Continue to Site

Furnace control, increase the temperature over time

asrock70

Full Member level 4
I built a high-temperature furnace, the first layer of Promaform insulation with a resistance of 1430 C, heating elements Kanthal wire A1 max 1400C. Installed power approx. 3 x 3000W. I am currently making heating elements.
Temperature measurement inside 2x Type K thermocouples, with 19 bit ADC.
Use, primarily for baking molds for precision casting, temperatures up to 1000C, but a precise increase in temperature must be observed, for example 150C/hour or hold a certain temperature for x hours.
Secondary, melting of copper and aluminum alloys, temperatures up to 1100C.
I'm secretly hoping that it could manage to heat gray cast iron to a casting temperature of 1300C, I really don't know if that's realistic.
I'm starting to figure out how to control it, whether it will be some kind of STM32 or ESP32 or something like Raspberry Pi. It's purely about the interface with the environment, in terms of performance, x52 would be enough.
If I calculate correctly, the output per second at full power and one phase is 3kJ/s, with three then 9kJ/k.
If the request is for a maximum temperature increase of 150 C, it is 0.0416667 per second, the resolution of temperature measurement with a 19-bit ADC is 0.0078125, i.e. 5.3 LSB, with a temperature increase of 100 C/h, it would be 3.5 LSB.
My idea of driving.
The operator will enter something like
1h 150C
1h 150C
1h 400C
3h 400C
2h 900C
3h 900C

Meaning, increase the temperature to 150 C in an hour, keep it at 150 C for another hour, then increase the temperature to 400 C in an hour, etc.
Does the idea of controlling temperature 18C increase to 150C per hour ie. deltaT/s = 0.036666 C/s i.e. 5 LSB rounded. If rise <= next second heating On, otherwise Off.

Is it a very bad idea, or is there a more sophisticated way, something like PID, but also monitoring the rate of temperature rise, not just the target temperature?

A lot of assumptions are missing about the distribution of heat, the size of the tank, the mass of metal, the thermal gradients, the oxidation process, the toxic fume management, the removal of slag. Have you considered circulating the hot air and an e^-T controlled ramp to limit the thermal gradients?

PID would, if implemented, use rates as well. But given thermal masses involved (I assume significant)
PID may not really add any value. eg. just first order measurements sufficient.

Operator table entry of profile certainly pertinent.

You using non contact T sensing or RTD....?

ESP32 analog not very good, still its a great part, and dual core allows one to
implement easily safety checks concurrently with normal processes. PI overkill
unless you want to run a full blown OS.....

Another series of parts, including 20 bit DelSig, and precision Vref is this SOC.
On chip these resources as well as multiple copies of many of them :

Even custom components can be created using schematic capture and/or Verilog.

Tons of projects to cut and paste from. Users have even created one chip Oscilloscope
using one chip. Here is RTD basic project, notice in resource window very few resources
used, eg. tons of other resources for other tasks available.

Regards, Dana.

Last edited:

Hi
Temperature measurement inside 2x Type K thermocouples, with 19 bit ADC.
In my eyes this makes no sense.
Let´s say you have a merange of 2000°C then the resolution would be 0.0038°C. You get nowhere near this value regarding precision or accuracy.
.. or 140nV = 0.14uV

Noise, unlienarity, measurement problems, timing ... will cause erros several decades worse.
--> be happy to get a useful resolution in the 12 bits range.
--> be happy to get the accuracy in the 8 bits range.

Even if you calculate those values like 0.036666 C/s ... it does not work this idealistic way. Reality is different.

rise <= next second heating On, otherwise Off.
This looks like an ON/OFF control. Here the resolution above makes even less sense.

Then you talk about several kW of heating power. How is it controlled?
Relay? SSR? phase angle control?

If ON/OFF per relay ... you probably want to reduce switching counts. .. using low frequency PWM style (like electric heat plates do)
With software you can improve a lot. Like calculating temperature difference, heat capacity, heating power, heater percentage ... even in look ahead mode.

Klaus

What is your criteria for the maximum oven hotspot?

If you use a Triac phase to adjust power for this higher temp and use a second thermistor to regulate the heat on and off for the target temp, you won't need anything more for servo control.

I suspect you have an SDC ADC with a 16 bit output and 3 bits are selected and discarded..

I'm back from vacation so let me give you some details.
The first layer of heat insulation is 10 cm boards made of aluminosilicates and aluminum oxide. Weight approx. 50 kg, c=1.13 kJ/kg.K.Second layer 50 mm mat weight approx. 15 kg, c=1.14 kJ/kg.K.
theoretically I can add a third layer of insulation 5 cm stone wool, 6 kg c=0.9 kJ/kg.K.
Heating is resistance wire Kanthal A1 3x 230V max 16A (in reality approx. 13A), switching 25A thyristors. The way it is switched is up to me, but within the 1s time slot I would like to stay only with On/Off angle control, in my opinion it makes no sense.Temperature measurement. MAX31856 (19bit ADC) + thermocouples K (B) Special Limits of Error: +/- 1.1C or 0.4%
Critical from the point of view of monitoring the temperature increase is the situation with the "baking" of the molds. The mold is a stainless steel tube with something like high-temperature gypsum in it. See for example here
The smallest mold can weigh around 200g, the largest one 1kg, up to 4 large ones can fit in the oven.

Hi,

so you have the weight, you have the thermal capacity, you have the temperature rise rate --> then you may calculate the required power

Additionally you will need power for
* thermal loss to the environment (thermal conductivity, radiation..)
* power required for the thermal reaction (may be negative)

****
Limits of Error: +/- 1.1C or 0.4%
I guess what for the application is important is the temperature of the mold.
So is your temperature sensor placed inside the mold?
If not: What temperature deviation do you expect between thermocouple and mold inside?

****
I would like to stay only with On/Off angle control
there is:
* ON/OFF control and
* angle control
This are two different modes. Never heard about "ON/OFF angle control".

Klaus

This makes an interesting thermodynamics problem if one estimates the Rth of each outer layer of insulation and the inner mold from surface to core center of mass. But one needs to know the max. mold gradient or thermal rate of change before thermal stress becomes destructive. This may require experimentation.

There will be an air thermal gradient away from the heat source and gravity without an air stirrer. Does this thermal mass time constant matter from bottom to top in air and from surface to core? This depends on the conductive resistance and thermal mass time constant. With a Kd derivative gain factor or "look ahead value", you can prevent overshoot of the surface temp. and if desirable, reduce and optimize bake time.

But if your proportional rate for temp is much slower than the thermal mass time constant, the overshoot error can be minimized with proportional control.

Last edited: