Condition-Invariant Fault Diagnosis for the Resource Constrained Edge: A Retrofitting Use Case

Industry 4.0, is a data driven revolution where mechatronic systems and in turn industries are pushed to perform at higher efficiencies. Cloud based knowledge base has become the new norm of this so known industrial revolution, which makes machines connected to the internet imperative. With multitude of machines that are to be retrofitted and almost every machine having to push huge data onto the internet, data load on the networks is rigorously increasing. Latency is another issue that is of great focus considering many applications in cloud based industrial intelligence and automation. Even the best of protocols and newest of the network technologies could fall short of giving the right flexibility. An efficient way of retrofitting the old machines in these case scenarios is with a self-intelligent system at the edge which reduces the micromanagement of the sensory data by cloud. In the context of this Doctoral research, I would like to develop a system and its architecture to retrofit machines. The goal is to provide guidelines and methodologies to develop scalable, robust and adaptive solutions for retrofitting legacy machines towards Industry 4.0 standards. The so developed system will accommodate a certain level of machine specific intelligence at the edge to process the raw data and capable of raising machine critical alerts required for the automation right there and then. Events and aggregated data that are required by the operational intelligence based on the cloud will be shared based on the requirements. Thus reducing data that needs to be put onto the cloud and decreasing the machine specific latency while elevating the performance of the whole system.