Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 341 - Change-Point Models in Quality Control Applications
Type: Topic-Contributed
Date/Time: Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
Sponsor: Quality and Productivity Section
Abstract #317578
Title: Change-point methods in disturbance identification and probabilistic labeling of events
Author(s): Emmanuel Yashchin* and Nianjun Zhou and Anuradha Bhamidipaty
Companies: IBM Research and IBM Research and IBM Research
Keywords: Control charts; Detection; Machine Learning; Monitoring; Segmentation; Run length

Environmental disturbances often cause failure or malfunction of assets and related outage events. However, it is quite common that the failures cannot be identified as being caused by a disturbance based on the data, due to the limited information available at the time of data compilation, time constraints, or personnel's insufficient training. The ability to label an outage event reliably as one caused by a disturbance is a key pre-requisite for analytic activities such as risk modeling, outage detection, prediction and management. Change-point methods play an important role in this process, enabling efficient identification of disturbances and establishment of temporal boundaries. We introduce a methodology for disturbance identification and illustrate its use in conjunction with complex processes governing weather-related outages, which include handling spatio-temporal effects, and outliers. We also discuss the use of this methodology for probabilistic labeling of outage service tickets.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2021 program