PSAM 16 - Session T05 Overview

PSAM 16 Conference Session T05 Overview

Session Chair: Craig Primer (craig.primer@inl.gov)

Paper 1 VI321

Lead Author: Vivek Agarwal Co-author(s): Matthew Yarlett and Brad Diggans

Digital Nuclear Platform for Automation of Maintenance Activities
This presentation would describe requirements that established through development and design of the Nuclear Digital Platform (NDP) by PKMJ Technical Services, referred to as PKMJ NDP. The PKMJ NDP was designed to be a centralized solution to support large-volume storage and access, large-volume data processing, advanced data analytics techniques (e.g., artificial intelligence/machine learning), data and information visualization, and reporting. PKMJ utilizes enhanced business intelligence techniques in support of NPP customers; however, these tools are used in industries throughout the world. The PKMJ NDP takes input from nuclear industry subject matter experts and combines it with input from mixed discipline teams of data experts in order to apply cutting-edge principles to rapidly explore data, unlock and test new ideas, and turn those ideas into services. The presentation would also describe the development of automated work package in the PKMJ NDP reducing the risk of human error, optimizing resources allocation, automating the process, and leading to cost-saving if implemented.

Digital Nuclear Platform for Automation of Maintenance Activities

This presentation would describe requirements that established through development and design of the Nuclear Digital Platform (NDP) by PKMJ Technical Services, referred to as PKMJ NDP. The PKMJ NDP was designed to be a centralized solution to support large-volume storage and access, large-volume data processing, advanced data analytics techniques (e.g., artificial intelligence/machine learning), data and information visualization, and reporting. PKMJ utilizes enhanced business intelligence techniques in support of NPP customers; however, these tools are used in industries throughout the world. The PKMJ NDP takes input from nuclear industry subject matter experts and combines it with input from mixed discipline teams of data experts in order to apply cutting-edge principles to rapidly explore data, unlock and test new ideas, and turn those ideas into services. The presentation would also describe the development of automated work package in the PKMJ NDP reducing the risk of human error, optimizing resources allocation, automating the process, and leading to cost-saving if implemented.

Paper VI321 | |

Name: Vivek Agarwal (vivek.agarwal@inl.gov)

Bio: Dr. Vivek Agarwal is a Senior Research Scientist and Technical Lead of the Fission Battery Initiative. He specializes in crosscutting applications and advancement of inter-disciplinary research to enable resilient real-time measurement and control of process variables within the nuclear and other critical industries. Dr. Agarwal received a Bachelor of Engineering in electrical engineering from University of Madras, India in 2001, M.S. in electrical engineering from University of Tennessee, Knoxville in 2005, and Ph.D. in nuclear engineering from Purdue University in 2011. He joined INL as a research in 2011 and since then has lead projects under multiple DOE programs such as Light Water Research Sustainability, Nuclear Energy Enabling Technologies – Advanced Sensors and Instrumentation Program, Technology Transition Office, and others. Dr. Agarwal was awarded 2019 Presidential Early Career Awards for Scientists and Engineers, and 2016 Laboratory Director Early Career Achievement Award.

Country: USA
Company: Idaho National Laboratory
Job Title: Senior Research Scientist

Paper 2 AH36

Lead Author: Ahmad Al Rashdan Co-author(s): Brian Wilcken Brian.Wilcken@inl.gov Kellen Giraud Kellen.Giraud@inl.gov Svetlana Lawrence Svetlana.Lawrence@inl.gov

