Data-driven Crisis Management
Way Kuo
City University of Hong Kong
Abstract:
Since the outbreak of Covid-19 at the end of 2019, little consolidated effort has been taken to deal with the epidemic worldwide. Misleading information are around. We list losses from recent major pandemic outbreaks and nuclear power plant accidents, together with air pollution, traffic accidents and suicides. Early simulations of possible scenarios by the Emergency Response Unit can prevent the negative impact on the general public, and hence reduce global losses.
An Analytic Tool Box for Optimizing CBM Decisions
Andrew K.S. Jardine
Department of Mechanical & Industrial Engineering
University of Toronto
Abstract
Engineers and asset managers associated with Operations and Maintenance (O&M) have to make many difficult decisions, and this presentation addresses one of them: the important maintenance tactic of condition-based maintenance (CBM). Historically, companies either waited until a piece of equipment failed before repairing or replacing it, or simply guessed at a good time to perform maintenance and hopefully avoid failure. With CBM, the guesswork is largely eliminated because equipment is closely monitored. Empirical proof of a change in condition now guides maintenance decisions. The problem now is an overabundance of information. With the fourth industrial revolution, Industry 4.0, data come from everywhere, and everything is linked to everything else. ISO 55001 stresses in Section 8.2.3: “The organization should have the capability to make evidence-based decisions on proposed changes and the ability to consider scenarios systematically across the entire organization.” This is all well and good, but if we want to make maintenance decisions based on evidence, which data are most relevant?
The focus of the presentation is evidence-based asset management (EBAM) and its application to CBM decisions. It illustrates the value of applying analytics to big data gathered by numerous condition monitoring technologies (oil sampling, vibration monitoring, pressure, temperature etc.) to ensure that an evidence-based decision is made and that the decision will, in fact, optimize the condition based maintenance decisions.
The methodology has been successfully applied in numerous sectors, such as the military, mining, transportation, pulp and paper, petrochemicals, food processing, and electricity generation.
Reliability in the 21st Century
William Q. Meeker
Department of Statistics
Center for Nondestructive Evaluation
Iowa State University
Ames, Iowa 50010
wqmeeker@iastate.edu
Abstract
Reliability is an engineering discipline that relies heavily on the application of probability and statistics. Changes in sensor, communications, and storage technologies are changing the nature of reliability field data. An increasing number of modern systems are being outfitted with sensors that capture information about how and when and under what environmental and operating conditions individual systems are being used. In some cases, the physical/chemical state of critical system components can also be quantified and reported. For many systems such information is being downloaded continuously into data farms. In addition, improvements in computing capabilities and investment in developing physics-based models for failure provide another important dimension of reliability information. There are many potential applications for using such data to improve safety and reduce costs but existing statistical methods for reliability assessment and prediction are inadequate for the tasks. This talk reviews some particular applications where the modern field reliability data are used and explores some of the opportunities to use modern reliability data to provide stronger statistical/physical methods that can be used to operate and predict the performance of systems in the field. We also provide some examples of recent technical developments designed to be used in such applications and outline remaining challenges.
Key words: Condition-based maintenance, Dynamic covariates, Materials state awareness, Prognostics, Structural health monitoring
Condition Monitoring and Fault Diagnosis of Technical Processes for
Reliability Enhancement with Applications to Nuclear Power Plants
Jing Jiang
Department of Electrical & Computer Engineering
University of Western Ontario
London, Ont. Canada
N6A 5B9
Abstract
To enhance the reliability of safety-critical technical systems or related equipment, condition monitoring and fault diagnosis techniques are often used. However, these techniques can be characterized as an ‘inverse’ problem, where one can only infer the condition of the underlying process through acquired external behaviours, often in the form of measurements. A distinctive feature in such inverse problems is that the solutions may not be unique, i.e. multiple causes can lead to a similar observation. This makes such a task very challenging. To assess the true condition of a technical process effectively and to diagnose any crucial faults, one has to increase the dimensionality of the observation space. Advanced signal processing techniques, models of the technical process, and emerging technologies, such as IoTs, digital twins, can help to increase the solvability of this problem by exploring causal relationships among different variables.
This talk will focus on the following aspects: (1) nonlinear relationships between the condition of a technical process being monitored and the observed behaviours through measurements; (2) existing signal processing methods and data analytics techniques to capture the genuine behaviours of the technical processes using models and measurement data; (3) advances in technologies that can gain insights of technical processes, that were not previously feasible; and (4) some application examples of the above developed techniques in condition monitoring and fault diagnosis in nuclear power plants.
More specifically, the talk will examine the relationships between the reliability and condition monitoring and fault diagnosis for technical systems. The process of condition monitoring and fault diagnosis is then described from an ‘inverse’ problem perspective. Explanation of why dimensionality of the observation data is critical in finding unique solutions to this problem. Different signal processing techniques, including time-frequency analysis, wavelets, and data fusion, will be described to extract key information, while removing noise or artifacts. Recent advancements in sensing technologies with distributed fibre optic sensors, printable electronics, no-touching eddy current probe arrays, as well as IoTs and digital twins have provided revolutionary means to acquire relevant information about the processes with no or minimal influence to its inherent operation. Some of the latest research results and demonstration examples will also be presented.