Approximate quality was classified by calculating data collected from different points of a diesel engine in different precision,and the most sens itive status information collectingpoint was acquired by further calculation of approximate dependence on condition attributes and decision attributes in each collecting point.The most efficient detecting point for work status in fault diagnosis system was found by processing the data from collecting points with the theory of variable precision rough set. Collection of efficient information could decrease the noise pollution on fault diagnosis to the maximum extent, which improves the work efficiency and accuracy of the fault diagnosis system.