Machine learning: approaches to predicting reliability and developing maintenance strategies

Abstract. Current approaches to maintenance of rolling stock bogies are focused on compliance wear limits as stipulated by OEM specifications. recommendations critical providing an industry wide approach safety and compliance. These not operation specific often the most cost-effective solutions. A system reliability is established that applied in less complex systems where relationships between components well defined with historical data predictable conditions. Extending this more multi-variate many intuitively obvious or mathematically presents a challenge. Machine learning techniques have been address such problems examples image recognition, tool prediction using multiple sensory inputs estimating railway bogie vibration inputs. [8,9,10] The aim study extend adapt machine-learning area developing strategies for optimal business benefit focus maintenance. This aims present insight into variables, which includes tracking condition affecting track side rate. finding in-depth each independent variable’s individual impact necessary step efficient modelling. include geometry, operating component variables. Track wear, curve radius, superelevation top variance found be significant predictors predictions consistent different rail tracks exhaustive. Specifically, performance requires inclusion. Combining these variables statistical inference convolutional theory maximum likelihood estimators would establish predictor rate operation. paper finally discusses relationship influences grinding frequency material type.