The quantitative prediction model of the lubricating oil moisture was established by using the extreme learning machine method,in which KennardStone method was used to divide the sample so as to reduce the working of modeling and improve the modeling speed.The quantitative prediction of the lubricating oil moisture for a special vehicle was carried out by the extreme learning machine method,and the predicted result was compared with that by the partial least square method and BP neural network method.Results show that the model based on the extreme learning machine method is more steady and the predicted result is more accuracy.The extreme learning machine method can be used as a rapid detection method of the lubricating oil moisture.