Data processing: generic control systems or specific application – Generic control system – apparatus or process – Optimization or adaptive control
Reexamination Certificate
1998-04-23
2001-04-03
Sheikh, Ayaz R. (Department: 2781)
Data processing: generic control systems or specific application
Generic control system, apparatus or process
Optimization or adaptive control
C700S028000, C700S029000, C700S030000, C700S031000, C700S047000
Reexamination Certificate
active
06212438
ABSTRACT:
FIELD OF THE INVENTION
The invention relates to a method and to a system for generating a model of an industrial production and to control such a production with the help of the model. The industrial production relates, for example, to producing, on a mass production scale, paneling material such as chipboards.
BACKGROUND INFORMATION
Process models serve for approximating, analyzing, and optimizing of industrial processes. Such processes of process steps may be performed by an entire production line, by a few system components cooperating as an aggregate, or even by individual system components of a production line. The basic function of such process models is to provide on the basis of input parameters, output values that can be expected or predicted on the basis of the input parameters. The output values may be used in a closed loop or positive feedback control for influencing or fully controlling the industrial process or production.
Conventional process models are based on a linear formulation and are usually analyzed by means of known statistical methods. However, by using conventional linear formulations, it is not possible to satisfactorily model complex processes having non-linear characteristics. Thus, for modeling non-linear processes it is customary to use non-linear models such as neural networks which are capable of mapping or displaying complex non-linear response characteristics.
In order to identify the parameters of such non-linear models it is necessary to use non-linear optimizing algorithms which require an extremely high computer investment expense and effort, particularly during the so-called learning phase. Another particular difficulty is encountered in the formation of the neural network structure, such as the selection of the number of the individual neural cells to be incorporated into the network and to select the internetting connections of these neural cells within the network.
For further background information reference is made to the following publications, the content of which is incorporated herein by reference.
(A) John Moody and Christian J. Darken,
“Fast Learning in Networks of Locally Tuned Processing Units”, published in: Neural Computation, Vol. 1, pages 281 to 294, published in 1989 by MIT, with regard to the “Radial Basis Functions Method” applying Gauss-functions;
(B) Mark J. L. Orr,
“Regularization in the Selection of Radial Basis Function Centres”, Neural Computation, Vol. 7, No. 3, pages 606-623, published in 1995 by MIT, with regard to the “Stepwise Regression Method”;
(C) S. Chen, C. F. N. Cowan, and P. M. Grant
“Orthogonal Least Squares Learning Algorithm for Radial Basis Function Networks”, IEEE Transactions on Neural Networks, Vol. 2, No. 2, pages 302 to 309; publication date: Mar. 2, 1991, with regard to the “Forward Selection Method”.
(D) G. Deco and D. Obradovic,
“An Information-Theoretic Approach to Neural Computing”, Publisher: Springer Verlag, 1996, with regard to the variation of the selection criterium as an estimate of the expected generalized error.
OBJECTS OF THE INVENTION
In view of the foregoing it is the aim of the invention to achieve the following objects singly or in combination:
to provide a method and system for the generating of a model of an industrial process or industrial production, whereby the model shall be capable of approximately a non-linear response characteristics of complex processes and systems with a relatively small investment expense and effort compared to conventional systems using exclusively neural networks for such a purpose;
to construct such a system that the process model is capable of optimally learning while having a simple, yet clear structure that can be accomplished with a minimum number of neural cells;
to utilize the information provided by the model for a control, preferably a closed loop control of the industrial process or production to optimize its performance; and
to advantageously utilize the capability of neural networks to map complex process characteristics in combination with well established statistical methods, especially linear methods in the same process model.
SUMMARY OF THE INVENTION
According to the invention there is provided a method for generating a model of an industrial production to provide control signals for optimizing said industrial production, comprising the following steps:
(a) gathering a number N of training data sets for said industrial production,
(b) connecting a first neural network (
5
) in parallel to a second linear network (
6
), said neural network being formed initially by a number of neural cells corresponding to said number N of training data sets, said linear network being formed by M linear paths and to a corresponding number M of linear inputs provided in common for said first and second networks (
5
,
6
), to provide a preliminary model for simultaneously performing linear combinations of input values in the first and second networks to thereby train and optimize said first and second networks together,
(c) processing by said neural cells with the application of radial basis functions an input vector into individual first activating values,
(d) applying by said neural cells first weighting factors to said individual first activating values to provide first weighted values,
(e) linearly combining said first weighted values to provide first combined values,
(f) supplying said first weighted values and said first combined values to an output summing circuit,
(g) applying further weighting factors to said input values in said linear second network to provide second weighted values,
(h) simultaneously linearly combining said further weighted values to provide second combined values,
(i) supplying weighted and combined second values to said output summing circuit,
(j) performing an R number of regression steps and terminating said regression steps in accordance with a stop criterium which determines an over-adaption when said initial number N of neural cells is reduced to a lower number K of neural cells in said neural network and when said number M of linear paths is reduced to a lower number M-R of linear paths in said linear network, o provide a reduced final model,
(k) ascertaining actual process or production parameters and supplying said actual process or production parameters to said reduced final model,
(l) determining by said reduced final model expected or rated quality characteristics of a product to be manufactured by said industrial production to provide respective rated quality output values,
(m) processing said rated quality output values with the help of an optimizing algorithm to provide production control values, and
(n) controlling said industrial production by said control values.
According to the invention, there is further provided an apparatus for generating a model of an industrial production, comprising a neural network (
5
) including a number K of neural cells (
7
), a linear network (
6
) including a number M of linear signal paths, conductors connecting said neural network (
5
) and said linear network (
6
) in parallel to each other for simultaneously training and optimizing said neural and linear networks, to form a parallel circuit for performing weighted linear combinations of maximally M input parameters, a number of (x
1
to x
nM
) of input terminals to which said parallel circuit is connected for receiving said M input parameters, and a summing point (
9
) connected in common to all said neural cells (
7
) and to all said linear signal paths (
6
A,
6
B, . . . ), wherein said M process parameters forming input values are correlated with a number L of production quality characteristics forming output values with the help of N training data sets, whereby K is smaller than or equal to N, to provide control signals for said industrial production.
The combination of a linear network made up of linear paths such as linear conductor paths and a neural network made up of neural cells, connected in parallel to each other as taught by the invention uses the ability of neural networks to model very complex pro
Fasse W. F.
Fasse W. G.
Jean Frantz Blanchard
Schenk Panel Production Systems GmbH
Sheikh Ayaz R.
LandOfFree
Method and apparatus for generating a model of an industrial... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for generating a model of an industrial..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for generating a model of an industrial... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2526805