-
公开(公告)号：EP4075346A1
公开(公告)日：2022-10-19
申请号：EP21168898.1
申请日：2021-04-16
发明人： Grimm, Stephan , Himmelhuber, Anna , Joblin, Mitchell , Ringsquandl, Martin , Runkler, Thomas , Zillner, Sonja
摘要： The invention relates to a computer-implemented method used for generating a predictive maintenance alert to perform a maintenance action concerning an entity of an industrial system such as a production system represented by a graph stored in a memory, said graph having nodes which represent entities of the industrial system such as components of the production system and having edges representing relations between entities of said industrial system, the method comprising the steps of: processing by a trained graph neural network the input graph to calculate at least one class prediction for selected entities of said industrial system represented by associated nodes; processing the calculated class prediction by a sub-symbolic GNN explainer to identify edges between nodes and associated features of nodes belonging to a sub-graph within the stored graph having influenced significantly the calculated class prediction; learning automatically logic class expressions belonging to explainer classes by running an inductive logic programming, ILP, algorithm using the at least one calculated class prediction as a target predicate on the basis of nodes, edges and features modelled as instances in a domain knowledge ontology of the technical domain of said industrial system and on the basis of nodes, edges and features of the stored graph and/or the sub-graph; and generating the predictive maintenance alert if the calculated class prediction belongs to a predefined maintenance relevant class, wherein the predictive maintenance alert is output via an interface along with at least one maintenance justification derived from the domain knowledge ontology on the basis of the associated learned logic class expressions to increase efficiency of the maintenance process and consequently efficiency of the monitored industrial system.
-
公开(公告)号：EP4009462A1
公开(公告)日：2022-06-08
申请号：EP20211327.0
申请日：2020-12-02
发明人： Beyer, Dagmar , Henselmeyer, Sylwia , Hildebrandt, Marcel , Joblin, Mitchell , Ringsquandl, Martin , Weber, Sebastian
摘要： Assistance apparatus for localizing errors in a monitored technical system (10) consisting of devices and/or transmission lines, comprising at least one processor configured to
- obtain values of actual attributes of the devices (11, 12) and/or of the transmission lines (13),
- determine an error probability for each device (11,12) and/or transmission line (13) by processing a graph neural network with the obtained actual values of attributes as input, wherein the graph neural network is trained by training attributes assigned to an attributed graph representation (26) of the technical system (10), and
- output an indication for such devices (11, 12) and/or transmission lines (13), whose error probability is higher than a predefined threshold.-
公开(公告)号：EP4235516A1
公开(公告)日：2023-08-30
申请号：EP22159221.5
申请日：2022-02-28
发明人： Akella, Chandra Sekhar , Hildebrandt, Marcel , Joblin, Mitchell , Mogoreanu, Serghei , Ringsquandl, Martin
摘要： A database stores a set of items, with each item (i 1 , i 2 , i 3 ) having technical attributes (x 1 , x 2 , x 3 ), and with each item representing a module that can be used in an engineering project of a first user, u 1 . A feature encoder (f) embeds each item based on its technical attributes into a low-dimensional vector space. Then, in a second step, a graph neural network (g) pools over these item embeddings to compute an updated user embedding for the first user e u 1 L . A decoder mapping (h) then addresses the recommendation task by outputting recommendation scores ( s u 1 , i ) for each item. That means, heuristically speaking, that the method and system lift the recommendation task to the level of technical attributes to overcome the sparsity problem caused by item sets that are not overlapping between user groups. Thus, when matching similar users, the method does not rely on users configuring exactly the same modules but rather on configured modules that are similar from a technical point of view.
-
公开(公告)号：EP4174711A1
公开(公告)日：2023-05-03
申请号：EP21204952.2
申请日：2021-10-27
摘要： The present invention relates generally to an assistance system and computer-implemented method for designing an engineering system, in particular an automation system.