Using Automated Trending to Inform Nuclear Power Plant Compliance
To remain competitive in current and future energy markets, the nuclear power industry must reduce operating costs while improving operational performance through comprehensive plant modernization and by transforming the model of operations from being historically labor-centric to one that is technology-centric. Many nuclear power plant (NPP) activities could be transformed using recent advances in automation. Specifically, activities performed in an NPP to achieve or demonstrate compliance with regulatory requirements could be performed more efficiently, both from a licensee and a regulator point of view. A corrective action program (CAP) is a key component of NPP compliance and performance-tracking activities. In addition to documenting and tracking the resolution of issues that occur in NPPs, a CAP is a primary resource for monitoring the trending of issues and causes of undesirable events across a plant, fleet of plants, or even the industry as a whole. Those conditions are documented in condition reports (CRs). Thus, CRs contain invaluable information about plant performance. However, CR analysis is one of the expensive aspects of regulatory compliance, due to the manual and time-consuming review and analysis involved. Trending the occurrence frequency and causes of certain issues across a plant, fleet of plants, or even the industry as a whole can lead to insights that are of particular interest regarding detecting degraded performance (e.g., increasing occurrences of a specific issue). Such trends accumulate the knowledge gained from the occurrence of issues and causes over time to generate conclusions for areas the plant should improve, or the regulator should monitor. Significant trends can be used as a new tool to risk-inform the compliance process. Automating the trending process to detect issues/causes found in CRs can reveal hidden trends that might not be uncovered manually. To achieve this objective, natural language processing (NLP) and machine learning (ML) methods were used to mine hundreds of thousands of CRs from multiple nuclear power utilities to generate a list of topics that cumulatively can describe any CRs created by the plant. Those topics resemble the typical keyword assignment applied to documents. Once a set of topics is automatically assigned to each CR, the CRs can be filtered for a specific combination of topics in order to extract and track any trends regarding a specific issue. Such trends can be automatically analyzed so that only trends of particular interest are presented for human consideration.

Using Automated Trending to Inform Nuclear Power Plant Compliance

To remain competitive in current and future energy markets, the nuclear power industry must reduce operating costs while improving operational performance through comprehensive plant modernization and by transforming the model of operations from being historically labor-centric to one that is technology-centric. Many nuclear power plant (NPP) activities could be transformed using recent advances in automation. Specifically, activities performed in an NPP to achieve or demonstrate compliance with regulatory requirements could be performed more efficiently, both from a licensee and a regulator point of view. A corrective action program (CAP) is a key component of NPP compliance and performance-tracking activities. In addition to documenting and tracking the resolution of issues that occur in NPPs, a CAP is a primary resource for monitoring the trending of issues and causes of undesirable events across a plant, fleet of plants, or even the industry as a whole. Those conditions are documented in condition reports (CRs). Thus, CRs contain invaluable information about plant performance. However, CR analysis is one of the expensive aspects of regulatory compliance, due to the manual and time-consuming review and analysis involved. Trending the occurrence frequency and causes of certain issues across a plant, fleet of plants, or even the industry as a whole can lead to insights that are of particular interest regarding detecting degraded performance (e.g., increasing occurrences of a specific issue). Such trends accumulate the knowledge gained from the occurrence of issues and causes over time to generate conclusions for areas the plant should improve, or the regulator should monitor. Significant trends can be used as a new tool to risk-inform the compliance process. Automating the trending process to detect issues/causes found in CRs can reveal hidden trends that might not be uncovered manually. To achieve this objective, natural language processing (NLP) and machine learning (ML) methods were used to mine hundreds of thousands of CRs from multiple nuclear power utilities to generate a list of topics that cumulatively can describe any CRs created by the plant. Those topics resemble the typical keyword assignment applied to documents. Once a set of topics is automatically assigned to each CR, the CRs can be filtered for a specific combination of topics in order to extract and track any trends regarding a specific issue. Such trends can be automatically analyzed so that only trends of particular interest are presented for human consideration.

Paper AH36 | |

Name: Ahmad Al Rashdan (ahmad.alrashdan@inl.gov)

Bio: Ahmad Al Rashdan, Ph.D., is currently a senior research and development scientist at INL. Dr. Al Rashdan holds a Ph.D. in nuclear engineering from Texas A&M University, a M.Sc. in information technology and automation systems from Esslingen University of Applied Science in Germany, and a B.Sc. in mechanical engineering from Jordan University of Science and Technology in Jordan. He has more than 15 years of industrial and research experience. His experience includes automated work processes using AI methods and advanced analytics, online condition monitoring of nuclear systems, control-systems design and development, anomalies detection, and automated modeling and simulation. Dr. Al Rashdan is an active contributor to and organizer of several DOE events and scientific conferences. He authored or co-authored more than 60 technical reports and journal papers and 7 patent applications and is a senior member of the IEEE and a member of the American Nuclear Society.

