EAF: The EAF process modeling site

PRMC > Processes > EAF > Interface

Interface definitions for EAF process models:

For a description of the general purpose process modelling software interface, refer to the PRMC Interface pages. Here we provide some motivation from an anticipated EAF community point of view. Then we define the input data vector xi, the parameters pk and the output vector yj for process models describing the electric arc steelmaking furnace (EAF).
Please refer to this interface by the following identifiers:
DATA_VERSION    = prm05.b (the PRMC Interface version)
PROCESS_NAME    = eaf     (the unique name of the process - never changing!)
PROCESS_VERSION = 0.5b    (the version of this process specific interface)
When there is no a(α) in the version specifiers, the interface is unchangeable because existing software and systems are using it. A b (β) indicates the interface is currently tested. New and changed interfaces will get a new number!

Motivation for an open EAF data format and model interface:

EAF process models should be comparable and exchangeable

While there are several commercial and academic EAF process models, a scientific validation and comparision is currently very difficult. The end users have almost no chance to choose a process model depending on technical criteria, e.g. best prediction of tap temperature for the 2007 data after training with the 2006 data.

EAF process models should be independent of a specific automation environment

Automation environments and process models can be developed independently and an open interface will allow the usage and/or test of different process models. On the long term, plant specific process models will fail to reach the state-of-the-art. Additionally open interfaces support modularization of the automation environments. Complexity management is much easier if there are open borders with well defined interfaces.

EAF plant operation data should be collectible, exchangeable and future-proof

By collecting and saving the detailed inputs and parameters required by the process model interface, a maximum chance for the usefullness of stored plant operation data in the future is maintained. This long term storage of detailed data will become more and more important because this data is the basis for model parameter determination by inverse modelling and thus for MPC and model based process optimization systems. The data scatter und missing data problem in the charging material properties (scrap, ...) requires long term data collection for statitical analysis.

University research can be linked to industrial practise

Due to the increased innovation rate new process routes for new steel grades need to find their way to production. Academic support may stimulate the permanent process optimization in the plants.

EAF process data format and model interface (draft):

Here, we define xi, yj and pk for the EAF:

Input data vector xi:

Include all time dependent data in SI units that may be required by process models based on physical conservation laws. Current proposal (03.09.2008): prm_process_eaf_inp.xls, prm_process_eaf_inp.csv.
For demonstration of the input data exchange opportunities, here is a template for a input file: eaf_inp.csv.

Output data vector yj:

Include all time dependent process model results data in SI units that may be required by plant automation systems, HMI's and for scientific analysis. Current proposal (03.09.2008): prm_process_eaf_out.xls, prm_process_eaf_out.csv.

Model parameter data vector pk:

Include all parameter data in SI units that may be required by process models based on physical conservation laws. Current proposal (03.09.2008): prm_process_eaf_para.xls, prm_process_eaf_para.csv.
For demonstration of the parameter data exchange opportunities, here is a template for a parameter file: eaf_para.csv.

Open questions in the EAF model interface:

Currently, there are many process observation systems using
  • Which parameters pk, inputs xi and outputs yj did we forget?
  • Are the scrap parameters sufficient?
  • Is the allowance for 9 different scrap categories sufficient (see below)? Is it always possible to classify the scrap fractions into less then 10 categories!?
  • Is the allowance for 2 different lances/injectors and 6 burners sufficient (see below)?
  • Parameters for acoustic and optical sensors!?
  • Slag foaming observation systems parameters!?
For practical reasons, the results of these measurements can be fed into the process model, even if an EAF process model should predict them in principle (e.g. Off-Gas analysis).
A problem with the electric measurements like arc voltage and current is the high sample rate. Thus the effective power and additional data like
  • Electric noise and flicker parameters
are required for process modells today. In the future, process models may take advantage of U(t)/I(t) measurements at high sampling rates (kHz)!?

Parameter definition policy:

  • The selection of parameters and inputs should not be limited to those currently obtainable or required by existing EAF process models
  • Parameters and inputs should have as much physical meaning as possible, e.g. scrap charging should be described by rates instead of total masses.
  • Only basic SI units are allowed, i.e. no tons, no BTU, no hours, no inches ...
  • Parameters of principal physical importance must be included, even if they are currently not measureable (e.g. false/leak air flow).
  • Input of data which can be predicted by one known model should be avoided, e.g. from melt temperature measurements. Such data can be used to reinitialize the model or for comparision with model outputs.
  • Parameters and inputs should be measureable in principle. The number of fit parameters determinable only by inverse modelling has to be minimized.

The scrap property and charging parameters:

The parameters describing the charging materials (scrap, DRI, ..) need to be physically sufficient and suitable for the determination of unknown scrap properties by inverse modelling. The time dependent charging process is modelled by charging temperatures (CONT_FEED_TEMP, BASKET_TEMP and SCRAP_TEMP) and mass addition rates (in kg/s) like CONT_FEED_RATE. The discontinous charging of scrap baskets can be modelled basket by basket (BASKET[1-3]_RATE) or by using up to 9 different scrap classes (SCRAP[01..09]_RATE). The modelling by scrap classes will allow for the specific determination of unknown scrap parameters by inverse modelling. E.g., these parameters are SCRAP[01..09]_RADIUS (mean radius of round pieces) or SCRAP[01..09]_SIZE_[X,Y,Z] (mean dimensions of flat pieces) and SCRAP[01..09]_HTC_MULT (mean heat transfer coefficient correction factor. The scrap/basket analysis parameters where included in order to allow for physical consistent models, although such an analysis can not be determined practically. The SCRAP[01..09]_SPEC are intended for replacing the SCRAP[01..09]_-parameters by best practise values for standard scrap classes.

Basket charging

As indicated above the contents of up to 3 baskets can be provided (BASKET[1..3]_..) and the charging can be modelled by BASKET[1-3]_RATE, e.g. by using a rate equal to the weight divided by the charging time.

Continous charging

By specification of the constant materials properties of DRI, HBI or continuos fed scrap and the time dependent temperatures (CONT_FEED_TEMP, DRI_TEMP and HBI_TEMP, currently not distinguished) and rates (CONT_FEED_RATE, DRI_RATE and HBI_RATE, currently not distinguished), the continous charging is specified.

Notes on other parameters and inputs

The most important parameters to be specified are electric power input (PEL_EFF, effective electrical power), burner gas fluxes (BURNER[1..6]_CH4..), lances (LANCE[1..2]_..), the false air inflow (FALSE_AIR) and the off gas (OFF_GAS_UP or OFF_GAS_DOWN). A physical exact model needs to now (t), while best practise models can also rely on off gas flux and analysis measurements. The process models should scale their individual fit parameters (FIT_P[01..020]) in such a way that using the default value of 1 gives a good starting point for fit parameter determination by inverse modelling.

Notes on the model outputs

A physical exact model should provide LIQUID steel, SLAG and GAS temperatures and compositions, the model outputs should also include best practise parameters provided by simplified models or required by the process model end users.

Please send us your comments!

© Dr. J. Wendelstorf  · Imprint