Sunteți pe pagina 1din 9

Engineering Applications of Articial Intelligence 14 (2001) 7785

A predictive system for blast furnaces by integrating a neural network

with qualitative analysis
Jian Chen*
Laboratory of Systems Engineering, School of Economics and Management, Tsinghua University, Beijing 100084, Peoples Republic of China
Received 1 June 1999; received in revised form 1 June 2000; accepted 1 July 2000
Silicon content in pig iron has long been used as one of the most important indices to represent the thermal state of a blast
furnace. In this paper, a predictive system for blast furnaces by integrating a neural network with qualitative analysis is presented.
The qualitative trend of the process in blast furnace is predicted through causal analysis and qualitative reasoning, and the relevant
variables as the inputs of a neural network model are determined. Then, a neural network model is constructed and trained with
appropriate data. Evaluation of the system is made by comparing the predicted values with observed data (totally 610 heats are
included), and the performance of the system is excellent. # 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Silicon content; Blast furnace; Qualitative reasoning; Neural networks
1. Introduction
The iron blast furnace is one of the useful tools for
human race. It provides the means by which iron is
rapidly and eciently reduced from ore and virtually it
is the basis for all primary steel making. It is a signicant
item in the economy of any country. Because of the large
quantity produced, even small improvements of the
process can result in considerable prot. Silicon content
in pig iron has long been used as one of the most
important indices to represent the thermal state of the
blast furnace. The control of silicon content in pig iron
at lower level concentration has been regarded as one of
the most important operational problems of a blast
furnace, due to the following reasons: (1) to improve the
quality of product; (2) to meet the need of steelmaking
process; (3) to decrease the running cost, etc.
Nowadays, the iron smelting process in a blast
furnace is mainly controlled by experienced operators.
However, they cannot work forever, and not all
operators can become experts. To improve the control
performance, we need a process model that could
accurately describe stoichiometric and enthalpy balance
for the process and be consistent with the observed
silicon content in the furnace stack. There are several
methods that have been proposed to build the model
(Fielden and Wood, 1968; Jin, 1986; Keyser and Van
Canwenberghe, 1981; Li et al., 1984; Unbehauen and
Diekman, 1982). A blast furnace is a kind of giant
reactor in which gas, liquid and solids coexist, and the
reaction in a blast furnace is very complicated. Its
complexity has prevented it from being controlled
optimally. On the other hand, with the success on
rule-based expert systems in AI, people expect that the
methods of expert operators could be computerized.
Some expert systems (Iida et al., 1989; Nakajima, 1987)
have been built for increasing the operational eective-
ness of a blast furnace. To obtain useful information
from a mathematical model and also from the oper-
ators experience, a predictive system based on combin-
ing an adaptive predictor and a knowledge base has
been developed (Chen, 1993). Generally, knowledge-
based systems have been developed solely through the
use of rule-based programming, which allow for easy
modeling of expert reasoning. However, experience and
analyses showed that there were serious limitations to an
inference method based on empirical associations
between observable ndings and diagnostic hypotheses.
In this paper, a predictive system for blast furnaces by
integrating a neural network with qualitative analysis is
presented, which tries to help push back the bounds on
*Tel.: +86-10-62789896; fax: +86-10-62785876.
E-mail address: (J. Chen).
0952-1976/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 2 - 1 9 7 6 ( 0 0 ) 0 0 0 6 2 - 2
both quantitative methods and qualitative methods
alone. Moreover, with the strong expressive capability
for incomplete knowledge/information by qualitative
reasoning, it may help to make use of the missing
normative information. The qualitative trend of the
process in blast furnace is predicted through causal
analysis and qualitative reasoning, and the relevant
variables and model structure are determined. Then, a
neural network model is constructed and trained with
appropriate data. Evaluation of the system is made by
comparing the predicted values with observed data, and
the performance of the system is excellent. The paper is
organized as follows. The process of a blast furnace,
neural networks and qualitative reasoning are intro-
duced briey in Sections 2, 3 and 4, respectively. Section
