| Multi-Class Classification for Ocean Turbine State Detection | ||||
| Author | Janell Duhaney
|
|||
| Co-Author(s) | Taghi M. Khoshgoftaar
|
|||
| Abstract | In this study, we utilize machine learning algorithms to automate ocean turbine state detection. Our research team has previously studied the question of applying data mining and machine learning techniques to determine if there was a fault present (i.e., the output is one of two states: normal or abnormal). In the present work, we extend those efforts to determine which of multiple states the system is currently in. Although the algorithms used here are by no means novel, their application to ocean turbine condition monitoring (as is done in this paper) separates this from other works. The case studies presented here show the results of applying five well-known machine learning algorithms to classifying experimental data gathered from six sensors mounted on the ocean turbine’s dynamometer test bed. Our findings reveal that when the data from all six sensors are combined using a feature level fusion technique, the 5-Nearest Neighbor algorithm can identify the ocean turbine’s state with 100% accuracy using just pre-processed vibration sensor data.
|
|||
| Keywords | machine learning, reliability, state detection, ocean turbine | |||
| Article #: 19112 | ||||
August 5-7, 2013 - Honolulu, Hawaii, U.S.A. |