Intelligent Condition Monitoring Solutions for Manufacturing Machines of Wind Turbine Gearbox Components

Driven by increasing competition in the renewable energy market, reducing the Levelized Cost of Energy (LCoE) has become crucial. Traditional condition monitoring methods are inadequate for production machines due to the large size of components and the limited number in wind turbine manufacturing. This thesis focuses on developing intelligent condition monitoring solutions for manufacturing machines used in the production of gearbox components of wind turbines. By leveraging machine learning, this research aims to develop an unsupervised approach that captures complex temporal patterns in process signals. This involves creating algorithms to monitor operational states, detect anomalies, and identify maintenance needs without pre-labeled data. The goal is to enhance the machine availability, ensuring reliable operation with minimal downtime, ultimately reducing lead times (i.e., ensuring that the machinery operates efficiently and without unexpected delays) and production costs. The study also explores the identification of key vibration sensors, automatic processing for constant and non-constant segments, and improvements in clustering techniques to support this advanced monitoring approach.