5 describes the structure of the predicting system, which
is followed by the qualitative relationship of a blast
furnace in Section 6. In Section 7, the evaluation of the
system is presented and nally, Section 8 concludes the
2. The process of a blast furnace
The iron blast furnace is a tall, vertical shaft furnace
that employs carbon, mainly in the form of coke to
reduce iron from its oxide ores. A schematic view of a
typical blast furnace is presented in Fig. 1. Iron ore as
metal source, coke as heat energy source and reductant
are fed to the top of the blast furnace in charges. These
move from the top to the bottom of the furnace in about
67 h and are melted during this period. Hot blast is
blown in from the tuyeres at the furnace bottom so as to
burn coke. The heat will be determined by the blast
system that diuses the furnace from bottom to the top.
In the furnace, iron ore reacts with the gas generated by
the combustion of coke, which promotes iron to reduce
and melt into hot metal.
The principal objective of the blast furnace is to
produce molten iron of constant composition at a high
rate. For a smooth progress of the reaction, various
conditions including in-furnace heat balance, burden
descending, gas permeability and the likes must be
maintained properly. If not, it will lead to reduced hot
metal output, and in extreme cases, a total lack of
output. The critical operating parameters are the
temperature and silicon content of the iron. According
to some reports (Jin, 1986; Liu, 1985), there is an
approximate linear relation between the temperature
and silicon content of pig iron. Furthermore, the silicon
content is a good measure of the heat content and the
course of the blast furnace process (Keyser and Van
Canwenberghe, 1981). Therefore, only silicon content is
chosen as an output variable for the process model in
this paper. The silicon content must be held as low as
possible with restriction that it should not be too low or
too long below the desired value (this is a dangerous
situation in running the blast furnace). The silicon
content in pig iron is uctuating stochastically. This
makes manual control far from easy. The purpose of the
predictive system is to advise the operator whether a
planned action will be favorable or not.
3. Neural networks
Articial neural networks (ANNs) are mathematical
models inspired by the organization and functioning of
biological neurons. ANNs are composed of a number of
very simple processing elements known as neurons. A
neuron typically consists of three components: (1) a
group of weights, (2) a weighted summer, (3) a nonlinear
activation function f (x) (e.g. sigmoidal function). The
weights are regression coecients to be estimated from
sample data. The bias term is comparable with the
intercept of conventional regression model, with the
input signal being always 1 for this term.
The structure of an ANNs model is dened by its
architecture (neurons and their interconnections) and its
activation function. Fig. 2 shows a typical structure of
an ANNs model (multi-layer perceptrons, MLP) with n
inputs, a single hidden layer with three processing units
(neurons) and one output.
In MLP, neurons are arranged in layers. The output
from each neuron in a given layer is connected to the
input of every neuron in the next layer. The size of the
input and output layers is dened by the problem faced,
and the hidden layer can contain an arbitrary number
of neurons. The architecture of the ANNs model
adopted in the predictive system of this paper is an Fig. 1. A schematic view of a typical blast furnace.
J. Chen / Engineering Applications of Articial Intelligence 14 (2001) 7785 78
MLP in three-layer structure and the sigmoid activation
function. Mapping vectors from input space R
to the
output space R
, an MLP can be expressed as
. The mapping process is a dynamic
process, in which neuron inputs and outputs are
updated sequentially from the input layer to the output
layer. A MLP is trained to represent an unknown
mapping by employing a training set, that is, a collection
of paired input and desired output vectors observed
from a given problem. The purpose in training an MLP
is to determine the values of the weights of the neuron
for the network so that the MLP can closely represent
the unknown mapping. An MLP is often trained by
using an algorithm known as error back propagation
(BP) (Rumelhart et al., 1986) and this combination is
referred as a BP network. BP typed neural network
model has powerful learning ability and simple structure
that are very attractive properties for a wide range of
Let M be total number of layers of an MLP including
the input (layer 1) and output layers, neu(s; i) denote
the ith neuron in the sth layer, and u
denote the input of
the neu(1; i). The operation of neu(s; i) is dened as

