The paper discusses several methods of modelling complex nonlinear dynamics using neural networks. Particular reference is made to the problem of modelling direction-dependent relationships. A typical example of this would be top product composition control in a distillation column, where it is easier (i.e. faster) to make the product less pure than it is to make it more pure by an equivalent amount. Recurrent neural networks are identified as a potential method of modelling this type of relationship. The particular architecture chosen for this example is referred to as ‘semirecurrent’, since only past values of the predictions of the network are fed back to the input layer. This architecture is successfully used to model direction-dependent relationships in both simulated and actual industrial process data.
References
-
-
1)
-
C. Peel ,
M.J. Willis ,
M.T. Tham
.
A fast procedure for the training of neural networks.
J. Process Control
,
4 ,
205 -
211
-
2)
-
K.J. Hunt ,
D. Sbarbaro
.
Neural networks for non-linear internal model control.
IEE Proc. D
,
5
-
3)
-
J. Thibault ,
V.V. Breusegem ,
A. Cheruy
.
On-line prediction of fermentation variables using neural networks.
Biotechnol. Bioeng.
,
1041 -
1048
-
4)
-
D.E. Rumelhart ,
G.E. Hinton ,
R.J. Williams
.
Learning representations by back-propagation errors.
Nature
,
533 -
536
-
5)
-
M.J. Willis ,
C. Di Massimo ,
G.A. Montague ,
M.T. Tham ,
A.J. Morris
.
Artificial neural networks in process engineering.
IEE Proc. D
,
3 ,
256 -
266
-
6)
-
R.G. Brown
.
(1963)
Smoothing, forecasting and prediction of discrete time series.
-
7)
-
Sorsa, T., Koivo, H.N.: `Application of neural networks in the detection of breaks in a papermachine', IFAC symposium on On-line fault detection and supervisionin the process industries, 22–24 April 1992, Newark, Delaware, p. 162–167.
-
8)
-
Koshijima, I., Niida, K.: `Neural network approach to fault detection under steady state operation', IFAC symposium on On-line fault detection and supervisionin the process industries, 22–24 April 1992, Newark, Delaware, p. 174–179.
-
9)
-
R.D. Tyagi ,
Y.G. Du ,
T.R. Sreekrishnan ,
J.P. Villeneuve
.
Neural model for operational control of activated sludge processes.
Process. Biochemistry
,
4 ,
259 -
267
-
10)
-
S.-Z. Qin ,
H.-T. Su ,
T.J. McAvoy
.
Comparison of four neural net learning methods for dynamic system identification.
IEEE Trans. on Neural Networks
,
1
-
11)
-
G. Cybenko
.
Approximation by superpositions of sigmoidal function.
Math. Control Signal Syst.
,
303 -
314
-
12)
-
S.A. Billings ,
H.B. Jamaluddin ,
S. Chen
.
Properties of neural networks with applications to modelling non-lineardynamical systems.
Int. J. Control
,
1 ,
193 -
224
-
13)
-
C.A. Terzuolo ,
T.A. McKeen ,
R.E. Poppele ,
N.P. Rosenthal ,
C.A. Terzuolo
.
(1969)
Impulse trains, coding and decoding, Systems analysis to neurophysiological problems.
-
14)
-
A.V. Holden
.
(1976)
Models of the stochastic activity of neurons.
-
15)
-
Glassey, J., Montague, G.A., Willis, M.J., Morris, A.J.: `Considerations in process applications of artificial neural networks', Colloquium on Neural networks and fuzzy logic in measurementand control, March 1993, John Moores UniversityLiverpool.
-
16)
-
R.J. Williams ,
D. Zipser
.
A learning algorithm for continually running, fully recurrent neuralnetworks.
Neural Computat.
,
1270 -
1289
http://iet.metastore.ingenta.com/content/journals/10.1049/ip-cta_19960061
Related content
content/journals/10.1049/ip-cta_19960061
pub_keyword,iet_inspecKeyword,pub_concept
6
6