PSAM 16 - Session T15 Overview

PSAM 16 Conference Session T15 Overview

Session Chair: Yail Kim (jimmy.kim@ucdenver.edu)

Paper 1 NI33

Lead Author: Nicola Tamascelli Co-author(s): Antonio Javier Nakhal Akel, Riccardo Patriarca, Nicola Paltrinieri, Ana Maria Cruz

Are we going towards "no-brainer" risk management? A case study on climate hazards
Industry is stepping into its 4.0 phase by implementing and increasingly relaying on cyber-technological systems. Wider networks of sensors may allow for continuous monitoring of industrial process conditions. Enhanced computational power provides the capability of processing the collected “big data”. Early warnings can then be picked and lead to suggestion for proactive safety strategies, or directly initiate the action of autonomous actuators ensuring the required level of system safety. But have we reached these safety 4.0 promises yet, or will we ever reach them? A traditional view on safety defines it as the absence of accidents and incidents. A forward-looking perspective on safety affirms that it involves ensuring that “as many things as possible go right”. However, in both the views there is an element of uncertainty associated to the prediction of future risks and, more subtle, to the capability of possessing all the necessary information for such prediction. This uncertainty does not simply disappear once we apply advanced artificial intelligence (AI) techniques to the infinite series of possible accident scenarios, but it can be found behind modelling choices and parameters setting. In a nutshell, “there ain't no such thing as a free lunch”, i.e. any model claiming superior flexibility usually introduce extra assumptions. This contribution will illustrate a series of examples where AI techniques are used to continuously update the evaluation of the safety level in an industrial system. This will allow us to affirm with certain confidence that we are not even close to a “no-brainer” condition in which the responsibility for human and system safety is entirely moved to the machine. However, this shows that such advanced techniques are progressively providing a reliable support for critical decision making and guiding industry towards more risk-informed and safety-responsible planning.

Are we going towards "no-brainer" risk management? A case study on climate hazards

Industry is stepping into its 4.0 phase by implementing and increasingly relaying on cyber-technological systems. Wider networks of sensors may allow for continuous monitoring of industrial process conditions. Enhanced computational power provides the capability of processing the collected “big data”. Early warnings can then be picked and lead to suggestion for proactive safety strategies, or directly initiate the action of autonomous actuators ensuring the required level of system safety. But have we reached these safety 4.0 promises yet, or will we ever reach them? A traditional view on safety defines it as the absence of accidents and incidents. A forward-looking perspective on safety affirms that it involves ensuring that “as many things as possible go right”. However, in both the views there is an element of uncertainty associated to the prediction of future risks and, more subtle, to the capability of possessing all the necessary information for such prediction. This uncertainty does not simply disappear once we apply advanced artificial intelligence (AI) techniques to the infinite series of possible accident scenarios, but it can be found behind modelling choices and parameters setting. In a nutshell, “there ain't no such thing as a free lunch”, i.e. any model claiming superior flexibility usually introduce extra assumptions. This contribution will illustrate a series of examples where AI techniques are used to continuously update the evaluation of the safety level in an industrial system. This will allow us to affirm with certain confidence that we are not even close to a “no-brainer” condition in which the responsibility for human and system safety is entirely moved to the machine. However, this shows that such advanced techniques are progressively providing a reliable support for critical decision making and guiding industry towards more risk-informed and safety-responsible planning.

Paper NI33 | Download the paper file. | Download the presentation PowerPoint file.

Name: Nicola Tamascelli (nicola.tamascelli@ntnu.no)

Bio: I am a second-year Ph.D. student enrolled in a jointly-supervised doctorate at the Norwegian University of Science and Technology (Norway) and The University of Bologna (Italy). My research focuses on the development of Machine Learning algorithms to support a dynamic and proactive approach to process safety. My interests include the analysis and monitoring of industrial alarm systems, the development of classification algorithms to predict the consequences of major accidents, the implementation of regression algorithms for prognostic purposes, and the integration between data-driven simulation models and risk analysis techniques to improve environmental risk management in the Waste-to-Energy industry.

Country: ITA
Company: Norwegian University of Science and Technology
Job Title: Ph.D. student

Paper 2 DI140

Lead Author: Diego Andrés Aichele Figueroa Co-author(s): Lavínia Maria Mendes Araújo - lavinia.mendes@ufpe.br Thais Campos Lucas - thaiscamposlucas@hotmail.com Marcio das Chagas Moura - marcio.cmoura@ufpe.br Isis Didier Lins - isis.lins@ufpe.br Enrique Lopez Droguett - eald@g.ucla.edu

Diagnosis of Failure Modes from Bearing Data via Deep Learning Variational Autoencoder Method
Bearings and gears are indispensable equipment in complex machinery. Many studies developed analyses to improve the effectiveness of predictive maintenance for these components. Thus, Deep Learning (DL) models for the diagnosis and prognosis of equipment failure modes can be highlighted. For this purpose, many applications have used supervised learning methods, such as Support Vector Machine, Multilayer Perceptron, and Convolutional neural networks. However, in practice, labeled data connected to the conditions of real-time systems can be more complex and costly to obtain. In this sense, we highlight unsupervised learning models, where the algorithm discovers by itself through data exploration, the possible relationships between data points. Hence, this paper aims to apply the unsupervised Variational Autoencoder method to diagnose failure modes of bearings and gears. Six databases available in the literature will be used for analyses purposes. Nevertheless, optimization of the model's hyperparameters is aggregated to perform more efficient assessments. Finally, the results will be compared with other methods to validate the model's effectiveness.

