Control strategy for an Annealing burner


Thread Starter

Marco Rivas

Dear List;
I am exposed to a Annealing burner control project. The burner's length is near 50 meters. The material (sheet steel) is rolling inside this gas operated burner in an inert atmosphere. A temperature profile has to be reached: Heating zone (926 oC), then soaking zone(900 oC) and controlled cooling zone (400 oC).

A colleague suggested me to hire another consultant to develop a model for the burner and then apply different control strategies. The burner is more 20 years-old.

Could you give me some tips to face this project?

Best regards

"If it ain't broke, don't fix it."
I am not sure of your meaning of "model" in this case, but if by model you meant generate a set of algorithms that represent the final surface or case hardness based on the key process variables, I have found this to be a difficult task with very few consultants available at an acceptable rate.

I only have a few years experience in this area, so if you get better advice, take it. But here is what I suggest.

1. Automate the process with a PLC to control and monitor the available key process input variables. These input variables include the temperature at each zone or multiple points within each zone, the flow of gases into each
zone, any measurements of the atmosphere such as CO, CO2, ect., and others that you can define better than me. Don't forget to include the cooling zone.

2. Another big variable that I think many companies overlook is the raw material variance. The percentage of carbon and alloys added to the steel greatly affect the outcome of the annealing process. I think that any model that doesn't include these raw material variables is unlikely to produce acceptable results unless the steel vendors can improve the consistency of the material that they send to you. It is almost a given that they can not at an acceptable price. They do usually have the quality information
available that they may give you if you ask. You need to know everything they put in it and what the final percentages are. These final percentages are inputs to your process model. You should also automate this process as well such as have the receiving office or quality enter the information after receiving the material.

3. When you check the material for quality after the annealing process, you should record this information automatically as well. Normally this
information includes the surface hardness and case hardness (hardness profile to some depth as I've defined case hardness).

4. Once all of this is automated, you can gather information for each of your products with special attention given to the special products. The products that run on the top and bottom of your specifications are very important to any process model. Normally all of this data is fed back to a central database.

5. To model this data, I recommend that you use a neural network process modeling software. First you select a set of records from the database that are representative of your entire product line and process and export these to a file that neural network software can import. This is normally flat text or comma separated files. Neural network packages can import this
data, create a model of the process, and export the model to a format such as C or BASIC.

6. To use the model as a process quality monitor, you can predict the outputs, which are your predicted quality variables, and monitor these predictions to signal an alarm or stop the process if they go out of specification.

7. To use the model for control, which is not recommended at first, you reverse your model so that the "desired" quality are inputs, and several
controllable key process variables are outputs. Your PLC can convert these outputs to signals used to control the devices which control these
variables. These devices may be temperature or gas flow devices.

8. You will then have something that is probably not perfect but much better than what you have today.

It is hard to fit much information about something so complicated into this space. If you have any questions, please email me directly.

I hope this is of some help.

Pierre Desrochers

Marco - This type of application has a few theorys beind it. One would call for theoric models reaching to point of being associated to artificial intelligence...
Others will talk about process constant. I deal with this type of process and I can tell you that the end product will determine the type of control
needed. As Tim wrote you must have an automated system to, at the very least gather the precise informations on wich to base the future decisions...
The control system should be easy to modify (PID and stuff) ...The fact is that there is a trend to have only one value (the constant) from which you could derived all other settings.

In the control strategy, the length of the Soak zone becomes a variable. You could have a 150 meter long process but need only 125 meter. Hence
speeding up the line respecting the Soak time. This process is a simple one. what is not simple is to make it work.

Entrance tunnel and preheat have a de-ox function and aneal, Soak deals with the anneal, Cooling zones are in relation with the zinc bath.

You end up with one constant for the preheater, one for the Soak(s), one for the cooling , etc

The strategy here should be to control ALL the Zone with the Zinc bath temperature controler. In theory the bath controls the Cooling zone, which controls the Soak ... all this back to the entrance tunnel.

Of course if you are developing a new strategy for the preheater zone you should only deal with precise pressure controls and the elimination of lag in the gaz/air ratio controlers. The last thing you need here is to have oxigen in the surface of the steel.

My 2 cents.

Pierre Desrochers
Integral Instrumentation Inc
[email protected]