Using a process control and optimization
system first developed for EAFs, an advanced statistical methodology can
predict end-point carbon from off-gas composition trends and process
parameters.
The Expert Furnace System Optimization
Process (EFSOP®) developed by Tenova Goodfellow Inc. (www.tenovagroup.com) is a
dynamic process control and optimization system based on the real-time
measurement of off-gas composition. Originally developed for electric arc
furnaces, it has been applied recently for end-point control in basic oxygen
furnace steelmaking. Through a three-year funding grant from Sustainable
Development Technology Canada, a not-for-profit foundation, this first EFSOP
system for the BOF was installed on a 165-ton BOF, used to convert a nominal mix
of 120 tons hot metal and 45 tons of scrap into high-performance automotive
steel.
 |
| Figure 1: Schematic of the EFSOP system applied to a basic oxygen furnace. |
- The EFSOP off-gas analyzer, for sample conditioning and analysis. A customized purging system keeps the probe clear of dust and eliminates plugging.
- A passive infrared gas pyrometer(s) for off-gas temperature.
- A supervisory control and data acquisition (SCADA) system.
This first application of the EFSOP analysis system proved highly reliable, with over 99% uptime. In addition to sampling and analysis, the EFSOP analyzer performs a secondary function of controlling the back-purging of the sampling circuit. To ensure a valid off-gas sample throughout the blowing period, the system is only purged during natural breaks in the process (e.g., during charging and tapping). This is more than sufficient to keep the probe from plugging.
Composition measurements, as well as operational alarms and outputs from the analyzer are linked to the plant’s PLC network. The EFSOP SCADA computer is linked to the same network and reads and logs off-gas data, as well as all relevant process data at a frequency of one second. In total over 300 BOF parameters are sampled and logged in real time. Both historical and real-time plots of the data are made available to the operator. Off-gas data, process data, and EFSOP system alarms are emailed to Tenova Goodfellow, allowing process engineers to follow the operation remotely.
 |
Figure 2: A profile of the measured downstream off-gas composition and temperature for a typical heat.
|
End-point detection
A primary objective of BOF steelmaking is
to achieve some desired end-point temperature and grade composition at the
lowest cost and in the shortest time. To do so, operators rely on standardized
BOF practices and static charge models. These models are mass and energy
balances that account for the initial conditions (scrap and hot metal
temperatures and compositions) and the desired end-point conditions of the bath
and slag, and indicate to the operator the expected total oxygen and fluxes
that are required to reach that end-point. The blow is stopped once the
pre-determined amount of oxygen has been reached. In addition to the charge
model, the operator relies on other cues, such as the change in the color of
the flame at the mouth of the vessel and a characteristic drop in the steam
flow in the fume system cooling circuit, to identify when carbon has been
depleted.
In practice, static charge models have a limited
ability to predict end-point because they do not account for process dynamics.
End-point accuracy also is affected by uncertainties in the initial conditions
(e.g. mass, temperature and composition of the hot-metal, mass and type of
scrap and fluxes added) and by variations in the efficiency of the oxygen
lance, not only within a heat as the height and flow of the lance is varied,
but also from heat to heat as the lance wears and the geometry of the vessel
changes with refractory wear.
The limitations of the charge-model based
end-point determination are demonstrated in Figures 3a and 3b, showing the
error between the aim and measured end-point. The data was collected over one
month of operation and is based on 400 heats. Temperature and bath carbon were
measured by “bomb Celox” samples taken at the conclusion of the heat, as
dictated by the charge model.
 |
| Figure 3: Error distribution in end-point carbon and temperature, using the charge model. |
Figure 3a is a histogram of the error in
carbon (measured as points of carbon, i.e. % × 100). The average error is about
0.1 points of carbon with a standard deviation of 0.8. The measured carbon at
first sample was, on average, lower than the aim. This particular operation
tends to over-blow its heats, with respect to carbon. Figure 3b is a histogram
of the error in temperature (measured in °Celcius). As indicated in that
figure, the average error in temperature is -20°C with a standard deviation of
25°. The temperature at first sample was, on average, higher than the aim;
indicating that the heats are also over-blown with respect to temperature.
Over-blowing impacts not only yield and productivity but also has a significant
environmental impact.
If too much carbon is removed from the bath
it must be replaced in the ladle to meet cast specifications. The extra carbon,
first removed and then replaced, unnecessarily contributes to additional
greenhouse gas emissions. Despite the tendency to overblow, approximately 7% of
the heats were found to be more than 1 point of carbon above the aim; meaning
that the operator had to re-blow the heat after the first sample.