The computer-implemented method for designing an engineering system comprises the following method steps performed by one or more components, wherein the components are software components which can be executed by one or more hardware and/or firmware components and/or one or more processors, wherein said engineering system is to be designed to meet predefined engineering constraints:
a) receiving, via a connector component (CC), a set of constraints (1) in a machine-readable representation, a domain model (DM) whose elements represent software components and/or hardware components of the engineering system (ES) and their relationships to each other and a set of engineering actions (AC);
b) creating and providing, by an assistant component (AS), at least one formal specification (R) of at least one engineering constraint by using recommendation from the received set of constraints (1), by using concepts related to elements in the domain model and by using a recommended set of constraint specification utilities (UT) preferably based on data types of the used concepts;
c) identifying, by an advisor component, at least one engineering action to be applied if violation of least one engineering constraint is discovered; and
d) displaying the identified, at least one, engineering action on a display component.-
公开(公告)号：EP3992856A1
公开(公告)日：2022-05-04
申请号：EP20204562.1
申请日：2020-10-29
发明人： Ringsquandl, Martin
摘要： The invention relates to a method for operating a device by using hierarchical reinforcement learning, whereby a complete operation task is to perform by said device, said method having the following steps:
a) Decomposing the complete operation task into a set of subtasks,
b) Representing dependencies among all subtasks of the complete task by a subtask graph (ggm) being direct acyclic and having nodes connected by edges, said nodes representing said subtasks and said edges representing relations among these subtasks,
c) using the subtask graph as input for a graph convolutional neural network in which from the input data by use of a convolutional operator for each network node a representation is obtained containing information about the relation of one network node and at least some neighbored network nodes and delivering encoded data in order to create a subtask graph vector (gv) in which a relation between nodes that are directly connected by an edge is encoded,
d) using the subtask vector (gv) obtained in step c) as input data for a deep neural network by which further encoded data are obtained;
e) Using said further encoded data to set up an action of the device (a) in order to perform at least one subtasks of the set of subtasks.-
公开(公告)号：EP4300382A1
公开(公告)日：2024-01-03
申请号：EP22181611.9
申请日：2022-06-28
发明人： Brikis, Georgia Olympia , Hasan, Rakebul , Hildebrandt, Marcel , Joblin, Mitchell , Koleva, Aneta , Mogoreanu, Serghei , Ringsquandl, Martin
摘要： Computer implemented method for classification of an object having a certain set of properties into a class containing objects with at least one common property,
said classes being arranged in a taxonomy tree being denoted by a class label and
said classification being performed by using a trained link prediction model which provides, for a textual description of an object, a prediction of a class label of said object,
wherein for the link prediction model the textual description of an object and the taxonomy tree (T) is used as input and
wherein the taxonomy tree comprises a set of nodes and a set of directed edges leading from a root node across nodes in at least one intermediate level, said nodes representing intermediate class labels, to leaf nodes at a final level, said leaf nodes not having child nodes and denoting a class label and said taxonomy tree containing a structure information defined by a branching ratio at each node and a number of levels, wherein for the application of the link prediction model, structure information is used so that the number of computations to obtain a class label for a given textual descriptions reduced by pursuing only selected paths from the root node to leaf nodes.-
公开(公告)号：EP4300375A1
公开(公告)日：2024-01-03
申请号：EP22182150.7
申请日：2022-06-30
摘要： Method and system for information retrieval
