Condition Monitoring by Model-of-Signals: Application to gearbox lubrication

In this work, we make use of the Model-of-Signal technique to perform lubrication monitoring of a large industrial worm gear motor. We assume sensor measurements to be modelled by autoregressive processes and exploit the edge-computing capabilities of programmable logic controllers to perform the Recursive Least Squares algorithm to identify them. Then, we use those models to compute indicators able to diagnose the lubricant level within the gearbox and compare them to statistical indexes, which are traditionally used for monitoring. The aim of this application is to show how to build a condition monitoring infrastructure in an industrial environment able to detect possible occurring faults locally and acquire knowledge about them by exchanging information with external computers, paving the way towards Intelligent Maintenance Systems in Industry 4.0.