Tutorials

Description: The main challenge of PHM is to predict the remaining useful life of components (RUL).  This is basically a three-step process. The first step is the detection of the fault, including the identification of the failure mode.  The second step is the assessment of the fault severity. The third step is the prediction of the RUL. Predicting the RUL requires a good understanding of the fault propagation mechanism for the specific failure mode, assessment of the fault severity trend, and a definition of the critical point that marks the end of life (the critical fault severity). The tutorial will focus on the second step – the assessment of fault severity, in the context of rotating machines. For rotating components, the assessment of the fault severity and the determination of the critical severity are extremely challenging.

Two main approaches are mostly investigated. The first approach involves techniques that rely on the physical understanding of the changes in the vibration signature, with respect to the fault severity. The second approach encompasses a range of data-driven methods, including statistical modeling, supervised learning, and unsupervised learning. Lately some studies propose hybrid methods that take advantage of the strengths of both approaches. The tutorial will present and compare the different approaches and their possible application in research of PHM, focusing on rotating components. The presentation will explain and illustrate the important role of physical models in different research phases, the development of new Condition Indicators that can bring an essential enlightenment to the task of prognostics in mechanical systems, as well as the possibilities to incorporate machine learning algorithms in hybrid methods.

Speaker Bio:

Dr. Renata Klein received her Ph.D. in the field of Signal Processing from the Technion, Israel Institute of Technology. For 17 years she managed the Vibration Analysis department in ADA-Rafael, the Israeli Armament Development Authority. Later, as the Chief Scientist in RSL- Electronics, she invented and led the development of a vibration based diagnostics and prognostics system that is used successfully in combat helicopters and UAVs of the Israeli Air Force.  Renata is the CEO and owner of R.K. Diagnostics. In this role, and per invitation from Safran Aircraft Engines, she developed a full set of vibration based diagnostics and prognostics algorithms for jet engines.  These algorithms are being integrated into the next generation of CFM jet engines. In recent years, Renata has focused on supervising academic research programs in the area of rotating machinery prognostics. Jointly with Prof. Jacob Bortman, she co-manages the BGU PHM Lab in Ben Gurion University of the Negev, teaches and provides supervision to MSc and PhD students.

Description: Technical Language is highly contextual communications found in industry, academia, and other scientific domains where terms may have very different use or implications than more common natural language. This tutorial introduces participants to Technical Language Processing (TLP), an engineering-oriented approach that leverages AI tools, including Natural Language Processing (NLP), for effective processing of technical language data. Central components of TLP will be covered, starting with engineering use cases related to Prognostics and Health Management (PHM). The tutorial will provide an overview of fundamental concepts from mainstream NLP, including NLP pipelines, pre-processing techniques, and word representation methods with clear emphasis on the applications and risks of using them in TLP. This tutorial will also highlight existing tools and procedures that aid in cleaning, processing, and analyzing information contained within technical documents of various sizes. We will conclude with TLP Q&A addressing successes, challenges, and next steps for the community.

Speaker Bios:

Dr. Sarah Lukens is a Data Scientist at LMI. Her interests are focused on data-driven modeling for reliability applications by combining modern data science techniques with current industry performance data. This work involves analyzing asset maintenance data and creating statistical models that support asset performance management (APM) work processes using components from natural language processing, machine learning, and reliability engineering. Sarah completed her Ph.D. in mathematics in 2010 from Tulane University with focus on scientific computing and numerical analysis. Sarah is a Certified Maintenance and Reliability Professional (CMRP).

Dr. Michael E. Sharp is a Reliability Engineer at the National Institute of Standards and Technology (NIST) located in Gaithersburg, MD. He received a B.S (2007), M.S. (2009), and Ph.D. (2012) in Nuclear Engineering from the University of Tennessee, Knoxville, TN, USA. His research interests include signal analytics, machine learning, artificial intelligence, optimization, and natural language processing. Michael has worked on a wide array of projects including image processing for elemental material recognition, navel reliability monitoring, and manufacturing robotics diagnostic monitoring. He currently works as a project lead on Industrial Artificial Intelligence Management and Metrology (IAIMM) at NIST in the Communications Technology Laboratory.

Description: TBD

Speaker Bio: TBD