Predicting asset degradation in industrial motors

Maintenance and repairs constitutes a large part of the production costs in the oil and gas industry. Companies active in the North Sea area has seen a dramatic increase in maintenance costs duringthe last ten years. One of the reasons being aging equipment due to unwillingness for new investments. Thus many assets are being usedfar beyond their original design life, requiring more costly maintenance and repairs. This has made better and more cost efficient maintenance strategies a key factor for companies to remain competitive. This thesis investigates whether it is possible to implement more efficient, predictive maintenance strategies, using the availablehardˇware and sensor systems. Data was collected from currentlyavailable sensor systems and was used to predict the winding temperatures of an electric motor. The winding temperature was chosen since previous research show that the winding temperature of an electric motor tend to increase with degradation.The model proposed in this thesis was able to predict the motors winding temperatures with an average absolute error of 0.188 degrees, when predicting on unseen data. The thesis also showed that service and repairs to the motor introduced a prediction bias of 1.65degrees, providing evidence that changes in average prediction error can be used as an indicator of asset degradation.Tryckt av: UppsalaDetta examensarbete är sekretessbelagt enligt avtal mellan projektägare, student ochUppsala universitet.IT 2