WO2002073323A2 - Schätzverfahren für eine grösse eines prozesses der grundstoffindustrie unter verwendung einer stützvektormethode - Google Patents
Schätzverfahren für eine grösse eines prozesses der grundstoffindustrie unter verwendung einer stützvektormethode Download PDFInfo
- Publication number
- WO2002073323A2 WO2002073323A2 PCT/DE2002/000899 DE0200899W WO02073323A2 WO 2002073323 A2 WO2002073323 A2 WO 2002073323A2 DE 0200899 W DE0200899 W DE 0200899W WO 02073323 A2 WO02073323 A2 WO 02073323A2
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- model
- vector
- values
- estimation method
- treasure
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B17/00—Systems involving the use of models or simulators of said systems
- G05B17/02—Systems involving the use of models or simulators of said systems electric
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
- B21B—ROLLING OF METAL
- B21B37/00—Control devices or methods specially adapted for metal-rolling mills or the work produced thereby
Definitions
- the present invention relates to an estimation method for a process variable of a process in the basic material industry
- each pair of values has a vector of detectable signals of the process and a measured value corresponding to this vector for the process variable to be estimated
- an empirical model for the dependency of the process variable on the detectable signals is created on the basis of the number of pairs of values, an estimated value for the process variable being determined on the basis of the model during the course of the process using at least one further vector of detectable signals,
- the model is given at least one further pair of values, the vector of which was determined during the process and the measured value of which was determined after the vector was ascertained,
- the model created is adapted on the basis of the at least one further pair of values on a process automation computer.
- Such estimation methods are e.g. based on neural networks, generally known. For example, reference is made to DE 197 27 821 AI.
- the estimation methods previously used in the raw materials industry have problems.
- the training effort for the model often unfavorably scales with the dimension of the input vectors. In the worst case, the training effort increases exponentially.
- the training process is often based on a non-linear optimization process. This can cause that the global optimum may not be found and / or the computing time requirement may be too great.
- the object of the present invention is to provide an estimation method for an empirical model that can be trained reliably and with an acceptable computational effort even with a higher-dimensional input variable vector.
- the task is solved in that the model is a support vector model.
- Training methods for the model are "chunking methods”, e.g. those according to Joachims or the “Sequential Minimal Optimization” according to Platt, because these methods work reliably and quickly.
- a support vector model is (with the exception of an offset) a linear combination of symmetric core functions that are Let mathematically write as a dot product of two identical functions.
- Support vector methods and support vector models are e.g. B. in the monographs
- Regression methods are generally used to predict physical process variables within the process automation (process control and regulation) of a plant, while classification or pattern recognition processes are used, for example, to detect faults in the plant, If the model can be partially derived analytically from the detectable signals, the online computational effort is minimal.
- the symmetric core functions can e.g. B. Gaussian radial basis functions.
- adjusting the model can be done during times when the computing power of the computer used to adjust is not needed otherwise.
- the process can be of any nature.
- it can be a coal and ore preparation process, a smelting process, a steel production process, a casting process, a rolling process and a pulp and paper production or processing process.
- 1 shows a plant of the raw materials industry and 2 shows a flow chart.
- a process variable of process 1 is to be estimated for a process 1 in the basic material industry.
- Process 1 can be any process 1 of the basic be industry. 1 shows a smelting process la, a steelmaking process 1b, a casting process 1c and a rolling process Id. However, other processes are also possible, in particular a processing process for coal or ore and a production or processing process for Pulp or paper.
- a certain type of support vector model 2 is first selected together with a training method.
- the training method is online-capable in terms of computing time and reliability, i.e. on a process automation computer 2 'for process 1. Because the process automation computer 2 'of the system preferably implements the treasure method.
- support vector model 2 e.g. - but not necessarily - the support vector model according to Vapnik described in US Pat. No. 5,950,146 can be used.
- a number N of pairs of values is collected from the plant on which process 1 is running (or a similar plant).
- This number N of pairs of values is used for offline training of a support vector model 2 of the selected type.
- An empirical support vector model 2 for the dependence of the process size on the detectable signals is therefore created on the basis of the number N of value pairs.
- the support vector model 2 pre-trained in this way is then implemented together with the selected online training method on the process automation computer 2 ′ of the system.
- a vector of detectable signals x x , ..., x n of process 1 is supplied to the support vector model 2 as an input size vector.
- the Stutzvek- Door model 2 uses the supplied vector to determine a treasure value y * for the process size during the course of process 1.
- the actual value y of the process size can often also be determined using process 1 (but always only afterwards).
- This value y is fed to the support vector model 2 together with the input size vector for online training.
- the support vector model 2 Based on the pair of values formed from the input variable vector and the process variable, i.e. on the one hand the vector of detectable signals xi, ..., x n and on the other hand the corresponding measured value y for the process variable, the support vector model 2 then goes online (ie on the process automation computer 2 ') retrained or adjusted.
- the creation and adaptation of the model 2 is preferably carried out in accordance with a “chunking method” as used for. B. is described by Joachims or how it was developed as a so-called “Sequential Minimal Optimization” ("Sequential Minimal Optimization") by Platt.
