AgressivStreetLamp

Quote

cdrake39 wrote:

Hey folks, it's been a while since I've posted an update. Roasting has been going very well - usually roasting once a week to keep up. Batch sizes are around 200-250g depending on the beans being roasted. FC occurs around the 9min mark on most roasts, but the heater is often maxed out around the 6 min mark.
I'm just curious if I could possibly increase the batch size with some design modifications? Currently using an 1800w element - am I being too optimistic in assuming I can roast more than 250g with this?
Of course the easiest route might be to add a second heating element - but I was hoping for some suggestions on design geometry changes that could improve heat efficiency? For reference, pipe diameters are 1.5" tri clamp and the sight glass is a 4" tri clamp. I bored out a hole in the valve to which I secured a mesh screen instead of the traditional perf plate design. This allows for high airflow and better lofting (max of 325g when using the cyclone, or 425g with the top of the roaster 'open').


Nevermind I see it in an earlier reply. Thanks anyway

HI, on this picture, it looks like you have a Y piece for bean drop, how are you ensuring that the beans come out the side pipe and not straight down onto the heater?
 
iPa

Quote

renatoa wrote:

.. apply a sharp power step impulse, about 10% of range, let's say from 40 to 50%.
You have to be stable at 40% power when doing this.
...


I would like to tune my pid parameters, for now I use value you gave me to start, P=2 I=0.08 D=0
Sorry probably due to my english I am not sure to understand the sentence above (see quote).
How to proceed ?
- "apply a sharp power step impulse" do you means need to draw a test profile, some thing like this (picture below)?
- How to, with an empty chamber (probably behavior will be too different), or charge with a batch of grain, even one already roasted.

i84.servimg.com/u/f84/20/23/21/77/captur21.png
Thank you
 
renatoa
Power your empty machine with a level close enough for a hypotetical preheat temperature, say 160 C. Let it stabilize, let say this happen at 40% power. None of these values are critical, just guidelines.
When stable, start Artisan logging, increase power by 10%, thus jump to 40+10=50%, and press charge in Artisan.
Then let it go until stable again, and stop the "roast" logging.
Then post the "roast" .alog file, better to crunch numbers than a graphic image.

For judging "stable", you can use the number labeled delta BT, on the Artisan right side. Or the graphic... for me is close to stable when less than 1.

