A real-time predictive maintenance system for machine systems - An alternative to expensive motion sensing technology

The machine system parameters can provide a large amount of useful information about the machine condition, maintenance requirement and productivity. Until now, the machine parameters are estimated using crude motion technology supported with motion sensing techniques. However, the motion sensing technology is expensive and difficult to maintain. Furthermore, the sensing technology is still not mature enough to estimate machine system parameters of a complex inter-dependent motion axis. The need for a better low-cost technique for machine parameter estimation motivates the development of a real-time predictive maintenance system. This paper describes a neural network based novel real-time predictive maintenance system for machine systems. The neural network approach is used for the prediction of the machine system parameters using the motion current signature. Unlike many neural network based condition monitoring systems, this approach is validated in an off-line proof of concept procedure, using data from an experimental test rig similar to a production machine. The experiment aimed to classify five distinct motor loads using the motion current signature, irrespective of changing tuning parameters. Comparison of the predicted and actual loads showed good agreement. The success of the proof of concept motivates the development of the real-time predictive maintenance system with additional parameters like friction and external torque. An application of the real-time predictive maintenance system to a clinical micro-drilling process, for example a stapedotomy procedure, is presented in this paper. The drilling in such processes is controlled using the drilling inertia. Therefore, the real-time predictive maintenance system is used to estimate the drilling inertia from the motor current signature, hence providing a mechanism of manipulating the drilling process in a stapedotomy procedure by the surgeons.