Data processing: artificial intelligence – Neural network – Learning task
Reexamination Certificate
2000-07-03
2003-05-06
Follansbee, John (Department: 2121)
Data processing: artificial intelligence
Neural network
Learning task
C700S041000, C700S042000, C700S043000
Reexamination Certificate
active
06560585
ABSTRACT:
CROSS-REFERENCE TO RELATED APPLICATION
This document claims priority under 35 U.S.C. § 119 to French Patent Application No. 99 08 417 filed on Jul. 1, 1999, the entire contents of which are hereby incorporated herein by reference.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention pertains to a procedure and equipment for carrying out a complex process which generally have a level of numerical command/control based on the so-called CIM (Computer Integrated Manufacturing) model.
2. Discussion of the Background
Such a procedure and equipment integrate traditional techniques which include: (1) the acquisition and the designation of numerical and symbolic data; (2) numerical computations; (3) signal processing and recognition of shapes; (4) artificial intelligence, especially of techniques for representation of indefinite, spatial, and temporal knowledge as rules of logic of first order predicates, as objects, and as reports, and associated reasoning techniques; (5) automatic control of continuous and/or discrete systems, in time and space; and (6) system control algorithms in real time.
In order to describe the state of a process and the evolution of the state of the process one calls on the following traditional definitions (1)-(12).
(1) A process is a system of transformation of an incoming flow and an outgoing flow of material, energy or information; in a system such as a blast furnace of a steel plant or a cement producing revolving oven, the process transforms the matter or energy.
(2) Such a system of transformation proceeds according to a large set of phenomena that are related to one another according to a common goal, which corresponds to the goals of production in an imperfect environment.
(3) The goal of production of the process can be expressed in terms of adherence to constraints which affect certain incoming flows and certain outgoing flows. These constraints can in turn be expressed in terms of specific aims or ranges of values. The aims pertaining to the incoming flows pertain for example to the position of the actuators of the process that the operator responsible for process behavior must respect, as, for example, in the case of a blast furnace, a minimum and maximum proportion of coal consumption, a minimum and maximum flow rate of oxygen, while the goals pertaining to the outgoing flows express some constraints pertaining to the output that the conduct operator must satisfy, which include for example a range of melt temperatures, a minimum daily flow rate of pig iron, a range of silicon content in the melt or the dross.
(4) The environment is called imperfect in the sense that it is defined in an imprecise, uncertain and incomplete way; this environment limits the production possibilities of the process.
(5) A model of behavior of a process is an organized set of knowledge, or a “body” of knowledge, which is used to predict the state of the system as a function of the value of recorded quantities of the system and, for example, values of parameter measurement of the model.
(6) A process is dynamic when the quantities which occur in its functioning model, like the state variables X, the input variables U, and the output variables Y, are related by temporal relationships.
(7) The behavior of one quantity can be defined by the relation between the value of its magnitude x and time t; it is then represented by x(t).
(8) According to the traditional rules defined by the entire set of the pairs of values (X′(t), X(t)) relative to each recorded quantity, where X′ designates the derivative with respect to time t of the quantity X; the process can then be described by a relationship of the type:
X′(t)=f[&thgr;,X(t), U(t)] and Y(t)=h[&thgr;, X(t), U(t)]
Where &thgr; designates some parameters, t the time variable here continues, f [ ] and h [ ] some functions that describe the process.
(9) A trajectory of state of a process for a time interval [↑
min
, ↑
max
] is defined by a sequence of points (X′(t), X(t)) in which the values of t are included in this range.
(10) A dynamic process is complex in one or the other of the following cases: absence of mathematical model of behavior or mathematical model of inoperative behavior; absence of a physical model of behavior, due to the inadequacy of the scientific knowledge, for example, or inoperative behavior physical model which does not yield any exploitable numerical calculation algorithm, for example: a non-reversible model, a model that cannot be calculated, a model that results in prohibitive calculation times with respect to the required response time, and a chaotic model.
In the case of complex dynamic processes one therefore will generally use symbolic models that resemble the entire set of the knowledge bits and the expertise obtained from observation of the behavior of the process. In a traditional manner one can construct such a model of behavior from knowledge possessed by experts during conduct of the process in question, by employing: logical formalisms that allow one to represent this knowledge; methodological tools that allow the acquisition of the knowledge for the purpose of their representation; and techniques for solving problems that have been posed to automatically find solutions to problems expressed according to these formalisms.
Among the formalisms employed one can mention: (1) the representation of the descriptive knowledge in the form of objects, classes, and meta-classes; (2) the representation of deductive knowledge by a logic of the first order predicates; (3) the representation of temporal knowledge in the form a reified temporal logic; and (4) the representation of spatial and temporal knowledge in the form of charts of discrete events.
Among the problem solving techniques we can mention: (1) reasoning through the memory of properties and behaviors; (2) reasoning based on the logic of the first order predicates, of the so-called “Modus Ponens” type for example; (3) control of reasoning directed by the events, by the compilation of rules with trees of binary events for example; (4) management of time constraints; (5) signal processing by filtering, for example, by time and/or space segmentation, and by parametric identification; and (6) shape recognition, by multiple linear regression for example.
Among the methodological tools which allow the automatic exploitation of the knowledge we can mention system design methodologies with a knowledge base, such as the software product called “Openkads™” or methodologies that work out the generic knowledge bases such as KADS.
The dynamic character of the behavior model is obtained by the use of formalisms that integrate the temporal constraints, by including concepts of reports and events for example to the logic of the first order predicates. One known example of formalism of this kind is called “DEVS” (Discrete Event System Specification), which allows one to define what is called discrete events from the inputs U, from the state X, from the outputs Y of internal transition functions X→X, of the external transition functions U×X→X, of the output functions X×U→Y and of life duration functions of a state.
FIG. 1
shows such an abstraction with the discrete events E
1
to E
4
of an inputs U-outputs Y relationship.
In summary, the techniques for solving problems supply the technological and methodological tools that allow one to automate the exploitation of the knowledge that is used by experts to carry out a complex process and to translate these bits of knowledge into a behavioral model of the complex dynamic process.
As an example of a complex dynamic process we can mention: a blast furnace; an electric steel production oven; a glass production oven; a cement production oven; and, a rolled strip unit.
SUMMARY OF THE INVENTION
The present invention has a goal of providing a procedure and equipment capable of using this model of a complex dynamic process for the purpose of guiding the conduct of the process in conf
Barles Michel
Dolenc Norbert
Le Goc Marc
Lesaffre François-Marie
Thirion Claude
Follansbee John
Holmes Michael B.
Oblon & Spivak, McClelland, Maier & Neustadt P.C.
Sollac
LandOfFree
Procedure for controlling a complex dynamic process does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Procedure for controlling a complex dynamic process, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Procedure for controlling a complex dynamic process will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3092924