The resulting graph should be something like the attached image.
Source is screen 41 of this slide, if you want to read more theory Grin:
https://www.slide...g-workshop
~~~
renatoa attached the following image:
step-test-data-and-dynamic-process-modeling-n.jpg
 
iPa
I am doing something wrong because I tried but each time I press CHARGE soon my heater value drop to 0.
What I tried:
1) Delete background profile
2) On PID control change "Set Value Mode" from "Background" to "Manual"
3) Press "ON" got connected to the TC4
4) Set Fan speed
5) Press "Heater 30%" wait for stable value (156°C)
6) Press "START" logging
7) Press "Heater" 40% (feedback heater value -> 40% OK)
8) Press "CHARGE", then Heater feedback value went to 0

By doing so I do not give a target value to the PID, shouldn't I rather play with the SV value?
 
renatoa
No PID ! Grin
The Control button should stay inactive all the time, please uncheck start PID at Charge.
What I wrote is 100% manual power control, we don't want any other "intelligence" to counteract the machine natural response.
You should add a slider that controls the heater, and every change you do will be done using this slider, nothing else.
Fan have to be constant the whole test.
 
iPa

Quote

renatoa wrote:
No PID ! Grin


oh oh ThumbsUp
here you are
iPa attached the following file:
roastalog.zip [6.32kB / 295 Downloads]
 
renatoa
I see you removed the post, but I typed too much, so will try to explain here as short and simple as I can, even this subject deserves a separate thread.
The power output, that controls the SSR is composed of three components:

Output = P term + I term - D term

- the proportional term, P, has a simple formula: kP * error, where error is the difference between the setpoint and the actual measured temperature.
When your temperature is far from setpoint, for example right after you change the setpoint, the power jump will be also great, for example 50% if the kP is 5 and the new setpoint is 10 degrees up or down.
When the temperature approach the setpoint, in an ideal process the P term should go to zero, because error goes to zero.
But a FB machine, or any other high convective machine, due to greater air turbulence, especially when paired with a very responsive temperature sensor, like a naked thermocouple, is far from the scenario where error goes to zero, you can have easily 2-3 degrees of temperature variations in a second, that's why P term should be kept low, else you would experience high power control variations, even near setpoint. Or use other technique, described later.

Let's move to second part of PID... if P term trends to zero, in an ideal process, who is keeping the heat on to a level enough to maintain the temperature near he setpoint ?
This is the I term, I from integrative, i.e. the sum of all errors I wrote above in the P term paragraph, added over long time, thus a formula something like ... Sum(kI * error)
As you can guess, these error values could be either positive either negative, as is the process temperature above of below the setpoint. When reaching an equilibrium, as P term trends to zero, this sum of errors trends to the value required to maintain the SV value, for example 40% for 160 C degrees.
However, how much is the value for this integrative term coefficient?...
A hint for estimating the value of I term is the speed of process setpoint variation we need to be able to follow.
For example, in one section of the roasting profile the rate of rise could be 10 C degrees per minute.
This means that every 6 seconds the temperature increase by one degree.
On the other side, how much power we need for this task, i.e. increasing temperature?
Every process has a specific feature, called ultimate gain, symbol Ku, that tell us how much input change we need for a specific output change. If in a roaster we reach 160 C degrees at 40% power level, then we can deduct that one percent of power change (input) will change temperature (output) by 4 degrees. This is the process gain.
Linking together the two parts described above, the result is that every 24 seconds the power should increase by one percent in order to satisfy both equations.
If we agree, as described above, that this power increase will be the mainly a contribution of the I term of the PID equation, then this 1% increase is the sum of the average error for 24 seconds, between the profile values and the real process values. If we are happy with one degree error, then this value is 1/24 = 0.04.
You should see this value as a minimal acceptable level, if we want the real profile don't lag behind the ideal profile we try following by more than a degree, for a slope of 10 C / minute.
For other parts of the profile, with steeper slope, this value should be higher, for other parts lower...
Same for the error of following, if we want a greater accuracy we need higher integrative coefficient, and vice-versa...
But beware about this last statement... if the error become comparable with the level of noise in your process, then we have an unstable process ! And a fluid bed machine is one of the most noisy we know, not only acoustical, but also as temperature turbulence...

...
The last term, the D from PID, has the role of the braking part of the controller output.
D comes from derivative, i.e. the speed of the evolution of a process.
The formula of D term is simple : kD * (input - lastinput), i.e. the difference between actual temperature and the last sample, typical one second earlier than actual sample.
It is subtracted from the output formula, not added, as you can see in the top of the post, thus the braking character of this term.
When a process evolves too fast, which typical is a sign of malfunction, the output power is reduced, to calm down the situation.
When could be this useful in a coffee roast, and how to set the kD coefficient of this term ?
First, we should define the limits of temperature evolution speed we consider acceptable in a roasting process.
As an overall average, the temperature slope in a roast is close to 20 C per minute, if we assume we start at 25 C and we end at 225 C in 10 minutes of roast. (225-25)/10 = 20
Sure, for various machine types this is different, FB are faster roasting even in 5-6 minute, some drum machines are slower, going to 15-20 minutes... but we consider an average, just to have some numbers to crunch.
This 20 C per minute average comes from a range of values spread from 30 C per minute in starting dry phase down to 5 C per minute in final development phase.
If we choose the value of 30 C per minute the maximum acceptable to not activate the brake term, then the variation per second between two samples will be 0.5 C. If we want this variation to produce a value that not change the output perceivable, i.e less than half percent, this led us to a value for kD = 1.
How effective/helpful would be having a D term coefficient of this value in certain scenarios ?
Apparently the D term will be inactive for the whole roast, because we agreed that the slope temperature will never exceed 30 C / minute...
But, for high convection machines, as I wrote in the P term paragraph, we can have instantaneous variations much higher than this average, due to the turbulence inside such machine
2-3 C degrees jumps in a single second are very common in such scenarios, were measured and recorded.
In such cases, a D term having the value of 1, as computed above, would led to 2-3 percents power reduction, that could be useful to calm such variations. But beware that a D term coefficient too high could easily produce exactly the contrary, i.e. amplify these variations ! That's why playing with this D term is highly not recommended to beginners Grin
There is one case though, specific to coffee roasters, that anyone want to use a D term should be aware, when we are dealing with preheated machines. The charge period has a huge temperature drop, having recorded in the first seconds of charge values bigger even than 150 C degrees per minute, which would led to a bump in power output of several percents, which is exactly the contrary we want in charge moment, when the power should be kept low. For this reason, ans some others, if using a D term, the PID is recommended to be disengaged during charge, and enabled after turning point.

That's all folks.
Edited by renatoa on 02/28/2022 3:20 AM
 
iPa
@renatoa
This ahternoon, while I was on a walk with my dogs, I thought about the problem encountered this morning. And suddenly I had a click, I was wrong, I had from the beginning ignored the PID parameters page of the CONTROL button because I understood that they were only used for the PID drived by Artisan, for the firmware PID it was necessary to modify those in the user.h file !!! Not having time to test again before tomorrow I deleted the post to prevent someone from spending time on it, again a failed…

Thank you for your precisions about PID process theory, I hope to have time tomorrow to try again, and curious about the rest of your explanations. Grin
 
iPa
This morning that was very interesting, here some results.
Some remarks:
- P There is always oscillations with any values, but at 2.0 and bellow temperature stabilize below SV.
- I raises the value remained too low by P but still mysterious how to tune it better smile
- D greatly reduces oscillations

i84.servimg.com/u/f84/20/23/21/77/pidtun10.jpg

Then I made some test with a background profile, the behavior is quite different, probably due to the acceleration of the setpoint temperature stay too below the curve, need to tweak, interesting result at with P=1.5 I=0.1 D=5

i84.servimg.com/u/f84/20/23/21/77/captur25.png

Waiting for next chapter of explanations from @renatoa (started 2 posts above) Grin
 
Jump to Forum: