Since the measurement of oil quantity in aviation engine lubrication systems is susceptible to multiple parameters， it is difficult to calculate and predict oil consumption rates. We propose an improved method for extracting oil quantity data and predicting oil consumption rates. Engine data are cleansed using density-based clustering algorithms， including density-based spatial clustering of applications with noise （DBSCAN）， to obtain stable oil quantity data during steady flight conditions. By the least squares method， oil consumption rates are derived with an average fitting goodness of 0.86. Subsequently， a multi-layer perception （MLP） is employed to establish the relationship between oil consumption rates and flight status parameters， resulting in a predicted average absolute percentage error of 1.15%， compared to actual values. The proposed method meets practical engineering requirements， providing a reliable reference for assessing the health status of aviation engine lubrication systems.