EP3449433A1 - Modellbasierte ermittlung eines systemzustandes mittels eines dynamischen systems - Google Patents
Modellbasierte ermittlung eines systemzustandes mittels eines dynamischen systemsInfo
- Publication number
- EP3449433A1 EP3449433A1 EP17728082.3A EP17728082A EP3449433A1 EP 3449433 A1 EP3449433 A1 EP 3449433A1 EP 17728082 A EP17728082 A EP 17728082A EP 3449433 A1 EP3449433 A1 EP 3449433A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- vector
- time
- correction
- time series
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 239000013598 vector Substances 0.000 claims abstract description 99
- 238000012937 correction Methods 0.000 claims abstract description 37
- 238000000034 method Methods 0.000 claims abstract description 33
- 238000013528 artificial neural network Methods 0.000 claims abstract description 19
- 230000007704 transition Effects 0.000 claims abstract description 14
- 230000000306 recurrent effect Effects 0.000 claims abstract description 13
- 238000004590 computer program Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000012886 linear function Methods 0.000 claims description 2
- 238000012502 risk assessment Methods 0.000 claims 1
- 238000005259 measurement Methods 0.000 abstract description 24
- 230000006870 function Effects 0.000 description 8
- 239000011159 matrix material Substances 0.000 description 8
- 238000004393 prognosis Methods 0.000 description 4
- 238000009795 derivation Methods 0.000 description 3
- 230000002123 temporal effect Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 241001123248 Arma Species 0.000 description 1
- 241000282941 Rangifer tarandus Species 0.000 description 1
- 241000728173 Sarima Species 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 239000000853 adhesive Substances 0.000 description 1
- 230000001070 adhesive effect Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 230000036962 time dependent Effects 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
Definitions
- Physical systems with temporal dynamics can be described with ma ⁇ thematic models. Observables can be observed in physical systems. They can be vectors that are a multitude of individual ones
- Observablen to describe a system state. Can observables for certain time points from system states or system state vector to be modeled by means of a model and therefore also describe a physical Sys tem ⁇ at this time. The time may be in the future, so that observables can be forecasted.
- time series analysis or to predict observable time series mathematical models are used which are derived on the basis of historical data a Sys ⁇ tem underlying functional, linear or non-linear input-output relationships. For example, regression models or so-called Autoregressive Moving Average models or short ARMA models, or Autoregressive integrated Moving Average or ARIMA models or Seasonal ARIMA, short SARIMA models or neural networks are used.
- the time series of the input and target variables usually come from sensor measurements.
- the data is processed as a time series. That is, vectors of quantities describing the system at different, successive times must be complete in order for the entire time series to be processed. Removing single missing timestamp is not possible without the model performance negatively beeinflus ⁇ sen.
- a system state is modeled that is corrupted by gaps in the time series. It is known to artificially fill in gaps in the time series using upstream algorithms and then to use the data for modeling. In a particularly simple case is held value with short gaps in the last known measurement until a new valid measurement exists ⁇ .
- the invention relates to a method for the model-based determination of a system state of a dynamic system with ⁇ means of a model, in which a recurrent neural network is provided as a model of the dynamic system, wherein the model of a time series of potentially detectable measured values, comprising detected and missing measurement values, is provided as an input,
- Time-dependent systems can be described particularly advantageously by recurrent neural networks. Since the inputs are not available for predicting future time steps for Availability checked ⁇ supply, the input and output variables are vAbLe by so-called observables, ie observable quantities replaced, in so-called Historically Consistent Neural Networks or historically consistent neural networks. Observable can to the present detected by measurements ⁇ map for points in time and can be derived as the target values for arbitrary points in time of a so-called system state or State Vector. A system state is based on observables as ⁇ be observable quantities and additionally hidden state variables or the so-called Hidden States. The modeling can be refined by replacing, for observables of system states in the past, the target values expected on the basis of the serviceable component in the system state by actually acquired observables. It is a procedure based on a so-called
- the application of a further developed ATF within the meaning of the subject-matter of the independent claim means that when a status vector changes at a first time t-1 to a state vector of a subsequent time t a correction is made which takes account of the fact that a 1 modeled value possible is not confirmed by a measurement made at time t-1, but there is a deviation between prognosis and real measurement. This deviation of the actually observed value from the predicted target value or model-herten target value is considered to determine the state vector describing the system at time t.
