Qualifying Evaluations from Human Operators: Integrating Sensor Data with Natural Language Logs
Michael Brundage, Michael Sharp, radu pavel
Even in the increasingly connected world of smart manufacturing and the Industrial Internet Of Things (IIOT), there will always be a need for human operators and evaluations. When creating equipment condition monitoring models and heuristics, the observations from human operators are often difficult to quantify or track. This situation can lead to the observations being underutilized, misunderstood, or ignored completely if autonomous sensors are employed. This work seeks to highlight the untapped potential for augmenting numeric data from sensors and control systems with human input and vice versa, by integrating documented natural language reports with data collection technology in a novel and intuitive way. This is a first step experiment and seeks primarily to establish a link between human-generated data and sensor driven information in order to motivate, justify, and guide future endeavors. This is an exploratory work that utilizes an experimental setup with a limited and controlled accelerated aging setup where human observations were recorded at regular intervals alongside streaming sensor data. The goal is to validate the relationship between observers' natural language, quantified sensed values, and some ground truth knowledge about the state of the tool. Recommendations for follow-on work and extensions of the performed the analysis are provided as part of a gaps and a next steps outline.
6th European Conference of the Prognostics and Health Management Society 2021
, Sharp, M.
and Pavel, R.
Qualifying Evaluations from Human Operators: Integrating Sensor Data with Natural Language Logs, 6th European Conference of the Prognostics and Health Management Society 2021 , Turin, , [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=932274
(Accessed September 18, 2021)