Tutorials

Abstract
Over the past decade, a large and diverse set of battery prognostics techniques have been developed for forecasting the capacity and predicting the remaining useful life of lithium-ion batteries. The past two years has seen increasing research efforts developing and validating machine learning algorithms and models for early life prediction. The idea is to predict battery cycle life using early-cycle data where battery cells show negligible capacity degradation. This tutorial will review past and ongoing research studies on battery capacity forecasting and early life prediction and discuss the long-term testing and methodology development efforts led by a team of researchers at Iowa State University.

Abstract
Ever-increasing qualities and quantities of data are routinely collected across a wide range of applications that offer a rich source of information. The developments in wearable sensors, smart home technologies, and the Internet of Things provide industries with opportunities to monitor patients’ health and machines’ health with the power of AI without significant disruption to their everyday activities. Analysing the large dataset, sometimes real-time collected data poses emerging challenges in data science as the data can have substantial artefacts; the dataset might be highly imbalanced and incomplete, and might contain high levels of variability. Moreover, data labelling can be expensive and time-consuming if there is an insufficient number of high-quality labels. In this tutorial, we will introduce several machine learning techniques to tackle these issues and provide robust solutions. Innovations arising from clinical science will be used to demonstrate how deep learning and probabilistic can facilitate rapid clinical intervention, transform a hospital-only treatment pathway into a cost-effective, home-based, combined alternative, and thus improve the overall quality of patient healthcare. The machine learning methods presented in these case studies can be implemented in other industrial domains that face similar challenges, such as tracking the health condition of machinery and predicting machine deterioration for maintenance scheduling.

Abstract
Ever-increasing qualities and quantities of data are routinely collected across a wide range of applications that offer a rich source of information. The developments in wearable sensors, smart home technologies, and the Internet of Things provide industries with opportunities to monitor patients’ health and machines’ health with the power of AI without significant disruption to their everyday activities. Analysing the large dataset, sometimes real-time collected data poses emerging challenges in data science as the data can have substantial artefacts; the dataset might be highly imbalanced and incomplete, and might contain high levels of variability. Moreover, data labelling can be expensive and time-consuming if there is an insufficient number of high-quality labels. In this tutorial, we will introduce several machine learning techniques to tackle these issues and provide robust solutions. Innovations arising from clinical science will be used to demonstrate how deep learning and probabilistic can facilitate rapid clinical intervention, transform a hospital-only treatment pathway into a cost-effective, home-based, combined alternative, and thus improve the overall quality of patient healthcare. The machine learning methods presented in these case studies can be implemented in other industrial domains that face similar challenges, such as tracking the health condition of machinery and predicting machine deterioration for maintenance scheduling.

Abstract
Mobility is a core technology with internet of things and artificial intelligence for the 4th industrial revolution. The mobility is also effective in the research field of prognostics and health management (PHM) in the sense that this technology not only provides an easy way to approach any facility of interest, but also replaces human’s duty in a dangerous environment for condition monitoring and fault detections. This tutorial presents the mobility toward PHM with smart inspection systems for overhung and underground power transmission lines. Inspection of power transmission lines at severe environments would be a good example to demonstrate effectiveness of mobility because power transmission lines are wide-spread, located at a variety of regions, and suffer from severe environments. A field demonstration of these platforms confirm inherent advantages of the mobility.

Abstract
Digital twin is a core technology to prognostic and health management (PHM). Model driven digital twin may provide more information of physical assets than data driven digital twin, but it requires much more computational burden. To handle this issue, many CAE software with twin builder (e.g. ANSYS, Siemens, etc) include various reduced-order modeling (ROM) techniques for linear and nonlinear systems. The tutorial presents an effective model driven digital twin framework of vibro-acoustic interaction problem with multiphysics model reduction. A real-time virtual sensor of fluid-filled pipe implemented by the framework will be also introduced. The virtual sensor can estimate the unmeasured time transient responses of a vibro-acoustic structure in real time. The tutorial includes a theoretically review of vibro-acoustic model reduction with its example codes.