- the prognosis or derivation of a model-based target value is improved.
- the prediction of the target variable vector associated with the time t which can be derived from the associated state vector, is improved.
- the target vector comprises a plurality of target values that represent ⁇ be observable sizes.
- the correction is made if a value within the observable vector has actually been measured or is present. On the other hand, if there is a gap in the time series, for example for some or more entries of one or more observable vectors at different times, these are not taken into account in the correction.
- the correction is carried out only in the case in which a value within the Observablenvektors also did was measured ⁇ neuter or present. In particular, no value determined by an algorithm based on the time series is used for the correction which has not actually been measured on the system. In time steps with missing
- Measured values of the state vector remains unchanged, that is, the target values of the observables for a respective subsequent time step suction ⁇ derived from the internal state of the model.
- a correction thus remains for entries or observables vablen from a system state or is not made, if there is for the corresponding observable in the previous To ⁇ was not reading.
- a purely internal derivation of a modeled target value from the state of the model takes place without consideration of the historical measurement data.
- the correction is thus carried out unadulterated by missing measured values.
- the underlying dynamic processes are not distorted by the fact that Internet or Extrapo ⁇ lationshabilit to complement missing values are applied in a time series.
- the proposed method for extending a historically consistent neural network model makes it possible to completely dispense with the preprocessing of measurement data and to have gaps in the data of the time series supplemented by the model itself. So too sparse time series processed consistently and thus also consistent modeling of the entire zugrundelie ⁇ constricting dynamics of the system are achieved.
- no expert knowledge is needed to manually select a suitable interpolation method.
- a suitable model should be chosen conveniently so ⁇ that automation of the correction method with ver ⁇ reasonable expense was not possible. In case of a faulty or inappropriate estimation of a
- the correction is carried out uninfluenced by the missing measured values of the time series by multiplying an observable vector present at each time of the time series by a vector and this vector is constructed from O entries such that the missing ones Measured values in the observable vector due to the O entries are not included in the correction of the system state.
- An index vector is therefore provided which excludes by O entries when multiplying those values in Observablenvektor for the correction or neglected, which have not been measured to herein are subject ⁇ membered time step, ie for which, for example, a measurement of a sensor no or no valid user or not processible Has delivered the measured value.
- the correction is performed unaffected by the missing measured values of the time series, by one present at each time of the time series
- Observable vector is multiplied by a vector, and this vector is constructed from 1-entries, that the detected measured values in the observable vector through the 1-entries received in the correction of the system state.
- the index vector is able to account for some or all of those values in the observable vector for correction by 1-entries which were actually measured at the previous time step, ie for which, for example, a measurement of a sensor has provided a valid or processable measurement.
- an Architectural Teacher Forcing carried out by known methods, so that a foodsvek- tor of the system status at a particular time by the Observablenvektor of the observed time series at the preceding time is corrected. It is essential that the 1 entry is not an O entry. An extension to entries deviating from a 1 entry is conceivable, as long as they are not O entries.
- the lack of measured values of the time series are not complemented by an algorithm or kill ⁇ estimates.
- only the measured values in unmodified form are present as input variables to the variable vector.
- an associated future model-based value is determined from a system state for a future point in time at a fixed point in time.
- a plurality of observational vectors of the time series at past time points enter the correction of a state vector associated with a time following a respective past time. Since each ATF each time optimizes the following respective system state, a forecast of future values be ⁇ Sonder is promising, if as many historical values enter into the ATF. For a multiplicity of state vectors belonging to past times, the respective previous observable vectors can be received.
- the correction is made in a learning phase of the dynamic system or in an operating phase.
- the method can be advantageously used both to improve modeling and to improve a prognosis.
- the modeling of observables at a large number of times in the past and the comparison with observables actually measured at the respective times can advantageously be used to train an HCNN.
- the ATF can be performed for a variety of Systemzu ⁇ stands of past times in an operating phase, in order to optimize the prediction of a future Observablenvektors.