- the model 2 can be adapted immediately after the new value pair has been determined. As a rule, however, it will make more sense to temporarily store the additional pair of values - possibly several additional pairs of values - and to adapt the model 2 at a later point in time with the previously accumulated pairs of values.
- the model 2 is given a number N of value pairs in a step 3.
- Each pair of values consists of a vector of detectable signals xj 1 , ..., X n 1 and a measured value y 1 corresponding to this vector for the Process size to be estimated. Based on these value pairs, model 2 is then trained in a step 4.
- the vector vector 2 is a vector of detectable signals x lr . , , , x n entered.
- it determines an estimated value y * for the process variable.
- the tat neuter ⁇ y value of the process variable determined.
- the vector of detectable signals xi, ..., x n is temporarily stored together with the value y in a step 8. Then in one
- Step 9 asked whether a re-training of the model 2 should take place. Depending on the result of the query, the system either jumps back to step 5 or does a follow-up training of model 2 in step 10.
- Support vector models for modeling continuous process variables are e.g. B. from the shape
- the entire support vector model 2 is a linear combination of these functions k.
- W j and b are coefficients to be determined by the training process.
Abstract
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AT02729789T ATE282221T1 (de) | 2001-03-14 | 2002-03-13 | Schätzverfahren für eine grösse eines prozesses der grundstoffindustrie unter verwendung einer stützvektormethode |
DE50201518T DE50201518D1 (de) | 2001-03-14 | 2002-03-13 | Schätzverfahren für eine grösse eines prozesses der grundstoffindustrie unter verwendung einer stützvektormethode |
EP02729789A EP1368714B1 (de) | 2001-03-14 | 2002-03-13 | Schätzverfahren für eine grösse eines prozesses der grundstoffindustrie unter verwendung einer stützvektormethode |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE10112267.5 | 2001-03-14 | ||
DE10112267A DE10112267A1 (de) | 2001-03-14 | 2001-03-14 | Schätzverfahren für eine Größe eines Prozesses der Grundstoffindustrie unter Verwendung einer Stützvektormethode |
Publications (2)
Publication Number | Publication Date |
---|---|
WO2002073323A2 true WO2002073323A2 (de) | 2002-09-19 |
WO2002073323A3 WO2002073323A3 (de) | 2002-10-31 |
Family
ID=7677437
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/DE2002/000899 WO2002073323A2 (de) | 2001-03-14 | 2002-03-13 | Schätzverfahren für eine grösse eines prozesses der grundstoffindustrie unter verwendung einer stützvektormethode |
Country Status (4)
Country | Link |
---|---|
EP (1) | EP1368714B1 (de) |
AT (1) | ATE282221T1 (de) |
DE (2) | DE10112267A1 (de) |
WO (1) | WO2002073323A2 (de) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE10327663A1 (de) * | 2003-06-20 | 2005-01-05 | Abb Patent Gmbh | System und Verfahren zur optimierenden Regelung der Dickenqualität in einem Walzprozess |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE19727821A1 (de) * | 1997-06-30 | 1999-01-21 | Siemens Ag | Verfahren und Einrichtung zur Steuerung und Voreinstellung eines Walzgerüstes bzw. einer Walzstraße |
US5950146A (en) * | 1996-10-04 | 1999-09-07 | At & T Corp. | Support vector method for function estimation |
US5991525A (en) * | 1997-08-22 | 1999-11-23 | Voyan Technology | Method for real-time nonlinear system state estimation and control |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US619360A (en) * | 1899-02-14 | Insulating alternating-current circuits | ||
DE4130164A1 (de) * | 1991-09-11 | 1993-03-18 | Bodenseewerk Geraetetech | Regler, insbesondere flugregler |
DE4338608B4 (de) * | 1993-11-11 | 2005-10-06 | Siemens Ag | Verfahren und Vorrichtung zur Führung eines Prozesses in einem geregelten System |
DE29617200U1 (de) * | 1996-10-02 | 1997-11-06 | Siemens Ag | Vorrichtung zur prädiktiven Diagnose der aktuellen Güte des technischen Arbeitsergebnisses einer technischen Anlage, insbesondere der aktuellen Güte der Schweißpunkte eines Punktschweißroboters |
DE19641432C2 (de) * | 1996-10-08 | 2000-01-05 | Siemens Ag | Verfahren und Einrichtung zur Vorausberechnung von vorab unbekannten Parametern eines industriellen Prozesses |
US6112195A (en) * | 1997-03-27 | 2000-08-29 | Lucent Technologies Inc. | Eliminating invariances by preprocessing for kernel-based methods |
DE19855873A1 (de) * | 1998-12-03 | 2000-06-15 | Siemens Ag | Verfahren und Anordnung zum Entwurf eines technischen Systems |
-
2001
- 2001-03-14 DE DE10112267A patent/DE10112267A1/de not_active Ceased
-
2002
- 2002-03-13 EP EP02729789A patent/EP1368714B1/de not_active Expired - Lifetime