Country: USA
Company: Idaho National Laboratory
Job Title: Senior Research and Development Scientist

Paper 3 KO218

Lead Author: Koushik Araseethota Manjunatha Co-author(s): Vivek Agarwal, vivek.agarwal@inl.gov Randall D. Reese, randall.reese@inl.gov

Federated Transfer Learning for Scalable Condition based Monitoring of Nuclear Power Plant Components
Condition-based monitoring (CBM) techniques are widely being adopted for maintenance activities in nuclear power plants. Asset operational data are collected by smart sensors mounted on and around the components. The sensed data is often gathered and processed by a monitoring and diagnostic center to garner various component fault signatures. These fault signatures are subsequently used as input to train predictive machine learning (ML) models for the specific component. Development of ML models require a significant amount of healthy and fault data. As faults are rare events, it is highly unlikely that all the potential fault modes are captured for a single component. Moreover, new components without historical data cannot contribute to ML model development. Additionally, fault signatures extracted from a single component cannot be robust enough to handle unseen fault patterns in same or different components. Privacy, security, legal, and commercial concerns often prevent data sharing across different plant systems. This research presents federated transfer learning (FTL) to scale ML models for CBM across components or plant systems by combining federated learning (FL) and transfer learning (TL) approaches as shown in Figure 1. FL enables developing local models at the component level across different units that are securely shared to a centralized server to aggregate into a global model. TL enables application of the developed aggregated global model to different but related systems within the same plant site, or to the same system at different plant sites. FTL was demonstrated for the circulating water system from two nuclear plant sites (representing three units) to predict the health of circulating water pumps. The FTL framework was verified using a multi-kernel adaptive support vector machine and an artificial neural network. The results show significant improvement in prediction performance while reducing over-fitting issues.

Federated Transfer Learning for Scalable Condition based Monitoring of Nuclear Power Plant Components

Condition-based monitoring (CBM) techniques are widely being adopted for maintenance activities in nuclear power plants. Asset operational data are collected by smart sensors mounted on and around the components. The sensed data is often gathered and processed by a monitoring and diagnostic center to garner various component fault signatures. These fault signatures are subsequently used as input to train predictive machine learning (ML) models for the specific component. Development of ML models require a significant amount of healthy and fault data. As faults are rare events, it is highly unlikely that all the potential fault modes are captured for a single component. Moreover, new components without historical data cannot contribute to ML model development. Additionally, fault signatures extracted from a single component cannot be robust enough to handle unseen fault patterns in same or different components. Privacy, security, legal, and commercial concerns often prevent data sharing across different plant systems. This research presents federated transfer learning (FTL) to scale ML models for CBM across components or plant systems by combining federated learning (FL) and transfer learning (TL) approaches as shown in Figure 1. FL enables developing local models at the component level across different units that are securely shared to a centralized server to aggregate into a global model. TL enables application of the developed aggregated global model to different but related systems within the same plant site, or to the same system at different plant sites. FTL was demonstrated for the circulating water system from two nuclear plant sites (representing three units) to predict the health of circulating water pumps. The FTL framework was verified using a multi-kernel adaptive support vector machine and an artificial neural network. The results show significant improvement in prediction performance while reducing over-fitting issues.

Paper KO218 | Download the paper file. |

A PSAM Profile is not yet available for this author.

Paper 4 M.110

Lead Author: Marcin Hinz Co-author(s): Doha Meslem, dmeslem.dm@gmail.com Stefan Bracke, bracke@uni-wuppertal.de

Application of Gaussian Mixed Model for the clustering of fine grinded surfaces
The optical perception of high precision, fine grinded surfaces plays a major role, especially in various consumer goods. The very complex manufacturing process of many of these products consists of a variety of parameters such as feed rate, cutting speed, grinding disc, cutting fluid, contact force or process time. The change of a parameter setting has a direct effect on the surface topography. Therefore, a standardized and optimized configuration of process parameters enables a desired quality of a product. By varying the process parameters of the high precision fine grinding process, a variety of cutlery samples with different surface topographies are manufactured. Surface topographies of grinded surfaces are measured by the use of classical methods (roughness measuring device, gloss measuring device, spectrophotometer). To improve the conventional methods, a new image processing analysis approach is needed to get a faster and more cost-effective analysis of produced surfaces. For the recognition of the product’s quality, a systematic analysis based on unsupervised learning techniques is needed. In this study, a multivariate analysis of interpreted data of knife surfaces is performed. For this purpose, three different knife types were measured in the lab and analyzed with the help of computer vision. The data is multivariate with over 40 extracted features, structured, and not full (some variables – mainly the production process data – are partially missing). Here, we discuss the theoretical and practical application of an unsupervised machine learning method on the mentioned data. Moreover, this research paper focuses on Gaussian Mixed Model, a probabilistic-based unsupervised machine learning method. This algorithm deals with soft-clusters, forming clusters and giving each datapoint certain probabilities. To create the most accurate output from an unsupervised machine learning method however, the data are clustered and subsequently compared to the already company-set criteria and to tangible knife properties such as roughness. This parameter study includes pre-processing and preparing the data using standardizing and normalizing techniques and adopting the algorithm’s parameters to monitor how this affects the clusters itself. To perform a comprehensive analysis of the algorithm’s parameters and ensure minimal discrepancy between the clustered data and classes proposed by the manufacturer, Gaussian Mixed Model parameters are tuned. This research paper therefore focuses on the study of different cluster-setups and the choice of the most accurate output based on the achieved results. The research has a generic character and can be applied to other sets of extracted data of fine grinded surfaces.

