The long term management of a production asset raises several major issues, among which rank the ... more The long term management of a production asset raises several major issues, among which rank the technical management of the plant, its economics, and the fleet level perspective one has to adopt. Decision makers are therefore faced with the need to define long term policies (up to the end of asset operation) that take into account multiple criteria including safety (which is paramount) and performance. In this paper we first remind the reader of the EDF three-level methodology for asset management. As introduced in PVP 2003 and PVP 2004, this methodology addresses the component/technical level (how to safely operate daily and invest for the future), the plant level (how to translate technical decisions into plant-wide consequences including economic performance), and the fleet level (how to manage a large number of similar assets). We then focus on the software tool that implements this methodology in order to allow decision makers to define, evaluate, and analyze long term plant o...
... Miguel Munoz Zuniga · Josselin Garnier · Emmanuel Remy · Etienne de Rocquigny ... A well-deve... more ... Miguel Munoz Zuniga · Josselin Garnier · Emmanuel Remy · Etienne de Rocquigny ... A well-developed state of the art of stochastic algo-rithms for the failure probability estimation can be found in Rubinstein and Kroese (2007) [34], Gille-Genest (1999) [16] and Cannamela (2007 ...
Volume 7: Operations, Applications, and Components, 2006
The life management of a nuclear power plant raises several major issues amongst which ranks the ... more The life management of a nuclear power plant raises several major issues amongst which ranks the aging management of the key components of the plant, both from a technical and an economic point of view. Decision-makers are thus faced with the need to define the best strategy in order to achieve the best possible performance which requires both a very precise modeling of the plant and a detailed analysis of all risks potentially incurred. In this paper, we wish to provide the reader with an overview of how advanced expert elicitation techniques can help identify, structure, quantify and feed sensitive data into a risk-based information system which can then be used for risk-based asset management evaluation. First we focus on the way knowledge management techniques allow EDF to structure and collect life-cycle management data into knowledge-based information systems. The elicitation of component experts is key, particularly in order to get technical information on the future behavior...
Considering reliability models In a Bayesian context, we propose an approach where experts opinio... more Considering reliability models In a Bayesian context, we propose an approach where experts opinions are supposed to come from fictitious data. Acting in such a way allows the analyst to weight the importance of the expert opinion in regard to the actual sample size in a sensible and reliable way. The control of experts knowledge becomes a control of the Fisher information on the reliability model parameters. Thus, the comparison between Feedback Experience Data (FED) and expert data through the reliability model and the prior distribution is easy and makes simple the calibration of the hyperparameters prior distribution to control the importance of expert contribution compared to the information provided by the observed sample. The presented approach is exemplified with Weibull models.
Within the structural reliability context, the aim of this paper is to present a new accelerated ... more Within the structural reliability context, the aim of this paper is to present a new accelerated Monte-Carlo simulation method, named ADS-Adaptive Directional Stratification-, and designed to overcome the following industrial constraints: robustness of the estimation of a low structural failure probability (less than 10 −3), limited computational resources and complex (albeit often monotonic) physical model. This new stochastic technique is an original variant of adaptive accelerated simulation method, combining stratified sampling and directional simulation and including two steps in the adaptation stage (ADS-2). First, we theoretically study the properties of two possible failure probability estimators and get the asymptotic and nonasymptotic expressions of their variances. Then, we propose some improvements for our new method. To begin with, we focus on the root-finding algorithm required for the directional approach: we present a stop criterion for the dichotomic method and a strategy to reduce the required number of calls to the costly physical model under monotonic hypothesis. Lastly, to overcome the limit involved by the increase of the input dimension, we introduce the ADS-2 + method which has the same ground as the ADS-2 method, but additionally uses a statistical test to detect the most significant inputs and carries out the stratification only along them. To conclude, we test the ADS-2 and ADS-2 + methods on academic examples in order to compare them with the classical structural reliability methods and to make a numerical sensitivity analysis over some parameters. We also apply the methods to a flood model and a nuclear reactor pressurized vessel model, to practically demonstrate their interest on real industrial examples.
Le Centre pour la Communication Scientifique Directe - HAL - memSIC, Oct 13, 2020
Cet article propose une revue bibliographique des principales approches de résolution du problème... more Cet article propose une revue bibliographique des principales approches de résolution du problème de déséquilibre des classes en apprentissage automatique.
International audienceThe (two-parameters) Weibull distribution is extensively used in reliabilit... more International audienceThe (two-parameters) Weibull distribution is extensively used in reliability engineering for modelling the lifetime of ageing systems (with increasing hazard rate) as well as improving systems (with decreasing hazard rate). Although this distribution is widely used, the checking of its relevance for a given data set is not always done, or done by elementary techniques such as Weibull plots. There exists more sophisticated techniques which aim to determine if a given model is adapted to a given data set, the goodness-of-fit (GoF) tests. Recently, Krit et al (2014) proposed a comprehensive review and comparison of GoF tests for the Weibull distribution, for complete data sets. However, in practice, data sets are often incomplete. We will consider here the case of type II censoring, where the largest lifetimes are not observed. The problem is to determine if the Weibull distribution is a relevant model for this kind of data sets. The aim of this paper is to presen...