= f (x
); (1b)
where y
is the output of the neu(s; i); n
is the
total number of neurons in the sth layer, w
is the
weight from neu(s; i) to neu(s 1; j), and f () is a
nonlinear activation function.
With BP algorithm, it performs the input to output
mapping by minimizing the following cost function for
each given pattern pair (U
; Y

^ y
; (2a)
where ^ y; y are the computed output and the real output,
E =

; (2b)
where N is the number of training samples. The amount
of change for connection weight w
i; j
at iteration time t
can be determined by the following equation:
(t) = m
(t 1); (3)
where m is the learning-rate coecient and a is a
momentum coecient, m; a > 0
= d
^ y
) if s = M;

_ _
) if 2 _ s _ M 1:
The standard BP algorithm, however, frequently causes
the local minima problem and/or very slow learning
owing to the typical handicaps of all steepest descent
approaches. There are many improved learning al-
gorithms that have been proposed, such as using
heuristic rules to nd optimal parameters, adopting
conjugate gradients, Newtons method and quasi-New-
ton techniques, etc. (Chen and Xu, 1998; Ergezinger and
Thomsen, 1995). From Eq. (3), we can nd that the
convergence rate lowers dramatically when the deriva-
tive of the activation function f
(x) is close to zero.
Suppose that the activation function f (x) is a sigmoidal
function, that is,
f (x) =
1 exp(x)
and its derivative takes the form
(x) =
(1 exp(x))
= f (x)(1 f (x)): (5)
When f (x) ~ 0 or 1, then, f
(x) ~ 0, and the standard
BP algorithm may converge to a local minimum. In this
case, to escape from a local minimum, it is important to
keep f
(x) ,= 0. According to the above analysis, we try
to promote f
(x)to a special value b at some points, that
is, we dene a new activation function f
(x) and its
derivative takes the form
(x) = f
(x) b; (6)
where b is a small number and b > 0 (e.g. 0.0050.025).
From Eq. (6), the value of f
(x) will never be zero. If b
is equal to zero, it is obviously that the improved
algorithm will degenerate to the standard BP algorithm.
If b is small enough, it is believed that the improved
algorithm may have the properties of the standard one.
The training procedure is the same as the standard BP
algorithm. The performance comparison between the
improved algorithm and the standard BP algorithm is
described in Chen and Xu (1998).
In general, there are some shortcomings with ANNs,
such as (1) ANNs models are dicult to interpret; (2)
ANNs contain more parameters to estimate than do
most statistical models, which may result in overtting.
Fig. 2. Illustration of an ANN.
J. Chen / Engineering Applications of Articial Intelligence 14 (2001) 7785 79
To improve the performance of ANNs models, in this
paper, we integrate an ANNs model with qualitative
analysis in a predictive system and expect to increase its
predictive accuracy.
4. Qualitative reasoning
Qualitative reasoning is a relatively new eld studied
originally from AI research and focuses on using
incomplete knowledge. It appears to be an appropriate
or even a necessary approach for complex systems (e.g.,
a blast furnace) where complete numerical information
for problem under study is not available at the time of
analysis. The basic procedure of qualitative reasoning is
to obtain system structure, i.e. components and con-
nections among them that are described by qualitative
equations or by causal constraints, then to solve these
equations or analyze the constraints, and all possible
future behaviors will nally be explored through reason-
ing. They provide greater expressive power for incom-
plete knowledge than quantitative methods. Over two
decades qualitative reasoning has developed and
achieved much both in theory and practice. Three main
techniques of qualitative reasoning have been proposed:
De Kleer and Browns ENVISION (De Kleer and
Brown, 1984), Forbus qualitative process theory (For-
bus, 1984) and Kuipers QSIM (Kuipers, 1986).
Although other interesting approaches have been
proposed, our work is based on Kuipers QSIM method,
in which the relationships between system variables are
represented by equation-like constraints. It allows us to
represent profound knowledge very easily and provides
the tree of all possible behaviors of the system studied.
The structure of a system is described by a set of
parameters linked by constraints. For each instant t, the
qualitative state of a parameter QS( f ; t) is dened by
the pair (qval, qdir), where qval stands for the quali-
tative value and qdir for the direction of change,
therefore the whole system qualitative states is a set of
the qualitative states of all the parameters. In the
simplest case, we can determine qval and qdir as follows:
qval =
if f (t) > 0;
0 if f (t) = 0;
if f (t)50;
qdir =
inc if f
(t) > 0;
sdt if f
(t) = 0;
dec if f
where inc, std and dec represent increasing, steady and
decreasing, respectively, and f
(t) denotes the time
derivative of f .
The concept of general qualitative algebra has been
introduced as the theoretical fundamentals. In the
simplest case, we can dene U = (; ; 0; ?) as the
universe of qualitative calculus, which has a lattice
structure with respect to set inclusion.
In this case, however, qualitative reasoning ap-
proaches work well only in a highly abstracted level.
When the size or complexity of the problem increases,
qualitative reasoning usually renders uninteresting
results, i.e. the results are too general to draw any
useful conclusion. In order to solve that, a more rened
qualitative universe should be represented by landmark
set. A landmark point is (1) a critical point that the
objective function or constraint equation reaches, (2)
stage segmentation point that the problem solver are
interested in. With rened landmark points, let us dene
it as a set L = L
; j= 1; . . . ; m, we can extend the
denition of qualitative value to
qval =
if f (t) = L
; L
) if L
5f (t)5L
QSIM has the ability of asserting new landmarks during
simulation [11]. Generally, seven basic types of
constraints are used in QSIM: Add (addition), Minus,
Mult (multiplication), M