Application of Gaussian Mixed Model for the clustering of fine grinded surfaces

The optical perception of high precision, fine grinded surfaces plays a major role, especially in various consumer goods. The very complex manufacturing process of many of these products consists of a variety of parameters such as feed rate, cutting speed, grinding disc, cutting fluid, contact force or process time. The change of a parameter setting has a direct effect on the surface topography. Therefore, a standardized and optimized configuration of process parameters enables a desired quality of a product. By varying the process parameters of the high precision fine grinding process, a variety of cutlery samples with different surface topographies are manufactured. Surface topographies of grinded surfaces are measured by the use of classical methods (roughness measuring device, gloss measuring device, spectrophotometer). To improve the conventional methods, a new image processing analysis approach is needed to get a faster and more cost-effective analysis of produced surfaces. For the recognition of the product’s quality, a systematic analysis based on unsupervised learning techniques is needed. In this study, a multivariate analysis of interpreted data of knife surfaces is performed. For this purpose, three different knife types were measured in the lab and analyzed with the help of computer vision. The data is multivariate with over 40 extracted features, structured, and not full (some variables – mainly the production process data – are partially missing). Here, we discuss the theoretical and practical application of an unsupervised machine learning method on the mentioned data. Moreover, this research paper focuses on Gaussian Mixed Model, a probabilistic-based unsupervised machine learning method. This algorithm deals with soft-clusters, forming clusters and giving each datapoint certain probabilities. To create the most accurate output from an unsupervised machine learning method however, the data are clustered and subsequently compared to the already company-set criteria and to tangible knife properties such as roughness. This parameter study includes pre-processing and preparing the data using standardizing and normalizing techniques and adopting the algorithm’s parameters to monitor how this affects the clusters itself. To perform a comprehensive analysis of the algorithm’s parameters and ensure minimal discrepancy between the clustered data and classes proposed by the manufacturer, Gaussian Mixed Model parameters are tuned. This research paper therefore focuses on the study of different cluster-setups and the choice of the most accurate output based on the achieved results. The research has a generic character and can be applied to other sets of extracted data of fine grinded surfaces.

Paper M.110 | Download the paper file. |

Name: Marcin Hinz (m.hinz@uni-wuppertal.de)

Bio: 1991 – 1995 Studies of Mechanical Engineering; University of Bochum 1997 – 1999 Doctor´s Thesis: „Quality strategies regarding to the reuse of components of technical products within the product remanufacturing (product recycling)” 1996 – 1999 University of Bochum, Chair of Manufacturing System Planning, Prof. Dr.-Ing. H. Schnauber, Section: Quality Planning and Control 1996 – 2000 Consultancy INNOSYS GmbH & Co. KG; Quality Management, Bochum, Germany 2000 – 2006 Dr. Ing. h.c. F. Porsche AG, Department of Quality Management, Stuttgart, Germany 2007 – 2010 Cologne University of Applied Sciences, Cologne, GermanyFaculty of Vehicle Systems and Production, Institute of Production,Professor; Department of Quality Management and Production Metrology Since 10.2010 University of Wuppertal, Germany, Faculty of Safety EngineeringProfessorship: Chair of Reliability Engineering and Risk Analytics Since 04.2016 Guest Professorship at Meiji University, Kawasaki, Tokio, Japan

Country: DEU
Company: University of Wuppertal
Job Title: Professor