专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明申请

WO2022263066A1 CENTRAL TECHNICAL COMPONENT FOR GENERATING AN AGGREGATED UPDATED MACHINE LEARNING PARAMETER 审中-公开
公开(公告)号：WO2022263066A1
公开(公告)日：2022-12-22
申请号：PCT/EP2022/062843
申请日：2022-05-11
申请人： SIEMENS AKTIENGESELLSCHAFT
发明人： KITTEL, Markus , KROMPASS, Denis , RUNKLER, Thomas
IPC分类号： G06N3/04 , G06N3/08
摘要： The invention is directed to Central technical component (10) for generating an aggregated updated Machine Learning parameter (12), wherein a. the central technical component (10) is configured for receiving at least one respective locally updated Machine Learning parameter (22) regarding a common Machine Learning system from each decentral technical component (20) of a plurality of decentral technical components; wherein b. the common Machine Learning system is consistently deployed on each decentral technical component (20) and the updates of the respective Machine Learning parameters (22) are performed locally on each decentral technical component (20); wherein c. the central technical component (10) is configured for receiving the number-counts of anomalous data samples (24) from each decentral technical component (20) of the plurality of decentral technical components; wherein d. the number-count of anomalous data samples (24) is derived from labeled data with a plurality of data samples which are each tagged with at least one label, wherein the labeled data is used during the local updates on the decentral technical units (20) and the labels are binary values that indicate if a data sample is from a normal operation or an anomalous operation; wherein e. the central technical component (10) is configured for generating the aggregated updated Machine Learning parameter (12) using an aggregation rule based on the received 30 plurality of locally updated Machine Learning parameters (22) and the associated number-counts of anomalous data samples (24); and f. the central technical component (10) is configured for transmitting the aggregated updated Machine Learning parameter (12) to each decentral technical component (20) of the plurality of decentral technical components for the common Machine Learning System. Further, the invention relates to a computer-implemented method, a computer program product, and a use of the central technical component.

2. 发明申请

WO2022253543A1 A METHOD FOR AMENDING OR ADDING FUNCTIONALITY TO AN AUTOMATION DEVICE 审中-公开
公开(公告)号：WO2022253543A1
公开(公告)日：2022-12-08
申请号：PCT/EP2022/062897
申请日：2022-05-12
申请人： SIEMENS AKTIENGESELLSCHAFT
发明人： ANICIC, Darko , REN, Haoyu , RUNKLER, Thomas
IPC分类号： G06F8/35 , G06F8/10 , G06N20/00 , G05B19/418
摘要： The present embodiments relate to a method for amending or adding functionality to an automation device in an automation systems domain.

3. 发明申请

WO2022223717A1 COMPUTER-IMPLEMENTED METHOD AND DATA MANAGEMENT SYSTEM FOR EXECUTING DATA MANAGEMENT OF PRODUCTION GOODS 审中-公开
公开(公告)号：WO2022223717A1
公开(公告)日：2022-10-27
申请号：PCT/EP2022/060597
申请日：2022-04-21
申请人： SIEMENS AKTIENGESELLSCHAFT
发明人： BALDAUF, Manfred , RUNKLER, Thomas , STEINBACHER, Frank , STOELBEN, Gunnar
IPC分类号： G06Q10/00 , G06Q50/02
摘要： In order to execute data management of production goods, by which the data being yielded in each phase of a production good-lifecycle is managed such that it can be used by any stakeholder and at changes in a data structure due to evolutionary production process of the good to optimize or certificate the production good and/or the production thereof, it is proposed to use lifecycle data (LCD, LCD1, LCD2,... LCDn) of a production good (PG) across a production good lifecycle and thereby (1) decentrally storing the lifecycle-data (LCD, LCD1, LCD2,... LCDn) and providing access to the stored lifecycle-data (LCD) through a "Peer-to-Peer-Network"-based blockchain technology (P2P-NW, NWN) for various stakeholders (SHI,...SHn) involved in a production good-lifecycle and participated regarding the "blockchain technology "-based storage and access as network nodes (NWN); (2) selectively accessing the lifecycle-data (LCD, LCD1, LCD2,...LCDn) through a cryptographic, in particular hierarchical, mechanism with a public key (PUK) and a private key (PRK, PRK1, PRK2), by which each stakeholder (SHI,... SHn) is able to grant or deny the access, in particular forndifferent stakeholders; and (3) enabling a dynamic data structure of the lifecycle-data (LCD) through a knowledge graph (KG) based on entities (ENT, ENT1, ENT2,... ENT9) being related to one other via edges (EDG, EDG1, EDG2,... EDG8), assigned attributes (ATR, ATR1, ATR2,... ATR8) and brought into thematic context or ontologies (OTG).