; M

(monotonic links be-

tween two parameters), Constant (specifying that a
variable remains so), Deriv (derivation).
In traditional numerical simulation, a set of dier-
ential equations over continuous variables is translated
into a set of dierence equations, each variable is
assigned a real number and behavior of the problem
evolves from one time-step to the next time-step
according to numerical computations. Here, the process
of the problem is described as a set of constraints over
interval valued variables, each variable is assigned an
interval, and because of the ambiguity thus introduced,
the behavior of the problem branches to encompass all
of the possible evolutions of the problem. For the details
of the reasoning algorithm please see Kuipers (1986).
Kuipers (1986) proved that QSIM is complete but not
always correct, that means, QSIM may generate some
false (spurious) behaviors. Some studies have been
done for eliminating such spurious behaviors (Fouche
and Kuipers, 1992; Say, 1998; Wang and Chen, 1995).
5. Structure of the system
With mathematical techniques, we describe dynamic
problems as dierential equations when the problems
are well structured and with precise quantitative
information. Solving the equations is trivial by giving
initial conditions, parameters and the like. Many real
problems, however, are often very complicated and ill-
structured, involving numerous factors, large uncertain-
ties, etc. For a complex problem (e.g., a blast furnace),
information is very likely to be highly qualitative, at
least initially and it virtually involves both qualitative
and quantitative aspects. Most of the time, precise
J. Chen / Engineering Applications of Articial Intelligence 14 (2001) 7785 80
quantitative information is not available or complete
information about the problem under study is not
available at the time of analysis. How do we know which
variable should be selected to model the problem faced?
Why do some ambient conditions omit when we use a
model? There are some problems for traditional
mathematical techniques when complex problems are
Reasoning with incomplete qualitative information
typically leads to ambiguous results. An alternative
approach to solve complex problems is to integrate
qualitative and quantitative methods to allow stronger
predictions than using qualitative reasoning or quanti-
tative methods alone (Wang and Chen, 1995). To
improve the predictive accuracy of silicon content
during the operating process of a blast furnace, an
integrated predictive system is developed by the synth-
esis of qualitative reasoning and a neural network. The
system attempts to help push back the bounds on the
quantitative models. Fig. 3 shows the conceptual
structure of the system.
The system includes the following main activities
Structuring problem: In this stage, the relationships
among factors are explored, according to the basic
principles of iron smelting process and experience
of expert operators. In many cases, however,
these relationships are qualitative with uncertainty,
e.g., If blast quantity increases smoothly, then
silicon content of pig iron may decrease smoothly.
Some variables are important to the smelting process,
but they can only be described qualitatively, e.g.,
speed of burden moving could be fast,
modest, slow, etc., according to the observation.
Based on the relationships between factors, a
conceptual model (e.g., causal network) is built, then
a qualitative model (equation-like) is obtained
Investigating conditions: With the qualitative
model, possible future behaviors of the process
are reasoned by QSIM, and certain hypotheses (i.e.,
the patterns of the operating process such as the
process is operating in stationary condition and
will run under stationary condition within 3 heats)
are given. Then, the relevant variables are selected
according to the hypotheses, that is, with dif-
ferent running pattern, dierent input variables
are selected. For example, the quantity of blast
(t), the temperature of blast V
(t) and the
quantity of coal powder V
(t) will be selected as
input variables when the process is running in
stationary condition. With qualitative analysis, the
interpretability of ANNs model may be improved to
certain extent.
Creating an ANNs model: Based on the results of
qualitative model, the inputs for an ANNs model are
determined in the last phase. With these variables, an
ANNs model is constructed with an MLP in three-
layer structure and the sigmoid activation function.
The number of neurons in hidden layer is determined
by following the empirical formula: s =