EFSOP end-point detection
The EFSOP strategy for end-point detection
uses real-time off-gas composition, along with measured process variables, to
determine more accurately when the temperature and carbon end-points have been
reached and signal the end of the heat. The online information is used in two
ways: Advanced multivariate statistical modeling of the process; and dynamic
state-space modeling of the process.
The statistical component, of the EFSOP
end-point predictor, is based on the fact that the off-gas profile is fairly
consistent, from heat to heat, with respect to shape. It is well accepted that
the kinetics of decarburization are driven by the rate of mass transfer of
dissolved carbon to the reaction interface between liquid metal and iron oxide.
At high carbon concentrations (approximately greater than 0.3% carbon), the
mass transfer rate is sufficiently high that the rate of decarburization is
controlled by the rate of oxygen supply to the steel bath. Below this
concentration, the rate of decarburization is limited by the rate of carbon
diffusion to the reaction interface. This mechanism is evident in the off-gas
profile where CO2 concentrations tend to remain fairly constant throughout the
heat, and then to decrease sharply as carbon is depleted near the end of the
blow.
This feature, along with other process
inputs, was used to develop a statistical model of the profile of decreasing
carbon consumption at the end of the blow. An evaluation of the methodology was
conducted to determine the accuracy with which bath carbon is predicted. Over
200 heats were evaluated off-line. Figure 4 plots the results and shows the
cumulative percent of heats that fall within the error interval indicated. The
error is determined as the absolute difference between the predicted carbon and
the measured carbon. The figure shows that the statistical model for carbon
end-point is able to predict within one point of carbon about 95% of the time.
This is a significant improvement over the plant’s historical operation, in
which the carbon end-point was within one point of the aim for only 85% of the
time.
The EFSOP dynamic model component makes use
of real-time off-gas composition to calculate a dynamic mass and energy balance
over the course of the blow — unlike the static charge model approach, in which
only initial and final conditions are taken into account. The off-gas
composition and temperature are used to calculate, in real-time, carbon, oxygen
and enthalpy balances of the gas-phase of the process. From the carbon balance,
the rate of decarburization is determined over the course of the blow. The
oxygen balance provides information not only of the total rate of oxidation,
but also the extent of post-combustion (CO to CO2) and the relative fraction of
oxygen reacting with either bath carbon or participating in slag-forming
reactions. An enthalpy balance of the gas phase makes it possible to calculate
energy leaving the system with the off-gas. Any remaining energy is either lost
through the walls of the vessel or attributed to heating the bath/slag or
melting and heating of scrap.
 |
Figure 4: Cumulative distribution of the prediction interval using EFSOP end-point model.
|
The greatest challenge encountered in the development of the dynamic model has been tuning the model to the process. The accuracy in the initial conditions (i.e. initial mass and composition of scrap and hot metal), as well as the accuracy in the measured final bath conditions (i.e. measured end-point carbon and temperature and slag analysis), will influence significantly the predictive ability of the model. Uncertainties in the precision of the input data are common in steelmaking, however the current EFSOP installation confirms that efforts to improve input data precision will ensure a reliable tool for end-point prediction.
Future work
Off-line evaluation of the EFSOP approach to end-point prediction indicates that carbon end-point can be identified with greater accuracy than is possible with the plant’s current charge-model approach. The EFSOP off-gas based statistical model is able to predict carbon within one point of the measured value, for 95% of cases. Based on these results, the plant has implemented the EFSOP off-gas system online for end-point carbon prediction. Online trials have started and are on-going at this writing.
A dynamic model of the BOF process, based on real-time offgas measurements, has been developed and tuned. The model is being validated off-line and initial results are promising.
In a subsequent phase of this project the EFSOP off-gas analysis system will be used to control post-combustion in the BOF vessel. Building upon its successes in optimizing post-combustion in EAF steelmaking, Tenova Goodfellow aims to develop a system to control and optimize post-combustion in the BOF. The off-gas model provides a dynamic measure of the extent of post-combustion occurring naturally in the process. Intentions are to use this information, in a feedback approach, to control both oxygen flow (primary for decarburization, and secondary for post-combustion) and lance height, in a dual-flow oxygen lance. It’s expected that energy recovered through the implementation of postcombustion will allow an increase in the scrap to hot-metal ratio to increase productivity in this hot-metal short operation. A reduction in the hot-metal to scrap ratio also will provide environmental benefits by reducing greenhouse gas emissions (kilograms of CO2 per ton of steel produced) from the integrated blast furnace/BOF process, measured as total CO2 per ton of steel produced.
Joseph Maiolo is the manager of Technology & Development, and Doug Zuliani is the director of Sales & Business Development, both with Tenova Goodfellow Inc.,
http://www.tenovagroup.com/
Fax +1 (905) 567 3899
_
