Deep Neural Network Hard Parameter Multi-Task Learning for Condition Monitoring of an Offshore Wind Turbine

Abstract Breaking the curse of small datasets in machine learning is but one major challenges that cause several real-life prediction problems. In offshore wind application, for instance, this issue presents when monitoring an asset attempt to reduce its infant mortality failures. Another challenge could emerge reducing number sensors installed order limit investment systems. To tackle these issues, aim article investigate impact data-set on conventional methods, and outline improvement achievable by implementation transfer approach. It provides a solution mitigate applying hard parameter multi-task approach supervisory control data acquisition from operational turbine, allowing smaller efficiently predict status gearbox’s vibration data. Two experiments are carried out paper. The first envisage possibility using two turbines. second compare results model findings deep neural network trained single turbine.