Diagnosis of Failure Modes from Bearing Data via Deep Learning Variational Autoencoder Method

Bearings and gears are indispensable equipment in complex machinery. Many studies developed analyses to improve the effectiveness of predictive maintenance for these components. Thus, Deep Learning (DL) models for the diagnosis and prognosis of equipment failure modes can be highlighted. For this purpose, many applications have used supervised learning methods, such as Support Vector Machine, Multilayer Perceptron, and Convolutional neural networks. However, in practice, labeled data connected to the conditions of real-time systems can be more complex and costly to obtain. In this sense, we highlight unsupervised learning models, where the algorithm discovers by itself through data exploration, the possible relationships between data points. Hence, this paper aims to apply the unsupervised Variational Autoencoder method to diagnose failure modes of bearings and gears. Six databases available in the literature will be used for analyses purposes. Nevertheless, optimization of the model's hyperparameters is aggregated to perform more efficient assessments. Finally, the results will be compared with other methods to validate the model's effectiveness.

Paper DI140 | Download the paper file. | Download the presentation PowerPoint file.

A PSAM Profile is not yet available for this author.

Paper 3 SI295

Lead Author: Sizarta Sarshar

Towards utilization of digital technology for railway infrastructure
The Norwegian railway is being modernized and emerging technologies allow for a digital transformation to improve efficiency and safety. New digital enablers include among others fast and high bandwidth communication systems, new sensors, and internet of things (IoT) solutions, use of artificial intelligence (AI) and digital twins to support e.g., automated train operations and smart asset management. To better understand some of the needs and challenges with introducing the digital enabler digital twins in the railway domain in Norway, a pre-project was established by the Norwegian Railway Directorate and carried out by the research institute IFE and infrastructure owner Bane NOR. The intention was to identify gaps where research is needed. Two use cases were studied to explore a holistic approach for utilizing digital twins where relevant end users were interviewed to identify needs and gaps: (1) Maintenance and inspection of a bridge which relate to asset management, and (2) Train operations which relate to planning and dispatching train movements This paper discusses the findings considering “what is the acceptable degradation” for utilizing the technologies in operation.

Towards utilization of digital technology for railway infrastructure

The Norwegian railway is being modernized and emerging technologies allow for a digital transformation to improve efficiency and safety. New digital enablers include among others fast and high bandwidth communication systems, new sensors, and internet of things (IoT) solutions, use of artificial intelligence (AI) and digital twins to support e.g., automated train operations and smart asset management. To better understand some of the needs and challenges with introducing the digital enabler digital twins in the railway domain in Norway, a pre-project was established by the Norwegian Railway Directorate and carried out by the research institute IFE and infrastructure owner Bane NOR. The intention was to identify gaps where research is needed. Two use cases were studied to explore a holistic approach for utilizing digital twins where relevant end users were interviewed to identify needs and gaps: (1) Maintenance and inspection of a bridge which relate to asset management, and (2) Train operations which relate to planning and dispatching train movements This paper discusses the findings considering “what is the acceptable degradation” for utilizing the technologies in operation.

Paper SI295 | Download the paper file. | Download the presentation PowerPoint file.

Name: Sizarta Sarshar (sizarta.sarshar@ife.no)

Bio: Dr. Sizarta Sarshar is research manager at Institute for Energy Technology (IFE) in Norway for the department Humans and Automation. He has worked with risk, safety and security the last 15 years and has PhD within the same topic.

Country: NOR
Company: Institute for Energy Technology (IFE)
Job Title: Research Manager

Paper 4 CY266

Lead Author: Young Ho Chae Co-author(s): Hyeonmin Kim, hyeonmin@kaeri.re.kr Poong Hyun Seong, phseong1@kaist.ac.kr

Development of a physics informed neural network based simulation methodology for DPSA
A nuclear power plant is a safety-critical system with large size and high complexity. Therefore, various methods were developed to identify possible accidents and deal with them. To broadly classify the methods, there are experiment-based methods and simulation-based methods. However, the experiment-based method, in reality, has several limitations. Therefore, various simulation-based analysis methods were developed. Most of the simulation-based analysis methods were highly dependent on numerical methods. Therefore, if the number of nodes and time units are divided to increase the analysis resolution, the time required for calculation tends to increase exponentially as the number of nodes is divided. Therefore, in this paper, to accelerate the simulation we developed artificial intelligence based simulation acceleration method. As an algorithm for AI-based simulation, physics informed neural network algorithm is implemented for convergence speed and extrapolation robustness. By using the suggested method, dynamic event tree based dynamic probabilistic safety assessment can be conducted which is almost impossible due to the calculation speed of the physical process model.

Development of a physics informed neural network based simulation methodology for DPSA

A nuclear power plant is a safety-critical system with large size and high complexity. Therefore, various methods were developed to identify possible accidents and deal with them. To broadly classify the methods, there are experiment-based methods and simulation-based methods. However, the experiment-based method, in reality, has several limitations. Therefore, various simulation-based analysis methods were developed. Most of the simulation-based analysis methods were highly dependent on numerical methods. Therefore, if the number of nodes and time units are divided to increase the analysis resolution, the time required for calculation tends to increase exponentially as the number of nodes is divided. Therefore, in this paper, to accelerate the simulation we developed artificial intelligence based simulation acceleration method. As an algorithm for AI-based simulation, physics informed neural network algorithm is implemented for convergence speed and extrapolation robustness. By using the suggested method, dynamic event tree based dynamic probabilistic safety assessment can be conducted which is almost impossible due to the calculation speed of the physical process model.

Paper CY266 | Download the paper file. |

Name: Young Ho Chae (cyhproto@kaist.ac.kr)

Bio:

Country: KOR
Company: KAIST
Job Title: Researcher (Ph.D Candidate)