Building Computing Environment for Predictive Analytics in Data-Driven Application Systems

Predictive maintenance improves productivity, product quality and safety in modern enterprises. It is based on the use of data to determine the current state of the equipment and make decisions about the necessary actions. Implementation predictive maintenance solution includes the decision to organize data storage, the definition of the preferred computing architecture. In addition, it is important to correctly set the task and determine the expected effect of the implementation. Today more and more factories are implementing predictive maintenance. Even a small percentage of the positive effect reduces costs and helps factories work more efficiently. The main aim of this master thesis is to build an anomaly monitoring system for industrial robotic system. Creating of the solution includes a description of the architecture, development of a machine learning algorithm for the detection of anomalies and development alert system prototype for enterprises staff. Moreover, this research takes a deeper look at machine learning techniques for predictive maintenance and principles of implementation such solutions in smart factory infrastructure.