- the state transition is performed by applying a non-linear activation function and a linear function.
- features such as the hyperbolic tangent and linear algebra are used.
- the state vectors which enter into the functions of the status transition are already adjusted in particular by Cor ⁇ rection.
- the system state is configured as a state vector of observables and hidden states. Only for the observables, the correction can be made, the hidden states describe unobservable quantities and are not by measurements
- a difference vector is applied to the observables of the state vector at the time corresponding to ⁇ at the state transition, wherein the difference vector ei ⁇ NEN difference between the observables and known observables an associated Observablenvektors the time series writes be ⁇ .
- This is for example a gcardi ⁇ ges Architectural Teacher forcing method in which the modeled values are replaced with actual measured values from historical time series.
- a generated from a system state value, in particular a predicted future value for controlling a technical installation to a control unit of the technical system is transmitted or Op ⁇ optimization of the model for the correction.
- the invention further relates to a computer program product with a computer program having means for carrying out the method according to one of the preceding claims, when the computer program is executed on a program-controlled device.
- a computer program product such as a computer program means may be playing as provided or ⁇ as a storage medium, such as memory card, USB stick, CD-ROM, DVD, or in the form of a downloadable file from a server on a network supplied. This can be done, for example, in a wireless communication network by transmitting a corresponding file with the computer program product or Compu ⁇ terprogramm agent.
- program-controlled Einrich ⁇ tion is in particular a control device, such as for example a microprocessor for a smart card or the like in question.
- Figure 1 is a schematic representation of a known
- Figure 2 is a schematic representation of a recurrent
- neural network for modeling a dynamic system with Architectural Teacher Forcing based correction method according to an embodiment of the invention.
- Recurrent neural networks for modeling a temporal behavior of a dynamic system typically comprise several layers having a plurality can include neurons, and based on training data from known states of the dynamic system learned in a suitable manner such that future states of the dynamic Sys ⁇ tems can be predicted.
- ⁇ NEN figures are corresponding clusters of neurons, which supply stand vectors or Observablenvektoren or Differenzvekto- model reindeer, represented by circles.
- recurrent neural networks can be used for the computer-assisted prediction of electricity prices or energy requirements or of raw material prices.
- FIG. 1 shows schematically a conventional network topology which is used to model a dynamic system. It is a Historically Consistent Neural Network, in which state vectors s ⁇ , Sj ⁇ , St_ j _, s t , s t +] _, s t + 2, also called state vectors, at times t-3, t
- t-1, t, t + 1, and t + 2 are used to provide target variable vectors y, Yt-2 'Yt-1' Yt 'Yt + 1' Yt + 2 corresponding to respective times
- a matrix B the target values Yt-3 'Yt-2' Yt-1 'Yt' Yt + 1 un d Yt + 2 respectively from the internal to ⁇ stand vectors SJ-.3, SJ, 2, s tl 's t ' s t + 1 unc ⁇ s t + 2 is derived by a linear transformation.
- the modeling is refined by a so-called Architectural Teacher Forcing, ATF for short, by potentially observable or measurable
- Observable vectors u ⁇ , u ⁇ , u ' which may comprise a plurality Obser ⁇ vablen per vector, are taken into account in the state transition from a state vector to temporally following state vector.
- the values of the observables u.sub.i, for example, are used to determine the state vector S.sub.j.sup.- ,
- the actually observed values in the vector are added negatively to the starting layer, illustrated in Figure 1 by -Id. From the state vector S j -, 3 are simultaneously filtered potentially be ⁇ observable observable from the state vector over the matrix [id 0].
- [Id 0] is the identity matrix with 0 entries for removing the hidden observables from the state vector.
- the observable states are added positively to the output layer.
- the difference vector can be formed between the modeled observables and the actually observed observables. This difference vector enters the state transition negatively.
- the state vector becomes the observable entries of the state vector
- St_3 is corrected by the actually observed observable entries, since by applying the matrix the difference vector, with corresponding O-entries for the hidden states, is subtracted from the state vector.
- the hidden states remain unchanged by the O entries.