4. 发明申请

WO2021073919A1 EVALUATION FRAMEWORK FOR TIME SERIES DATA 审中-公开
公开(公告)号：WO2021073919A1
公开(公告)日：2021-04-22
申请号：PCT/EP2020/077805
申请日：2020-10-05
申请人： SIEMENS AKTIENGESELLSCHAFT
发明人： ARNOUT, Hiba , BRONNER, Johanna , KEHRER, Johannes , RUNKLER, Thomas
IPC分类号： G06N3/04 , G06N3/08
摘要： Various examples relate to an evaluation framework for a generated dataset of a data generation algorithm such as a generative adversarial network. The generated dataset includes a plurality of iterations of multiple instances of generated time series of data points. The evaluation framework provides multiple views. A first view includes at least one distance measure. The at least one distance measure is between the multiple instances of the generated time series and multiple instances of a reference time series, as a function of the plurality of iterations.

5. 发明申请

WO2021013466A1 NEURAL RELATION EXTRACTION WITHIN AND ACROSS SENTENCE BOUNDARIES 审中-公开
公开(公告)号：WO2021013466A1
公开(公告)日：2021-01-28
申请号：PCT/EP2020/067835
申请日：2020-06-25
申请人： SIEMENS AKTIENGESELLSCHAFT
发明人： GUPTA, Pankaj , RAJARAM, Subburam , ANDRASSY, Bernt , RUNKLER, Thomas
IPC分类号： G06F40/30 , G06F40/279 , G06N3/02
摘要： The invention provides a computer-implemented method for intersententially determining a semantic relationship between a first entity and a second entity in a natural language document, comprising at least the steps of: generating (S10) a first dependency parse tree, DPT, for a first origin sentence of the document which comprises the first entity, wherein each DPT comprises at least a root node; generating (S20) a second DPT for a second origin sentence of the document which mentions the second entity; linking (S30) the root nodes of the first DPT and the second DPT so as to create a chain of words, COW; determining (S40) for each word in the COW a subtree; generating (S50) for each word in the COW a subtree embedding vector (SEV) c w which is based at least on word embedding v ectors (WEV) x w of the words of the subtree; generating (S60) a representation vector pw for each word in the COW; and classifying (S80), using a recurrent neural network, the semantic relationship between the first entity (e1) and the second entity (e2), based on the input representation vectors pw.

6. 发明申请

WO2022028690A1 TOPICBERT: DECARBONIZING BERT WITH TOPICS FOR EFFICIENT DOCUMENT CLASSIFICATION 审中-公开
公开(公告)号：WO2022028690A1
公开(公告)日：2022-02-10
申请号：PCT/EP2020/072039
申请日：2020-08-05
申请人： SIEMENS AKTIENGESELLSCHAFT
发明人： GUPTA, Pankaj , RUNKLER, Thomas , SAXENA, Khushbu
IPC分类号： G06N3/08 , G06N3/04 , G06F40/30 , G06N3/0454 , G06N3/084
摘要： The present invention relates to a computer-implemented method of fine-tuning Natural Language Processing (NLP) models and a data processing system for hosting NLP models. A NLP model is fine-tuned by jointly training a NN-based Topic Model (TM) and a parallel large-scale pre-trained Language Model (LM) based on a projected vector ℎp that is a combination and projection of a document topic proportion ℎTM , which is generated by the NN-based TM based on its scalable TM parameters from an input training text document of a multitude of training text documents, and of a contextualized document representation ? oCLS, which is generated by the large-scale pre-trained LM based on its scalable LM parameters from the same input training text document.