Congrès Lambda Mu 21 « Maîtrise des risques et transformation numérique : opportunités et menaces », Oct 16, 2018
Cette communication présente une comparaison de plusieurs algorithmes de classification par appre... more Cette communication présente une comparaison de plusieurs algorithmes de classification par apprentissage automatique sur un cas d'usage réel du domaine de la production d'électricité nucléaire. Leurs performances sont examinées et une discussion est proposée sur leur intérêt pratique, leurs avantages et inconvénients, leur précaution d'usage et leur pertinence dans le cadre applicatif industriel concerné.
L'objectif de cette communication est de presenter, dans le cadre des sessions interactives, ... more L'objectif de cette communication est de presenter, dans le cadre des sessions interactives, le logiciel MARS (Maintenance Assessment of Repairable Systems), developpe conjointement par EDF R&D et le Laboratoire Jean Kuntzmann de l'INP Grenoble. Cet outil logiciel inedit permet de mettre en œuvre les modeles de maintenance imparfaite et d'estimer conjointement les effets du vieillissement et des maintenances sur des jeux de donnees issus du retour d'experience.
Subsequence anomaly (or outlier) detection in long sequences is an important problem with applica... more Subsequence anomaly (or outlier) detection in long sequences is an important problem with applications in a wide range of domains. However, the approaches that have been proposed so far in the literature have severe limitations: they either require prior domain knowledge or become cumbersome and expensive to use in situations with recurrent anomalies of the same type. In this work, we address these problems and propose NormA, a novel approach, suitable for domain-agnostic anomaly detection. NormA is based on a new data series primitive, which permits to detect anomalies based on their (dis)similarity to a model that represents normal behavior. The experimental results on several real datasets demonstrate that the proposed approach correctly identifies all single and recurrent anomalies of various types, with no prior knowledge of the characteristics of these anomalies (except for their length). Moreover, it outperforms by a large margin the current state-of-the art algorithms in terms of accuracy, while being orders of magnitude faster.
International audienceThe aim of this paper is to present new likelihood based goodness-of-fit te... more International audienceThe aim of this paper is to present new likelihood based goodness-of-fit tests for the two-parameter Weibull distribution. These tests consist in nesting the Weibull distribution in three-parameter generalized Weibull families and testing the value of the third parameter by using the Wald, score and likelihood ratio procedures. We simplify the usual likelihood based tests by getting rid of the nuisance parameters, using three estimation methods. The proposed tests are not asymptotic. A comprehensive comparison study is presented. Among a large range of possible GOF tests, the best ones are identified. The results depend strongly on the shape of the underlying hazard rate
Volume 7: Operations, Applications and Components, 2014
Nuclear Reactor Pressure Vessel (RPV) integrity is a major issue concerning plant safety and this... more Nuclear Reactor Pressure Vessel (RPV) integrity is a major issue concerning plant safety and this component is one of the few within a Pressurized Water Reactor (PWR) whose replacement is not considered as feasible. To ensure that adequate margins against failure are maintained throughout the vessel service life, research engineers have developed and applied computational tools to study and assess the probability of pressure vessel failure during operating and postulated loads. The Materials Ageing Institute (MAI) sponsored a benchmark study to compare the results from software developed in France, Japan and the United States to compute the probability of flaw initiation in reactor pressure vessels. This benchmark study was performed to assess the similarities and differences in the software and to identify the sources of any differences that were found. Participants in this work included researchers from EDF in France, CRIEPI in Japan and EPRI in the United States, with each organi...
Volume 7: Operations, Applications and Components, 2007
The long term management of a production asset raises several major issues among which rank the t... more The long term management of a production asset raises several major issues among which rank the technical management of the plant, its economics and the fleet level perspective one has to adopt. Decision makers are therefore faced with the need to define long term policies (up to the end of asset operation) which take into account multiple criteria including safety (which is paramount) and performance. In this paper we first remind the reader of the EDF three-level methodology for asset management. As introduced in PVP 03 [1] and PVP 04 [2], this methodology addresses the component/technical level (how to safely operate daily and invest for the future), the plant level (how to translate technical decisions into plant-wide consequences including economic performance) and the fleet level (how to manage a large number of similar assets). We then focus on the software tools — introduced in PVP 04 [2] and PVP 05 [4] — that implement this methodology in order to allow decision makers to d...