- 2002-03-13 WO PCT/DE2002/000899 patent/WO2002073323A2/de not_active Application Discontinuation
- 2002-03-13 DE DE50201518T patent/DE50201518D1/de not_active Expired - Fee Related
- 2002-03-13 AT AT02729789T patent/ATE282221T1/de not_active IP Right Cessation
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5950146A (en) * | 1996-10-04 | 1999-09-07 | At & T Corp. | Support vector method for function estimation |
DE19727821A1 (de) * | 1997-06-30 | 1999-01-21 | Siemens Ag | Verfahren und Einrichtung zur Steuerung und Voreinstellung eines Walzgerüstes bzw. einer Walzstraße |
US5991525A (en) * | 1997-08-22 | 1999-11-23 | Voyan Technology | Method for real-time nonlinear system state estimation and control |
Also Published As
Publication number | Publication date |
---|---|
EP1368714A2 (de) | 2003-12-10 |
DE50201518D1 (de) | 2004-12-16 |
EP1368714B1 (de) | 2004-11-10 |
ATE282221T1 (de) | 2004-11-15 |
WO2002073323A3 (de) | 2002-10-31 |
DE10112267A1 (de) | 2002-11-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
DE102016011532B4 (de) | Maschinenlernvorrichtung und Maschinenlernverfahren zum Optimieren der Häufigkeit einer Werkzeugkorrektur einer Werkzeugmaschine und Werkzeugmaschine mit der Maschinenlernvorrichtung | |
EP2106576B1 (de) | Verfahren zur rechnergestützten steuerung und/oder regelung eines technischen systems | |
DE4338615B4 (de) | Verfahren und Vorrichtung zur Führung eines Prozesses in einem geregelten System | |
DE112019001512T5 (de) | EINSPRITZGIEßMASCHINENSYSTEM | |
DE4338608B4 (de) | Verfahren und Vorrichtung zur Führung eines Prozesses in einem geregelten System | |
DE4008510A1 (de) | Regeleinheit mit optimal-entscheidungsmitteln | |
DE4301130A1 (en) | Control system with highly nonlinear characteristic - uses learning function and derives control values from selected parameter values of set point adjustment model | |
EP2112568A2 (de) | Verfahren zur rechnergestützten Steuerung und/oder Regelung eines technischen Systems | |
DE4338607B4 (de) | Verfahren und Vorrichtung zur Führung eines Prozesses in einem geregelten System | |
WO2007124981A1 (de) | Verfahren zum betrieb eines mühlensystems | |
DE19781103B4 (de) | Verfahren und Einrichtung zur Identifikation bzw. Vorausberechnung von Prozeßparametern eines industriellen zeitvarianten Prozesses | |
EP0756219A2 (de) | Verfahren zur Überwachung von Produkteigenschaften und Verfahren zur Regelung eines Herstellungs-prozesses | |
EP1055180B1 (de) | Verfahren und anordnung zum entwurf eines technischen systems | |
DE112019007889T5 (de) | Bearbeitungsprogramm-umwandlungseinrichtung, numerische-steuereinrichtung und bearbeitungsprogramm-umwandlungsverfahren | |
DE19731980A1 (de) | Verfahren zur Steuerung und Voreinstellung eines Walzgerüstes oder einer Walzstraße zum Walzen eines Walzbandes | |
EP1368714B1 (de) | Schätzverfahren für eine grösse eines prozesses der grundstoffindustrie unter verwendung einer stützvektormethode | |
EP0416368A2 (de) | Adaptiver Regler | |
DE19641432C2 (de) | Verfahren und Einrichtung zur Vorausberechnung von vorab unbekannten Parametern eines industriellen Prozesses | |
EP1157317B1 (de) | Verfahren und anordnung zur reduktion einer anzahl von messwerten eines technischen systems | |
DE102020213527A1 (de) | Verfahren zum Optimieren einer Strategie für einen Roboter | |
DE102019208922A1 (de) | Verfahren und Vorrichtung zum Kontrollieren eines Produktionsprozesses | |
EP1157311B1 (de) | Verfahren und anordnung zum entwurf eines technischen systems | |
EP0735443B1 (de) | Neuronales Netz zur Positionsabweichungskompensation bei Werkstückbearbeitungseinrichtungen | |
EP1071999B1 (de) | Verfahren und anordnung zur rechnergestützten ermittlung einer abbildungsvorschrift | |
DE102018217310A1 (de) | Verfahren zum Auswerten von Zeitseriensignalen |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A2 Designated state(s): CN JP NO US |
|
AL | Designated countries for regional patents |
Kind code of ref document: A2 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR |
|
AK | Designated states |
Kind code of ref document: A3 Designated state(s): CN JP NO US |
|
AL | Designated countries for regional patents |
Kind code of ref document: A3 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
WWE | Wipo information: entry into national phase |
Ref document number: 2002729789 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 2002729789 Country of ref document: EP |
|
WWG | Wipo information: grant in national office |
Ref document number: 2002729789 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: JP |
|
WWW | Wipo information: withdrawn in national office |
Country of ref document: JP |