Framework for Early Failure Detection of Sensor

Industry are adopting modern machinery to fulfil their expectations of production. The advancement in the technology uses complex design and multiple sensors. These machines need optimal maintenance policy to keep them properly functioning. Maintenance is systematic activity done for keeping an equipment in proper services. The condition-based maintenance relies on sensor data to establish a baseline of prediction. The sensor supplying data is assumed to be providing true data based on present condition. The sensor installed in the equipment may become faulty and not able to complete its specified lifetime. Various independent parameters, such as temperature, vibrations, EMI and variations in input supply voltages, present in the working place are responsible to early failure of sensor. The objective of this paper is to collect ground truth using supervised learning and using support vector machines model that analyses data for classification and regression analysis for providing early detection sensor failure.