Data processing methods and solutions for IoT-based cyber-physical systems

OData support
Dr. Varga Pál
Department of Telecommunications and Media Informatics

Industrial systems are under constant utilization, often in a demanding environment, therefore wear-outs have a likely occurrence. High availability is expected from these systems, while costs must still be kept low. Traditionally, the maintenance process happens in a corrective fashion, meaning that any kind of intervention only happens after the component (or the whole system) broke down. In contrast to this, proactive maintenance solutions are gaining popularity, which aim to have the intervention before the breakdown.

Realization of such a solution is non-trivial, but can be done with already existing technologies. Developing a proactive maintenance solution for a whole network consisting of Cyber-Physical Systems (CPS) can be done with the help of cheap commercial sensors and embedded systems, which are already in place to monitor the systems. The task then is to unite the data flows from these devices in a local automation cloud based solution. Such clouds should guarantee real-time processing while meeting strict, industrial grade security requirements, and provide the needed resources for processing and analyzing of the huge amount of inbound data. A platform like this should give failure forecasts and, for example, predict remaining useful life (RUL) for various equipment components. Based on this, maintenance related tasks can be properly scheduled then, which helps reducing costs.

MANTIS [1] is a European project which aims to solve this challenging task. It plans to provide a proactive maintenance platform for systems consisting of CPSs. To achieve this, we need to create a highly flexible and efficient data processing methodology. Great amount of sensory data, good data preprocessing and well-applied machine learning algorithms are necessary for achieving this goal.

In this thesis, I give a brief overview of the most used paradigms in industrial automation and then discuss the structure and goals of the MANTIS project. The main topic of this thesis is to present an industrial forklift-related use case for condition based maintenance (CBM). In here, I describe in detail my proceedings in the offline data processing phase. Finally, I analyze the achieved results and lay out future work.

[1] Official MANTIS site:


Please sign in to download the files of this thesis.