Manufacturing analysis and data acquisition in advanced machining

Abstract

To cut cost of maintenance, being able to stop the machines at the right time before fault and the possibility to implement zero defect manufacturing is an important part of manufacturing aeroplane engine parts. The purpose of this report is to test a statistical method on different data sources of GKN Aerospace Norway, and Sweden to see what kind of available dataset is best suited to prevent high maintenance cost and identify faults in the machine. The data sources are based on calibration data of a probe and temperature from a Carnaghi vertical turning lathe machine at GKN Aerospace Norway, and vibration and runout data from the spindle on a GROB milling machine at GKN Aerospace Sweden. Using Principal component analysis and Mahalanobis distance to analyse these datasets will give a better picture of what kind of data to use when implementing condition-based maintenance or optimising fault recognition. After testing both datasets, the dataset available from the GROB machine made it possible to see when the spindle of the machine slowly started to deteriorate before a fault happened. This was possible since the dataset had a known fault. The calibration dataset shows it is possible to identify deviations from normal calibration, and may make it easier to analyse deviation on later calibration. The method is not implemented into either of the sites, but this report may give more background for further work.