Abstract
Technological advancements such as the industrial Internet of Things now allow companies to continuously monitor the operating conditions of expensive equipment using sensors. With the tremendous amount of sensor data flowing in continuously, equipment makers are seeking innovative analytical solutions to turn operational data to help guide their tactical and strategic decisions. Using sensor data on wind turbine operations and service records from a top Fortune 100 company in the energy industry, we showcase techniques to map out operational-level data for analysis, and develop several analytical models (a sequence analysis, a logistic regression and a survival model) to help predict and evaluate equipment failure risks. Our analyses highlight the significant value propositions of sensor data in the big data era. Practical implications as well as extensions of the proposed predictive models are discussed.
Acknowledgement
The corresponding author would like to thank the anonymous Fortune 100 company for providing the wind turbine data, and the business sponsors for offering their helpful insights in understanding the data and interpreting the results.
Notes
1 While some of the attributes may not be intuitive to understand, we spent a significant amount of time researching each of the variables with field engineers in the wind energy industry.
2 It is noted that one size (length of time window) does not fit all; in practice, various time window lengths should be tried. The results of the sequence analysis presented below are based on a 6-month time window.