Wear debris found in gear lubricating oil provides extremely valuable information on the nature and severity of gear faults as well as remaining gear life. The conventional off-line process of taking samples of oil for testing of wear debris is a hindrance because it is laborious, expensive, delays information collection, and is expert oriented. In view of these limitations, the development of automating wear debris particle analysis using various approaches has been ongoing for years. However, existing online technology does not encourage widespread use of wear debris analysis (WDA) in the industry. High costs coupled with expert and labour requirements have led users to use other types of condition-based maintenance, such as vibration. There is a need to develop a WDA technique that is relatively cheap, online, requires little expertise to handle, and provides more information for maintenance decision-making. This PhD thesis proposes a WDA technique which uses image processing and three-dimensional image reconstruction to diagnose the health of machinery. Its emphasis is on using the thickness and volume of the particles generated over time to predict the onset of gearbox failure, so that maintenance action can be taken before gears reach catastrophic failure.