The long term management of a production asset raises several major issues, among which rank the ... more The long term management of a production asset raises several major issues, among which rank the technical management of the plant, its economics, and the fleet level perspective one has to adopt. Decision makers are therefore faced with the need to define long term policies (up to the end of asset operation) that take into account multiple criteria including safety (which is paramount) and performance. In this paper we first remind the reader of the EDF three-level methodology for asset management. As introduced in PVP 2003 and PVP 2004, this methodology addresses the component/technical level (how to safely operate daily and invest for the future), the plant level (how to translate technical decisions into plant-wide consequences including economic performance), and the fleet level (how to manage a large number of similar assets). We then focus on the software tool that implements this methodology in order to allow decision makers to define, evaluate, and analyze long term plant o...
... Miguel Munoz Zuniga · Josselin Garnier · Emmanuel Remy · Etienne de Rocquigny ... A well-deve... more ... Miguel Munoz Zuniga · Josselin Garnier · Emmanuel Remy · Etienne de Rocquigny ... A well-developed state of the art of stochastic algo-rithms for the failure probability estimation can be found in Rubinstein and Kroese (2007) [34], Gille-Genest (1999) [16] and Cannamela (2007 ...
Volume 7: Operations, Applications, and Components, 2006
The life management of a nuclear power plant raises several major issues amongst which ranks the ... more The life management of a nuclear power plant raises several major issues amongst which ranks the aging management of the key components of the plant, both from a technical and an economic point of view. Decision-makers are thus faced with the need to define the best strategy in order to achieve the best possible performance which requires both a very precise modeling of the plant and a detailed analysis of all risks potentially incurred. In this paper, we wish to provide the reader with an overview of how advanced expert elicitation techniques can help identify, structure, quantify and feed sensitive data into a risk-based information system which can then be used for risk-based asset management evaluation. First we focus on the way knowledge management techniques allow EDF to structure and collect life-cycle management data into knowledge-based information systems. The elicitation of component experts is key, particularly in order to get technical information on the future behavior...
Considering reliability models In a Bayesian context, we propose an approach where experts opinio... more Considering reliability models In a Bayesian context, we propose an approach where experts opinions are supposed to come from fictitious data. Acting in such a way allows the analyst to weight the importance of the expert opinion in regard to the actual sample size in a sensible and reliable way. The control of experts knowledge becomes a control of the Fisher information on the reliability model parameters. Thus, the comparison between Feedback Experience Data (FED) and expert data through the reliability model and the prior distribution is easy and makes simple the calibration of the hyperparameters prior distribution to control the importance of expert contribution compared to the information provided by the observed sample. The presented approach is exemplified with Weibull models.
Within the structural reliability context, the aim of this paper is to present a new accelerated ... more Within the structural reliability context, the aim of this paper is to present a new accelerated Monte-Carlo simulation method, named ADS-Adaptive Directional Stratification-, and designed to overcome the following industrial constraints: robustness of the estimation of a low structural failure probability (less than 10 −3), limited computational resources and complex (albeit often monotonic) physical model. This new stochastic technique is an original variant of adaptive accelerated simulation method, combining stratified sampling and directional simulation and including two steps in the adaptation stage (ADS-2). First, we theoretically study the properties of two possible failure probability estimators and get the asymptotic and nonasymptotic expressions of their variances. Then, we propose some improvements for our new method. To begin with, we focus on the root-finding algorithm required for the directional approach: we present a stop criterion for the dichotomic method and a strategy to reduce the required number of calls to the costly physical model under monotonic hypothesis. Lastly, to overcome the limit involved by the increase of the input dimension, we introduce the ADS-2 + method which has the same ground as the ADS-2 method, but additionally uses a statistical test to detect the most significant inputs and carries out the stratification only along them. To conclude, we test the ADS-2 and ADS-2 + methods on academic examples in order to compare them with the classical structural reliability methods and to make a numerical sensitivity analysis over some parameters. We also apply the methods to a flood model and a nuclear reactor pressurized vessel model, to practically demonstrate their interest on real industrial examples.
Le Centre pour la Communication Scientifique Directe - HAL - memSIC, Oct 13, 2020
Cet article propose une revue bibliographique des principales approches de résolution du problème... more Cet article propose une revue bibliographique des principales approches de résolution du problème de déséquilibre des classes en apprentissage automatique.
International audienceThe (two-parameters) Weibull distribution is extensively used in reliabilit... more International audienceThe (two-parameters) Weibull distribution is extensively used in reliability engineering for modelling the lifetime of ageing systems (with increasing hazard rate) as well as improving systems (with decreasing hazard rate). Although this distribution is widely used, the checking of its relevance for a given data set is not always done, or done by elementary techniques such as Weibull plots. There exists more sophisticated techniques which aim to determine if a given model is adapted to a given data set, the goodness-of-fit (GoF) tests. Recently, Krit et al (2014) proposed a comprehensive review and comparison of GoF tests for the Weibull distribution, for complete data sets. However, in practice, data sets are often incomplete. We will consider here the case of type II censoring, where the largest lifetimes are not observed. The problem is to determine if the Weibull distribution is a relevant model for this kind of data sets. The aim of this paper is to presen...