7. 发明申请

WO2013092702A1 A METHOD FOR COMPUTER-ASSISTED DETERMINATION OF THE USAGE OF ELECTRICAL ENERGY PRODUCED BY A POWER GENERATION PLANT, PARTICULARLY A RENEWABLE POWER GENERATION PLANT 审中-公开
标题翻译：一种用于计算机辅助确定发电厂生产的电能使用的方法，特别是可再生能源发电厂
公开(公告)号：WO2013092702A1
公开(公告)日：2013-06-27
申请号：PCT/EP2012/076136
申请日：2012-12-19
申请人： SIEMENS AKTIENGESELLSCHAFT
发明人： EGEDAL, Per , GROTHMANN, Ralph , STERZING, Volkmar , RUNKLER, Thomas
IPC分类号： G06N3/04 , G06Q50/06
CPC分类号： G01R21/1333 , G06N3/02 , G06N3/0454 , G06N7/005 , G06Q50/06 , H02J3/383 , H02J3/386 , H02J2003/003 , Y02E10/563 , Y02E10/763 , Y04S10/54
摘要： The invention relates to a method for computer-assisted determination of the usage of electrical energy produced by a power generation plant, particularly a renewable power generation plant. The method uses a plurality of neural networks (NN1, NN2) having a different structure or being learned differently for calculating future energy amounts produced by a power generation plant. To do so, the energy outputs of the power generation plant forecasted by the plurality of the neural networks (NN1, NN2) are used to build histograms. Based on those histograms, energy amounts for different confidence levels describing the likelihood of the availability of the energy amount are determined and different uses are assigned to different energy amounts. Particularly, energy amounts having a higher likelihood of availability in the future are sold at higher prices than other energy amounts. The method of the invention is preferably used for renewable power generation plants for which the energy output is fluctuating according to environmental conditions. E.g., the invention may be used for a wind power generation plant or a solar power generation plant.
摘要翻译：本发明涉及一种用于计算机辅助确定由发电厂，特别是可再生能源发电厂产生的电能的使用的方法。该方法使用具有不同结构的多个神经网络（NN1，NN2）或不同的学习来计算由发电厂产生的未来能量。为此，使用由多个神经网络（NN1，NN2）预测的发电厂的能量输出来构建直方图。基于这些直方图，确定描述能量可用性的可能性的不同置信水平的能量量，并将不同的用途分配给不同的能量。特别地，将来可能性更高的能量以比其他能量更高的价格出售。本发明的方法优选用于能量输出根据环境条件波动的可再生能源发电厂。例如，本发明可用于风力发电厂或太阳能发电厂。

8. 发明申请

WO2005103374A2 VERFAHREN UND ANLAGE ZUR AUFBEREITUNG VON ALTPAPIER 审中-公开
标题翻译：方法和系统进行处理废纸
公开(公告)号：WO2005103374A2
公开(公告)日：2005-11-03
申请号：PCT/EP2005/051691
申请日：2005-04-18
申请人： SIEMENS AKTIENGESELLSCHAFT , DINKEL, Markus , MICKAL, Volkmar , RUNKLER, Thomas , SIEBER, Albrecht , VILLFORTH, Klaus
发明人： DINKEL, Markus , MICKAL, Volkmar , RUNKLER, Thomas , SIEBER, Albrecht , VILLFORTH, Klaus
IPC分类号： D21C9/10
CPC分类号： D21C5/02 , Y02W30/648