A database stores a collection of spreadsheet documents (CSD) describing an industrial system. The aim of the method is to create a digital twin of the industrial system on demand and on-the-fly in order to answer a given query. To this end, the spreadsheet documents are processed by a spreadsheet pre-processor (SPP) and a table encoder (TE). The query is processed by a graph pre-processor (GPP) and a graph encoder (GE). A table retriever (TR) computes a relevance score for each table that is proportional to a probability that the table is relevant to answer the query. Focusing on the most relevant tables, a cell retriever (CR) retrieves for each cell a relevance score that is proportional to a probability that the cell is relevant to answer the query. Finally, a user interface outputs the cell with the highest relevance score as an answer to the query. In other words, the method first retrieves relevant spreadsheet documents for a given query and then retrieves matching cells. As a result, a novel neural retrieval model and end-to-end training procedure are provided that allow to retrieve relevant information from spreadsheet documents with on-demand queries. Users do not have to manually look for relevant documents from which information should be extracted. This saves substantial experts' time. The neural retrieval model can be trained with respect to any standard schema depending on the domain of the industry (eClass for automation, Bricks for building technology, ISO-95).-
公开(公告)号：EP4137959A1
公开(公告)日：2023-02-22
申请号：EP21191938.6
申请日：2021-08-18
发明人： Brikis, Georgia Olympia , Hasan, Rakebul , Hildebrandt, Marcel , Joblin, Mitchell , Koleva, Aneta , Ringsquandl, Martin
IPC分类号： G06F16/215 , G06F16/23 , G06F16/332 , G06F16/901 , G06N3/08
摘要： An auto-encoder model (AEM) processes a datasets describing a physical part from a part catalogue in the form of a property co-occurrence graph (G), and performs entity resolution and auto-completion on the co-occurrence graph (G) in order to compute a corrected and/or completed dataset. According to an embodiment, the encoder (E) consists of a recurrent neural network (RNN) and a graph attention network (GAT). The decoder (D) contains a linear decoder (LD) for numeric values and a recurrent neural network decoder (RNN-D) for strings. The auto-encoder model provides an automated end-to-end solution that can auto-complete missing information as well as correct data errors such as misspellings or wrong values. The auto-encoder model is capable of auto-completion for highly unaligned part specification data with missing values. This has multiple benefits: First, the auto-encoder model can be trained completely unsupervised (self-supervised) as no labeled training data is required. Second, the auto-encoder model can capture correlation between any part specification property, value, and unit of measure. Third, the auto-encoder model is a single model instead of many models (for example, one for each property and unit) as would be the case in a Euclidean (table-based) missing data imputation algorithm. Fourth, the auto-encoder model can natively handle misspelled property and values terms and learn to align them. A further advantage is the ability for interactive user involvement. As the auto-encoder model operates purely on character-level, immediate feedback to the user can be given, for example after each character that the user is typing or editing.
-
公开(公告)号：EP3805993A1
公开(公告)日：2021-04-14
申请号：EP19202473.5
申请日：2019-10-10
摘要： The invention specifies a computerized method for generating graph-structured representations of a brownfield system (9), comprising the steps:
- Collecting (M1) training data of training systems, whereby training data consists of training pairs, with each training pair consisting of training sensor observations (3) and a training digital twin model (1),
- transforming (M2) the training digital twin models (1) into training graph-structured representations (2), whereby the training graph-structured representations (2) consist of nodes and links, whereby the nodes represent components of the training system and whereby the links represent relations between the components of the training system,
- training (M3) a graph generative model (3) to generate graph-structured representations of the brownfield system (9) using the training sensor observations (3) and the training graph-structured representations (2) of the training digital twin models (1), and
- generating (MS4) graph-structured representations of the brownfield system (9) using the trained graph generative model (3) and sensor observations of the brownfield system (8).-
10.发明公开
公开(公告)号：EP3151076A1
公开(公告)日：2017-04-05
申请号：EP16190483.4
申请日：2016-09-23
IPC分类号： G05B19/418 , G05B19/042
CPC分类号： G06F17/50 , G05B17/02 , G05B19/0426 , G06N99/005
摘要： Bei dem Verfahren zur Modellierung eines technischen Systems (TES) wird zunächst ein semantisches Systemmodell (SSM) des technischen Systems (TES) generiert wird und nachfolgend die Abhängigkeiten innerhalb des Systemmodells mittels einer Abhängigkeitsanalyse (DEA) analysiert, welche auf Eigenschaften des semantischen Systemmodells (SSM) beruht. Mit dem Verfahren ist es möglich z.B. Qualitätsprobleme bei Türen oder Ursachenaufklärung (ROC) einer anomalen Brennstofftemperatur in einer Gasturbine, auf der Grundlage vorangehender Ereignisse und Messungen vorherzusagen.
摘要翻译： 在技术系统建模方法中，生成技术系统的语义系统模型，并通过基于语义系统模型属性的依赖性分析来分析系统模型内的依赖关系。
-
-
-
-
-
-
-
-
-
搜索结果
22 条记录 (耗时 0.014 秒)
