Home grown Kiln controller

[Robert Coyle]

Hi Cone sixers:

I am working on a Java based ceramic kiln controller. I have a proto roughed out that works through an Arduino and a solid state relay. I have implemented control logic but I could use some input on how to make it pretty “fail safe”. By this I mean it turns off the kiln if things go wrong and try's to compensate for variation in temperature tracking.

So far, I have put a check in the Arduino software to turn the kiln off permanently if the temperature measured is greater than 1200 C. This is close to the limit of a K type thermocouple, but can be adjusted to just above cone 6 max. There is also a check to see if the Arduino is still getting commands from the computer. If it doesn't get a command in one minute  it shuts off the kiln.

On the Java side, I also have a check that aborts the run if the temperature reading is greater than 150% of what the program is looking for at any given time. I extend the run time if the maximum temperature designated is not reached by the end of the run until it does.

The complications, of course, is that there may be a complex up and down temperature gradient programmed, with temperature hold times between ramps. This is fine if the temperature tracks the gradient, but if it is slow, I'm not sure what would be the best approach to compensate.  No problem with too fast... all I can do is turn it off. Anyway,  usually cool down rate can be controlled easier than heat up, except maybe for glass.

Any feedback you have will be appreciated.

Bob C.

[Norm Stuart]

Robert - The programmed fire needs to calculate and update a desired target temperature.  The fail safes need to compare actual kiln temperature with the calculated target temperature, rather than an absolute.

The predicted temperature only has to be modified by absolute ramp boundaries when the user has indicated a temperature change the kiln is not capable of carrying out, such as a 9999 degrees per hour up or down.

I appreciate the pre-programmed Cone Fires in the Bartlett, where I simply call for a Cone 6, then a particular firing speed, and the software adjusts the firing of that particular load as needed to accrue a Cone 6. With that part of the firing controlled by the software, all we need to do is add on options like: a peak hold; our slow cool; or preheat for candling.

Of course we could hand-program something really complex, but even a fairly complex cycling-cool for macro-crystalline glazes can be stored in the "User 6" Slow-Cool.

Bartlett V6-CF Controller User Guide PDF

Th Error Codes in this Bartlett Manual on pages 17 and 18 are certainly a guide for what should be monitored.

A solid state relay should eliminate the very rare type of failure where a relay fails in the on-position - and there's nothing the software can do to fix that.

Good luck with your controller project.

[Robert Coyle]

Thanks Norm

The Bartlett control docs gives me plenty if things to think about. I'm not sure how fine the time slices have to be for a kiln ramp. Right now I am using a 1 min increment. That means that even a 600 deg/hr gradient will have less than a 10 degree change per minute.  I get almost that amount of signal jitters anyway.

I will go along with Bartlett and use a 50F over temperature trigger to shut things down rather than an absolute %.

I would also like to throw in a running temperature times time  "heat work" calculation, if that might more closely follow cone behavior. Not sure whether anyone else does this, or if it would tell you anything useful. Just a thought.

[Norm Stuart]

If you install solid state relays, in the hidden menu you can choose a CYCL time of 0 for turning the elements on and off, which creates a 200 ms cycle time, according to the V6-CF technical manual.

http://www.bartinst.com/kilns/1

My recollection of talking to the Tech Support at Bartlett was the heat work was determined by recording the temperature every three seconds. They could confirm the actual timing for you, after they return in January.  It's a pretty robustly designed controller.  Making your own controller is a fun project, but the V6-CF controller by itself has a retail price of only $219.

The Bartlett V6-CF is on the left.  Our Cress E23 kiln has a customized version of the V6-CF shown on the right.

.

Most buy a more costly $530 retrofit kit which includes a box for the controller along with a step-down transformer, fuse, power-switch, and thermocouple designed to replace a kiln-sitter.  It's the controller on the far left. I'm sure you'd be buying only the V6-CF controller from Bartlett and adding the rest yourself.

http://www.clay-king.com/kilns/olympic_kilns/electro_sitter.htmls

[Robert Coyle]

Yeah Norm, the price of controllers has come down a lot over the past few years. The advantage of  a PC based system is portability. I have three kilns and just plug in the themocouple leads and relay leads into the PC controller. They all have single thermocouple control to solid state relays, so I just need one. I also can use the PC to print out custom run reports which is handy.

[Chuck Colht]

I am on the same journey but perhaps not as far along. My setup uses 2 40 amp SSRs, one for each set of 4 elements in an old Duncan kiln. The new element sets draw about 20 amps each so the 40 amp each rating covers that. SSRs can fail on so I am keeping the mechanical timer to cut power if it fires too long. The SSRs are on the cold side of that. I also put a NO relay on the control lines that must be armed with a pushbutton (power feeds back from the cold side of the contacts to the coil) to allow the signal to reach the relays. In other words, if power is lost or a safety trips, the SSRs will stay off even if the controller wants them on. If they stick on, the mechanical timer will cut the main power eventually. 

Heat work can be measured as the area under the curve of the temp over time. Math speak for adding up the difference between some 'floor' (say 1900 or where ever the material starts to be affected) and the current  temp every sample time. This means that a fast rise to a high temp is the same as a slow rise to a lower (but still high enough) temp. If you graph the temp near the top using different rates, you should see what I mean. Temp going down but over the floor still counts. The magic is in that floor temp. I don't have any good refs on that and will start out just guessing, based on measurements from my computerized Skutt.