1, where s; n and m are the number of
hidden, input and output units (in our model, m = 1).
The blast furnace is a dynamic system. To adapt to
the change of the process, the sample set will be
updated when new data are available. The number of
samples in the sample set for training the model is 50
and the oldest sample of the sample set will be
eliminated when a new one is available and it will be
added to the sample set. Then, an improved BP
algorithm mentioned in session 3 is used to train the
model. When training error E (dened by Eq. (2b)) is
less than a given value e = 0:005, the training process
stops. The iteration number for training the model is
between 3500 and 5000 depending on how many
variables are selected as inputs.
Predicting silicon content of next heat: After the
model is trained, it is used to predict the silicon
content of next heat with proper input data.
Evaluating performance: The predictive value will be
compared with the observed one when it is available.
Consequently, according to the comparison, some
renements will be made for both qualitative model
and the ANNs model. For the qualitative model,
some landmarks or hypotheses may be adjusted if the
bias is signicant. And the ANNs model will be
retrained with the new data. With the feedback from
the comparison, the system will be adapted to the
change of the process. Fig. 3. Overview of system structure.
J. Chen / Engineering Applications of Articial Intelligence 14 (2001) 7785 81
With the integration of qualitative analysis and
quantitative methods, we could expect to make full use
of the information/knowledge (both qualitative and
quantitative) and furthermore to improve the predictive
6. Qualitative structure of a blast furnace
In order to reason the behaviors of complex systems
qualitatively, it is necessary to build a qualitative model
for the system. The causal approach is widely used to
model systems qualitatively. Through causality, we can
focus on any part or component of a system, without
having to deal with the model as a whole. This is useful
when trying to understand and explain the behaviors of
a system. The causal inuences (or causal explanations)
between variables are usually represented as an oriented
graph, where nodes in the graph represent variables or
physical parameters, arcs, oriented from an input or a
causal variable to an eect variable, are representative
of the relationship between variables. The silicon
content of the pig iron is inuenced greatly by the
operating situations, which are strongly related with
various conditions including in-furnace heat balance,
burden descending, gas permeability and the likes. In
actual operations, operators of a blast furnace obtain
information relating to the blast furnace conditions
from many sensors. By combining all these pieces of
information (absolute values, relative values, amount of
changes and so on), they judge the type of the blast
furnace conditions at any given time.
After making a thorough investigation and study on a
blast furnace, the qualitative structure of the operating
process in a blast furnace is worked out in the form of a
causal graph. Fig. 4 describes a part qualitative structure
of #9 blast furnace in Anshan Iron and Steel Company
(AISC) that is the largest Iron and Steel Company in
China. According to the causal graph, a qualitative model
is constructed with qualitative equation-like constraints.
The initial landmarks (important values) of each
variable are determined based on the specic parameters
of the blast furnace by discussing with experts.
7. Empirical results
To empirically implement the system described above,
610 heats from #9 blast furnace in AISC are checked.
According to the operating conditions of #9 blast
furnace, the following variables are considered to be
the input variables of the neural network for predicting
silicon content in pig iron:
(t): the quantity of blast
(t): the temperature of blast
(t): the pressure of blast
(t): the quantity of coal powder
(t): the index of ventilating performance
(t): the pressure of top gas
(t): the temperature of top gas
(t): the quantity of raw material
(t): the coke ratio
(t): the additional quantity of coke
Here, V
(t); V
(t); V
(t); V
(t); V
(t); V
(t) and V
are control variables and V
(t); V
(t) and V
(t) are
state variables. Generally, V
(t); V
(t) and V
(t) are
most frequently used to adjust the operating process,
(t) is constant for some weeks. However, some
additional quantity of coke V
(t) is fed some times if
necessary. In #9 blast furnace of AISC, the time
constant (T) of the iron-smelting process is about 6 h
and 10 heats are produced every day, that is, about
144 min on average for producing a heat. According to
engineering experience (Wang and Fang, 1992), the
sampling period t can be determined as t =
_ _
that is, the sampling period for #9 blast furnace of AISC
could be 3660 min. Considering the average time
interval between two heats is 144 min, we set the
sampling period t of 48 min. In this case, the interval
between two heats is exactly 3t. Our objective is to
predict the silicon content in pig iron of next heat (heat
t 1) based on the information of heat t, that means we
need predict 3 sampling steps ahead (k = 3), which
could help the operators to run the blast furnace with a
more ecient control strategy. With experiments and
analysis, the time delay indices of inputs are determined
as d
= 1; d
= 1; d
= 1; d
= 2; d
= 1; d
= 3; d
2; d
= 6; d
= 6; d
= 6. For example, under station-
ary condition, V
; V
and V
are selected as input
variables. Corresponding to y(t), the inputs are
(t 1); V
(t 1) and V
(t 1).
Once the qualitative states of the process in
blast furnace are derived through qualitative
analysis, the relevant input variables of the fore-
casting model can be determined, and the model will
be built automatically. With the model, silicon content
in pig iron will be predicted. Fig. 5 shows the
comparison between the predicted values and observed
Input and output data are divided into two sets:
the initial training set which includes 50 samples, and
the test set that includes 610 samples. The initial
training set is used for learning the initial parameters
(weights) of the ANN model. The test set is used to
assess the performance of the system. The model will be
retrained when a new sample from test set is available
and the oldest sample of the training set will be
eliminated at the same time to keep the size of training
set. To evaluate the performance of the prediction,
some important criteria used in practice are considered
J. Chen / Engineering Applications of Articial Intelligence 14 (2001) 7785 82
as follows:

_ _
1 where y(t) ^ y(t=t k) [ [50:1
0 else;
y(t) is the observed value at time t and ^ y(t=t k) the
predicted value k steps ahead at time t k

(y(t) ^ y(t=t k))

(y(t) y(t=t k))
y(t=t k)

Generally, we say that the prediction hits the target in
the case [y(t) ^ y(t=t k)[50:1 and the percentage of
hits (J
) is the most important criterion for operating a
blast furnace. Table 1 gives the comparison of dierent
methods for predicting silicon content in pig iron.
It is always believed that to predict the tendency of the
process (i.e. the direction of change of the process) is the
most important guide for taking correct actions.
Consider those heats where prediction does not hit the
target with the criterion J
, we also counter the
prediction hits the target if the predicting tendency is
consistent with real one, that is, the sign of
(^ y(t=t k) ^ y(t 1=t k 1)) is the same to the sign
of (y(t) y(t 1)). With this criterion, the percentages
of hits for IS and KAP will increase to 93.6 and 89.7%,
respectively. In practice, sometimes the criterion J
changed as follows to meet higher requirement of

_ _
1 where y(t) ^ y(t=t k) [ [50:075
0 else
The percentages of hits for IS and KAP are 81.5 and
68.1%, respectively. From the results of comparison
with other methods, we can nd that the performance of
Fig. 4. Illustration of a part qualitative structure for a blast furnace.
J. Chen / Engineering Applications of Articial Intelligence 14 (2001) 7785 83
Fig. 5. Comparison between the predicted values and observed values.
Table 1
Comparison of dierent methods
system (IS)
Combining knowledge-base
with an adaptive predictor
(KAP) (Chen, 1993)
Adaptive predictor
(Chen, 1993)
Experienced operator
(Chen, 1993)
(%) 88.7 81.3 68.9 62.0
0.0062 0.0082 0.0117 0.0214
(%) 14.3 16.3 19.3 24.9
For the detailed description of the method, please see Keyser and Canwenberghe (1981).
J. Chen / Engineering Applications of Articial Intelligence 14 (2001) 7785 84
the system presented in this paper is much better than
those of others.
8. Conclusion
In this paper, we present a predictive system for blast
furnaces to predict silicon content in pig iron. A blast
furnace is a very complex system, in which gas, liquid
and solids coexist. For a complex system, information is
very likely to be highly qualitative and it involves both
qualitative and quantitative aspects. Unlike traditional
methods, in this paper, we have designed a predictive
system by integrating a neural network with qualitative
analysis and expect to make use of both quantitative and
qualitative information. Six hundred and ten heats of
pig iron are used for verifying the performance of the
system. From the empirical results, it is evident that
excellent performance has been obtained, at least for the
test samples. It is very important that the predictive
system only makes a suggestion to help the operators to
make the right decision, that is, they can use it as a guide
or can ignore it and follow their own intuition. In this
sense, the more accurate the system performs, the more
helpful the system will be. We believe that to synthesize
qualitative reasoning with quantitative methods is a way
for treating complex problems. Further application of
this system to expand areas of operations is highly
expected. To improve the performance of the system
furthermore, we need to acquire more deep knowledge
about the process of blast furnaces, and explore ecient
methods for representing and processing qualitative
knowledge as well. On the other hand, eorts should
also be made toward a closed loop operation of blast
furnace heat control beyond the stage of guidance levels.
The author thanks the anonymous referees and the