Congrès Lambda Mu 21 « Maîtrise des risques et transformation numérique : opportunités et menaces », Oct 16, 2018
Cette communication présente une comparaison de plusieurs algorithmes de classification par appre... more Cette communication présente une comparaison de plusieurs algorithmes de classification par apprentissage automatique sur un cas d'usage réel du domaine de la production d'électricité nucléaire. Leurs performances sont examinées et une discussion est proposée sur leur intérêt pratique, leurs avantages et inconvénients, leur précaution d'usage et leur pertinence dans le cadre applicatif industriel concerné.
L'objectif de cette communication est de presenter, dans le cadre des sessions interactives, ... more L'objectif de cette communication est de presenter, dans le cadre des sessions interactives, le logiciel MARS (Maintenance Assessment of Repairable Systems), developpe conjointement par EDF R&D et le Laboratoire Jean Kuntzmann de l'INP Grenoble. Cet outil logiciel inedit permet de mettre en œuvre les modeles de maintenance imparfaite et d'estimer conjointement les effets du vieillissement et des maintenances sur des jeux de donnees issus du retour d'experience.
Subsequence anomaly (or outlier) detection in long sequences is an important problem with applica... more Subsequence anomaly (or outlier) detection in long sequences is an important problem with applications in a wide range of domains. However, the approaches that have been proposed so far in the literature have severe limitations: they either require prior domain knowledge or become cumbersome and expensive to use in situations with recurrent anomalies of the same type. In this work, we address these problems and propose NormA, a novel approach, suitable for domain-agnostic anomaly detection. NormA is based on a new data series primitive, which permits to detect anomalies based on their (dis)similarity to a model that represents normal behavior. The experimental results on several real datasets demonstrate that the proposed approach correctly identifies all single and recurrent anomalies of various types, with no prior knowledge of the characteristics of these anomalies (except for their length). Moreover, it outperforms by a large margin the current state-of-the art algorithms in terms of accuracy, while being orders of magnitude faster.
International audienceThe aim of this paper is to present new likelihood based goodness-of-fit te... more International audienceThe aim of this paper is to present new likelihood based goodness-of-fit tests for the two-parameter Weibull distribution. These tests consist in nesting the Weibull distribution in three-parameter generalized Weibull families and testing the value of the third parameter by using the Wald, score and likelihood ratio procedures. We simplify the usual likelihood based tests by getting rid of the nuisance parameters, using three estimation methods. The proposed tests are not asymptotic. A comprehensive comparison study is presented. Among a large range of possible GOF tests, the best ones are identified. The results depend strongly on the shape of the underlying hazard rate
Volume 7: Operations, Applications and Components, 2014
Nuclear Reactor Pressure Vessel (RPV) integrity is a major issue concerning plant safety and this... more Nuclear Reactor Pressure Vessel (RPV) integrity is a major issue concerning plant safety and this component is one of the few within a Pressurized Water Reactor (PWR) whose replacement is not considered as feasible. To ensure that adequate margins against failure are maintained throughout the vessel service life, research engineers have developed and applied computational tools to study and assess the probability of pressure vessel failure during operating and postulated loads. The Materials Ageing Institute (MAI) sponsored a benchmark study to compare the results from software developed in France, Japan and the United States to compute the probability of flaw initiation in reactor pressure vessels. This benchmark study was performed to assess the similarities and differences in the software and to identify the sources of any differences that were found. Participants in this work included researchers from EDF in France, CRIEPI in Japan and EPRI in the United States, with each organi...
Volume 7: Operations, Applications and Components, 2007
The long term management of a production asset raises several major issues among which rank the t... more The long term management of a production asset raises several major issues among which rank the technical management of the plant, its economics and the fleet level perspective one has to adopt. Decision makers are therefore faced with the need to define long term policies (up to the end of asset operation) which take into account multiple criteria including safety (which is paramount) and performance. In this paper we first remind the reader of the EDF three-level methodology for asset management. As introduced in PVP 03 [1] and PVP 04 [2], this methodology addresses the component/technical level (how to safely operate daily and invest for the future), the plant level (how to translate technical decisions into plant-wide consequences including economic performance) and the fleet level (how to manage a large number of similar assets). We then focus on the software tools — introduced in PVP 04 [2] and PVP 05 [4] — that implement this methodology in order to allow decision makers to d...
Uploads
Papers by Emmanuel REMY