Unsupervised Anomaly detection

Abstract

The only way for the world to move into the bright future is to move from nonrenewable resources into renewable ones. Creating and maintaining new economic spheres always requires human care and supervision, the magnitude of which can be lowered by using machine learning techniques. This work demonstrates the models that are created to solve the task of anomaly detection in an unsupervised fashion. This kind of methodology imposes a lot fewer restrictions on the data used while providing a framework to find cracks on a wind turbine. Moreover, it is a good building block for later research of unsupervised anomaly detection in the fields, where getting data might cost a lot, and the cost of mistakes is high. The success of the work can reduce the amount of time, money, and human resources for the big companies that utilize green energy and invest in the future of our planet.