editor for their valuable comments and suggestions. This
work is supported partly by the National Science
Foundation of China, Natural Science Foundation of
State Education Commission and National Defense
Chen, J., 1993. Predicting system based on combining an adaptive
predictor and a knowledge base as applied to a blast furnace.
Journal of Forecasting 12 (2), 93102.
Chen, J., Xu, D., 1998. An improved back propagation algorithm.
In: Gu, J. (Ed.), Proceedings of Systems Science and Systems
Engineering. Scientic and Technical Documents Publishing House,
Beijing, pp. 408411.
De Kleer, J., Brown, J.S., 1984. A qualitative physics based on
conuence. Articial Intelligence 24, 783.
Ergezinger, S., Thomsen, E., 1995. An accelerated learning algorithm
for multilayer perceptrons: optimization layer by layer. IEEE
Transactions on Neural Networks 6, 3141.
Fielden, C.J., Wood, B.I., 1968. A dynamic digital simulation of a
blast furnace. Journal of Iron and Steel Inst. 206, 650658.
Forbus, K.D., 1984. Qualitative process theory. Articial Intelligence
24, 85168.
Fouche, P., Kuipers, B.J., 1992. Reasoning about energy in qualitative
simulation. IEEE Transactions on Systems, Man and Cybernetics 22
(1), 4763.
Iida, O. et al., 1989. Blast furnace control by articial intelligence.
In: Rodd, M.G. Suski, G.J. (Eds.), Articial Intelligence in Real-
Time Control. Swansea, UK, pp. 7378.
Jin, G., 1986. A mathematic model for predicting the Si-content and
temperature of pig iron. Iron Smelting 5 (2), 816.
Keyser, R., Van Canwenberghe, A., 1981. A self-tuning multistep
predictor application. Automatica 17, 167174.
Kuipers, B., 1986. Qualitative simulation. Articial Intelligence 29,
Li, T., et al., 1984. Self-learning algorithm of fuzzy semantic inference.
Acta Automatica Sinica 10, 329336.
Liu, Y., 1985. The inuence of the relation between the temperature
and Si-content of pig iron by technique and melting condition. Iron
Smelting 4 (4), 5557.
Nakajima, R., 1987. Operation control system of blast furnace by
articial intelligence. AIME 46th Iron-Making Conference Proceed-
ings, pp. 155157.
Rumelhart, D.E., et al., 1986. Learning internal representations
by error propagation. In: Rumelhart, D.E., McClelland, J.L.
(Eds.), Parallel Distributed Processing: Explorations in the Micro-
structures of Cognition, Foundations, Vol. I. MIT Press, Cam-
bridge, MA.
Say, A.C.C., 1998. LHopitals lter for QSIM. IEEE Transactions on
Pattern Analysis and Machine Intelligence 20 (1), 18.
Unbehauen, H., Diekman, K., 1982. Application of MIMO identica-
tion to a blast furnace. Sixth IFAC Symposium on Identication
and System Parameter Estimation, pp. 235240.
Wang, Y., Chen, J., 1995. A framework of decision analysis integrating
qualitative reasoning and quantitative models. Proceedings of 1995
IEEE International Conference on Systems, Man and Cybernetics,
Vancouver, pp. 346450.
Wang, J., Fang, C., 1992. In: Computer Systems for Process Control.
Tsinghua University Press, Beijing, p. 37.
Jian Chen received the B.Sc. degree (1983) in Electrical Engineering,
M.Sc. degree (1986), and the Ph.D. degree (1989) both in Systems
Engineering from Tsinghua University, China. He is now a Professor
and Chairman of Management Science Department, Tsinghua
University. He serves as a member of the Administrative Committee
of IEEE systems, man and cybernetics society, a member of the
Standing Committee of Systems Engineering Society of China. He has
over 80 technical publications and has been a principal investigator for
about 20 grants or research contracts. His main research interests
include modeling and control for complex systems, forecast and
optimization techniques.
J. Chen / Engineering Applications of Articial Intelligence 14 (2001) 7785 85