9. 发明申请

WO2005103373A3 VERFAHREN ZUM STEUERN EINES BLEICHPROZESSES FÜR DIE ALTPAPIERAUFBEREITUNG, SOWIE EINE BLEICHVORRICHTUNG ZUR DURCHFÜHRUNG EINES SOLCHEN VERFAHRENS 审中-公开
公开(公告)号：WO2005103373A3
公开(公告)日：2005-11-03
申请号：PCT/EP2005/051729
申请日：2005-04-19
申请人： SIEMENS AKTIENGESELLSCHAFT , HORN, Joachim , LENZ, Henning , RUNKLER, Thomas , DINKEL, Markus , VILLFORTH, Klaus , GÖTZ, Reinhold , MICKAL, Volkmar , SCHMIDT, Thomas , SIEBER, Albrecht
发明人： HORN, Joachim , LENZ, Henning , RUNKLER, Thomas , DINKEL, Markus , VILLFORTH, Klaus , GÖTZ, Reinhold , MICKAL, Volkmar , SCHMIDT, Thomas , SIEBER, Albrecht
IPC分类号： D21C9/10
摘要： Die Erfindung betrifft ein Verfahren zum Steuern eines Bleichprozesses für die Altpapieraufbereitung, mit folgenden Schritten: - Bereitstellen einer Faserstoffsuspension; - Messen der Weiße der Faserstoffsuspension; - Vermischen der Faserstoffsuspension mit einem Bleichmittel; - Zuführen der vermischten Faserstoffsuspension in einen Bleichbehälter (7); - Entnehmen der vermischten Faserstoffsuspension aus dem Bleichbehälter nach einer bestimmten Zeitdauer; dadurch gekennzeichnet, dass das Bleichmittel ein Stoffgemisch aus mehreren Stoffen aufweist, wobei die Menge und/oder die Zusammensetzung des zugesetzten Bleichmittels abhängig von der gemessenen Weiße der Faserstoffsuspension so eingestellt wird, dass eine möglichst konstante Sollweiße der entnommenen Faserstoffsuspension erreicht wird.

10. 发明申请

WO2022013125A1 VERFAHREN ZUM KONFIGURIEREN EINES STEUERUNGSAGENTEN FÜR EIN TECHNISCHES SYSTEM SOWIE STEUEREINRICHTUNG 审中-公开
公开(公告)号：WO2022013125A1
公开(公告)日：2022-01-20
申请号：PCT/EP2021/069269
申请日：2021-07-12
申请人： SIEMENS AKTIENGESELLSCHAFT
发明人： RUNKLER, Thomas , SWAZINNA, Phillip , UDLUFT, Steffen
IPC分类号： G06N3/00 , G06N3/04 , G06N3/08 , G06N5/00 , G06N7/00
摘要： Zum Konfigurieren eines Steuerungsagenten (POL) werden vorgegebene Trainingsdaten eingelesen, die Zustandsdatensätze (S), Aktionsdatensätze (A) sowie resultierende Performanzwerte (R) des technischen Systems (TS) spezifizieren. Anhand der Trainingsdaten wird ein datenbasiertes Dynamikmodell (NN) darauf trainiert, anhand eines Zustandsdatensatzes (S) und eines Aktionsdatensatzes (A) einen resultierenden Performanzwert (R) zu reproduzieren. Ferner wird ein Aktionsbewertungsprozess (VAE) darauf trainiert, anhand eines Zustandsdatensatzes (S) und eines Aktionsdatensatzes (A) nach Vornahme einer Informationsreduktion den Aktionsdatensatz (A) zu reproduzieren, wobei ein Reproduktionsfehler (DR, DO, Dl) ermittelt wird. Zum Training des Steuerungsagenten (POL) werden dem trainierten Dynamikmodell (NN), dem trainierten Aktionsbewertungsprozess (VAE) sowie dem Steuerungsagenten (POL) Trainingsdaten zugeführt. Dabei werden vom trainierten Dynamikmodell (NN) ausgegebene Performanzwerte (RI, R2) in eine vorgegebene Performanzfunktion (P) eingespeist. In die Performanzfunktion (P) werden weiterhin vom trainierten Aktionsbewertungsprozess (VAE) ausgegebene Reproduktionsfehler (DO, Dl) als performanzverringernde Einflussgroßen eingespeist. Damit wird der Steuerungsagent (POL) darauf trainiert, anhand eines Zustandsdatensatzes (S) einen die Performanzfunktion (P) optimierenden Aktionsdatensatz (A) auszugeben.

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式