We have a customer in locomotive industry to reduce the maintenance cost by developing a mathematical model to monitor the health of bearings in locomotives. Learn more.
Bearing monitoring algorithm development for locomotives
The bearings of locomotives get degraded after a long run; hence they have to be replaced periodically for better performance of the locos. The objective of the project was to develop a mathematical model to monitor the health of bearings in locomotives and simulate the model to predict when unhealthy bearings in locos have to be replaced.
The engineers at QuEST Global started off by extracting the on-site measurements of the metal concentrations that wear out of the bearing. The team then did a thorough study to understand the wear out nature of different metals and correlated the required parameters to develop the model. Features from the model was generated by calculating the accumulated lead/tin which comes out of bearing and gets mixed with the oil. The team also generated the bearing health index for each loco by using curve fitting method and Artificial Neural Network Technology. Based on Bearing health index, predict the bearing which is healthy or unhealthy.
The following tools, Neural Network Tool box, Curve Fitting Tool Box, were used on Windows platform. The programming language used was Matlab
With the help of this prediction, the customer was able to reduce the maintenance cost as bearing aren’t to be replaced periodically. This also brought about improvement in performance of locomotives as manual work of finding the failed bearing is automated, as a model can produce prediction results faster and real time. Another advantage is that this model can be customized according to the different end customer requirement.