- a non-linear activation function is applied, e.g. a hyperbolic tangent function, illustrated in FIG. 1 by a circle with tanh therein, and with linear portions, illustrated by a weight matrix A, to arrive in the model at the subsequent state vector s.
- Observable values such as the quality of the vectors Yt + 1 'Yt + 2 unc ⁇ the corresponding observable values.
- a bias vector S Q is given as the initial state
- wel ⁇ cher is learned in a learning phase of the neural network together with the weight matrices A and B.
- the modeling tion of the dynamic behavior of a technical system which is described by a number of observables and by hidden states, not all training data are taken into account when learning the network, but the learning is based on a section of the network structure for a number of successive state vectors, for which known observable vectors from training data are available.
- the network is thus learned in time windows, which different ⁇ sections of successive
- FIG 2 is shown how the Architectural Teacher Forcing is modified from Figure 1 to achieve an improved Mo ⁇ model- ling at sparse time series.
- the Architectural Teacher Forcing is excluded from the Observableneinträgen of the state vector at a certain time t-3 and by the observed at this time actual Observablenvektor ⁇ - n formed Diffe ⁇ ence vector. It is also an index vector
- the described adaptation is applied in the generation of the state vectors for improved modeling of the system.
- Observablenvektors that, a quality characteristic value is to one of a sensor GELIE ⁇ ferten measured value addition forth. Missing entries can be identified, for example, by means of an entry "not available” or, in the simplest case, the measured value can only contain an invalid value (NaN, "Not a Number”). For example, the index vector is automatically adjusted for each time after the measured values have been acquired, so that the 0 and 1 entries match.
- Measurements can be taken into account the few available measured values in an advantageous manner for improving the modeling. For example, only at the times t-3 and t-1 is there a measured value for a temperature profile. The outside is located at the time t-3, only the measured value for the temperature ⁇ raturverlauf before and the other potentially detectable measured values ⁇ such as pressure or wind force, etc., are not available at this time. At the other times t-2, t-1 and t, for example, the other measured values such as pressure, wind force, etc. are completely available. It is not necessary an interpolation of Tempertatur tone for the intermediate state or measured value at time t-2 to advantageous ⁇ exemplary manner.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE102016209721.0A DE102016209721A1 (de) | 2016-06-02 | 2016-06-02 | Modellbasierte Ermittlung eines Systemzustandes mittels eines dynamischen Systems |
PCT/EP2017/062239 WO2017207317A1 (de) | 2016-06-02 | 2017-05-22 | Modellbasierte ermittlung eines systemzustandes mittels eines dynamischen systems |
Publications (1)
Publication Number | Publication Date |
---|---|
EP3449433A1 true EP3449433A1 (de) | 2019-03-06 |
Family
ID=59014574
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP17728082.3A Withdrawn EP3449433A1 (de) | 2016-06-02 | 2017-05-22 | Modellbasierte ermittlung eines systemzustandes mittels eines dynamischen systems |
Country Status (5)
Country | Link |
---|---|
US (1) | US20200320378A1 (de) |
EP (1) | EP3449433A1 (de) |
KR (1) | KR20190015415A (de) |
DE (1) | DE102016209721A1 (de) |
WO (1) | WO2017207317A1 (de) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111194451B (zh) * | 2017-10-20 | 2024-02-27 | 渊慧科技有限公司 | 门控激活单元运算的并行执行 |
CN113255208B (zh) * | 2021-04-21 | 2023-05-12 | 杭州新剑机器人技术股份有限公司 | 用于机器人的串联弹性执行器的神经网络模型预测控制方法 |
-
2016
- 2016-06-02 DE DE102016209721.0A patent/DE102016209721A1/de not_active Withdrawn
-
2017