I am starting with a 60 second cycle, dividing the rate/hour by the cycle and the difference between current and target by the same number to get the next set temp.  But a kiln is pretty sluggish going up compared to going down and I'm having trouble tuning the PID.

Right now I am controlling it manually using a front end I found here: 

http://brettbeauregard.com/blog/category/pid/front-end/

Great stuff and an excellent into to PID theory. 

[Robert Coyle]

Interesting link Chuck, I downloaded the code.

My application is pretty basic. I don't use PID, just measure temp every second or so and turn things on or off. If you look close there is some temperature wobble but over a 1000 + degrees, it looks to follow the ramp curve pretty close. I have been using this type of control for about  six years now and never had a problem.  I switched over from VB to java for my front end and slightly modified the control algorithms.  Gave it a test run the other day and it seems to work fine.  I need to add logic that insures end points of ramp cycles. The old VB logic didn't proceed to the next ramp until the endpoint of the last was detected. This meant that if the temperature lagged,  you still had to ramp up at the  rate designated by the next segment, even though that might be slower and you might want  catch up.  I might add some logic that sees we are running slow, and tries to ramp up faster to try to get back on time.

As for heat work, I just start my calculations at 1200 F.  I figure anything before then is just driving out water.  My ramp endpoints are still temperature, but it would be possible to use calculated heat work maybe instead of maximum attained temperature. I don't think anyone does this.

If you ( or anyone) wants a copy of my app to play with, send me an e-mail and I'll get it to you.

[Marina Reijsmeijer (Kleierij)]

I have a home built controller as well and because it is having problems I have a question for you. My thermocouple is connected to a multimeter with an RS232 interface to a laptop. The multimeter is old and I need to replace it. Does your controller use a multimeter too or can Arduino measure voltages? I am trying to find a replacement multimeter with a published protocol, preferably C or C#. I have found multimeters with an RS232 or USB output, but no api/protocol..

[Robert Coyle]

Marina Reijsmeijer (Kleierij) said:

I have a home built controller as well and because it is having problems I have a question for you. My thermocouple is connected to a multimeter with an RS232 interface to a laptop. The multimeter is old and I need to replace it. Does your controller use a multimeter too or can Arduino measure voltages? I am trying to find a replacement multimeter with a published protocol, preferably C or C#. I have found multimeters with an RS232 or USB output, but no api/protocol..

The Arduino has multiple 10 bit A/D converters. It reads from 0 to 5 Volts in 1/1024 divisions. This is plenty for a kiln controller. The Arduino is programmed in C. There is lots of info on the Arduino web site on just how to do this and how to  interface from a thermocouple either through the analog in or through the  MAX 31855 chip,which is designed for this application. The Arduino connects directly to the computer through a USB port.

[Robert Coyle]

I put all of the code, jars, dll,s etc into a zip file on my website

kiln controller link

Buy a cheap Arduino UNO and give it a try.... Read the readme.txt file for more info

[Marina Reijsmeijer (Kleierij)]

Thanks, it is an excellent idea. It will be much more fun to learn to interface with Arduino than to try and communicate with an unsupported voltmeter. Very kind of you to share the code.

I think the discussion covered all features a controller needs. Mine works with a firing plan that can be input with mouse clicks or by selecting desired temperatures and the rates. I have no warnings for a firing that does not reach the required temperature. I just go and look how it goes when it is 1100C. Sometimes it stays below the curve and i change the graph so it will reach it. The blue line is the input, the red the actual firing. I'd be happy to share the code with you, but it is not very well documented.  The program ignores the spikes from the thermocouple.

Robert Coyle said:

I put all of the code, jars, dll,s etc into a zip file on my website

kiln controller link

Buy a cheap Arduino UNO and give it a try.... Read the readme.txt file for more info

[Robert Coyle]

I attached a pdf report file from my app. and a couple of screen shots of the user interface  It is not a real run I did it with a potentiometer attached to the Arduino  rather than a thermocouple.

I am coding some logic to address the problem of kiln lag time. I am going to start out simple. Every minute I will check to see if the measured  temperature is within 10 degrees of the target temperature. If it is, I go on to the next increment, if not I do not advance the temperature. The run time plot on the graph will keep advancing in either case, so you will know what the kiln really did.

Any ideas would be helpful

[Marina Reijsmeijer (Kleierij)]

Hi,

The only check we do is whether the kiln temperature is below or above the graph (stookplan). If it is below the kiln relais switches the kiln on, always full power. As you can see from my previous picture, at 1100C the temperature does not reach the graph anymore, so it keeps firing. which is ok as long as the ramp/plan does not start the cooling until the required temperature and hold are reached. I do a manual extension of the peak of the graph to establish this and I watch the firing until it can start cooling. No clever algorithm here. I just watch it for a half hour or so.

Below the ui for creating and changing the graph. Sorry all in Dutch :)

I think your algorithm will work as long as you take an average of several temperature measurements and reject spikes. Do you intend to change the firing plan accordingly? You may want all 3, the original plan, the adapted plan and the actual firing data.

The x is hours y is temp C. I ran it without a thermocouple, hence the error.

Robert Coyle said:

I attached a pdf report file from my app. and a couple of screen shots of the user interface  It is not a real run I did it with a potentiometer attached to the Arduino  rather than a thermocouple.

I am coding some logic to address the problem of kiln lag time. I am going to start out simple. Every minute I will check to see if the measured  temperature is within 10 degrees of the target temperature. If it is, I go on to the next increment, if not I do not advance the temperature. The run time plot on the graph will keep advancing in either case, so you will know what the kiln really did.

Any ideas would be helpful