The technology uprising in the premises of the 4th industrial revolution has led to the modernization of the maintenance field and the migration from preventive to predictive maintenance through machine learning methods and techniques. This diploma thesis aims, through research of classical and state of the art algorithms in the timeseries anomaly detection and classification domain, to the development of a user friendly and accurate tool of fault identification. To achieve this, it is essential to research for the most suitable machine learning techniques and consequently implement, adjust and evaluate their results in a real industrial environment.