Fault diagnosis in a complex system, such as a nuclear plant, using probabilistic reasoning
AssigneeSiemens Corporate Research, Inc.
InventorsChristian Darken, Nugroho I. Santoso, Jochen Erdmann
Life of this patent
Estimating the term…
Abstract
A system for generating a diagnosis of the probable cause of detected faults. Preferably, a user interface is provided and a Bayesien network is employed wherein probabilities are generated automatically and a manual process is used to construct a probability table. The system also allows multiple hypotheses and/or diagnoses to be presented simultaneously to the operator.
Claims14
A method for making a diagnoses in a complex system, comprising the following steps:
- providing a causal network which is trained to provide a diagnosis based on input of extracted features from real time measurement values taken from said complex system, and
- providing a graphical user interface which is responsive to input of said extracted features for presenting to an operator responsible for making a diagnoses of the complex system, simultaneous multiple hypotheses as to the current condition of said complex system;
- wherein said extracted features are determined using sensor measurement data produced by a simulation of said complex system, a graph for the causal network is constructed in a manual manner based on expert knowledge, and conditional probability tables for the causal network are constructed in an automated manner.
A method for making a diagnoses in a complex system, comprising the following steps:
- providing a causal network which is constructed so as to model at least part of the complex system;
- said providing step comprising the sub steps of:
- simulating a specific fault condition,
- extracting features from a result of said simulation, and
- training the causal network with said extracted features,
- obtaining real time measurement values from said complex system,
- extracting features from said real time measurement values, and
- applying said extracted features from said real time measurement values to said causal
- network, so as to develop therefrom a diagnoses of said complex system;
- wherein a graph for said causal network is constructed in a manual manner based on the knowledge of an expert, and conditional probability tables are constructed for said causal network in a automated manner.
Apparatus for assisting in diagnosing faults in a complex system, comprising:
- a causal network which is trained to provide a diagnosis based upon input of extracted features from real time measurement values taken from said complex system, and
- a graphical user interface which is responsive to input of said extracted features from said causal network for presenting to an operator responsible for making a diagnosis of the complex system, simultaneous multiple hypotheses as to the current condition of said complex system;
- wherein said extracted features are determined using sensor measurement data produced by a simulation of said complex system, a graph for the causal network is constructed in a manual manner based on expert knowledge, and conditional probability tables for the causal network are constructed in an automated manner.
Dependencies are inferred from the claim text.