In this paper, a real-time detecting strategy for outliers and abnormal fluctuation during the data acquisition in length sensors measurement is investigated. The appearance of outliers and abnormal fluctuation could lower the accuracy of the measurement system using the accurate noncontact sensors. A mathematic model consists of the real value, the noise as well as the outliers is built to represent the final observation result of the length sensors used for measuring the machining quantity. Then, different kinds of the characteristics of numerical data are analysed for the real-time effectiveness for outliers detecting and abnormal fluctuation assessment. Based on the real-time characteristic of the acquired data, a detecting strategy is designed to eliminate the outlier and judge the fluctuating extent of the data. The simulation results illustrate the applicability and the effectiveness of the proposed approach.