- 2017-05-22 WO PCT/EP2017/062239 patent/WO2017207317A1/de unknown
- 2017-05-22 US US16/305,751 patent/US20200320378A1/en not_active Abandoned
- 2017-05-22 EP EP17728082.3A patent/EP3449433A1/de not_active Withdrawn
- 2017-05-22 KR KR1020187038200A patent/KR20190015415A/ko not_active Application Discontinuation
Also Published As
Publication number | Publication date |
---|---|
DE102016209721A1 (de) | 2017-12-07 |
KR20190015415A (ko) | 2019-02-13 |
WO2017207317A1 (de) | 2017-12-07 |
US20200320378A1 (en) | 2020-10-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP2112568B1 (de) | Verfahren zur rechnergestützten Steuerung und/oder Regelung eines technischen Systems | |
EP2724296A1 (de) | Verfahren zur rechnergestützten modellierung eines technischen systems | |
EP3662418A1 (de) | Verfahren und vorrichtung für maschinelles lernen in einer recheneinheit | |
AT511577A2 (de) | Maschinell umgesetztes verfahren zum erhalten von daten aus einem nicht linearen dynamischen echtsystem während eines testlaufs | |
WO2011128313A1 (de) | Verfahren zum rechnergestützten lernen eines rekurrenten neuronalen netzes zur modellierung eines dynamischen systems | |
WO2020216530A1 (de) | Verfahren zum bestimmen von restnutzungszyklen, restnutzungszyklusbestimmungsschaltung, restnutzungszyklusbestimmungsvorrichtung | |
CN107423496A (zh) | 一种新的降雨事件随机生成方法 | |
WO2017207317A1 (de) | Modellbasierte ermittlung eines systemzustandes mittels eines dynamischen systems | |
DE102016216945A1 (de) | Verfahren und Vorrichtung zum Ausführen einer Funktion basierend auf einem Modellwert eines datenbasierten Funktionsmodells basierend auf einer Modellgültigkeitsangabe | |
EP3599583A1 (de) | Bestimmung des verbrauchs an heiz- oder kühlenergie einer baulichen untereinheit | |
DE102020200051A1 (de) | Verfahren zum Bestimmen von Restnutzungszyklen, Restnutzungszyklusbestimmungsschaltung, Restnutzungszyklusbestimmungsvorrichtung | |
CN101520644A (zh) | 基于数据挖掘的分散控制系统特性函数在线拟合方法 | |
WO2020229050A1 (de) | Verfahren zur validierung von systemparametern eines energiesystems, verfahren zum betrieb eines energiesystems sowie energiemanagementsystem für ein energiesystem | |
DE102006020176A1 (de) | Verfahren zur Bereitstellung einer Patternprognose | |
DE102018125909A1 (de) | Verfahren zur Bestimmung zur Bestimmung einer Prozessgröße | |
DE102008014126B4 (de) | Verfahren zum rechnergestützten Lernen eines rekurrenten neuronalen Netzes | |
DE102013206274A1 (de) | Verfahren und Vorrichtung zum Anpassen eines nicht parametrischen Funktionsmodells | |
DE102011076969B4 (de) | Verfahren zum rechnergestützten Lernen einer Regelung und/oder Steuerung eines technischen Systems | |
DE102018128315A1 (de) | Verfahren, Vorrichtung, Computerprogramm und Computerprogrammprodukt zur Überprüfung eines ersten adaptiven Systemmodells | |
DE102019123447B4 (de) | Verfahren zum Betrieb von Fahrzeugkomponenten | |
EP3959571A1 (de) | Verfahren zum bestimmen von restnutzungszyklen, restnutzungszyklusbestimmungsschaltung, restnutzungszyklusbestimmungsvorrichtung | |
DE19527521C1 (de) | Lernverfahren für ein neuronales Netz | |
EP2384830B1 (de) | Verfahren zum Ermitteln von Parametern eines Modells für eine Walzanlage | |
EP4046149A1 (de) | Verfahren zum bestimmen mindestens eines zu bestimmenden restzeitwerts für eine anlage | |
DE102020116022A1 (de) | Verfahren zur zustandsbasierten Instandhaltung einer Zugangsvorrichtung |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
17P | Request for examination filed |
Effective date: 20181127 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
AX | Request for extension of the european patent |
Extension state: BA ME |
|
DAV | Request for validation of the european patent (deleted) | ||
DAX | Request for extension of the european patent (deleted) | ||
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN |
|
18D | Application deemed to be withdrawn |
Effective date: 20190625 |