Automatic Cycle Border Detection for a Statistic Evaluation of the Loading Process of Earth-moving Vehicles 2007-01-4191

In the earth-moving industry manymachines work in typical loading cycles that are repeated periodically. For a statistic examination of the overall load configuration and the dynamic fatigue of these machines, it is necessary to develop an adaptive algorithm for the separation of the individual cycles. This article presents methods for an automatic detection of the cycle borders. Adaptive algorithms are constructed for a reliable separation at different points during the loading cycle. Additionally, each cycle can be divided into different operating phases by extending the algorithms to a tool for the identification of each single phase. To avoid problems during the cycle detection, the data are checked for outliers and sensor faults first. To guarantee a meaningful statistical analysis, the separated cycles have to be tested for incorrect or atypical characteristics. Therefore, statistical classification numbers are calculated and compared for each cycle. Furthermore, the cross-correlation function between the data and an average loading cycle is considered. Robustness with respect to outliers is especially addressed and suitable procedures are incorporated. The presented algorithms have been tested on several measured data sets of different machine-types and -configurations as well as different drivers. These experimental results prove that an excellent and reliable performance can be achieved.