The Networking Method as a Tool for Prediction of Failures of Open Pit Mine Trucks

The authors present a practical problem solving scheme for the prediction and control of mining machine condition using real data from sensors tracking basic parameters of open pit mine trucks. The network analysis of signals from 15 sensors is performed for two systems of Komatsu HD1500-8 dump truck: internal combustion engine and gear shift transmission. The network nodes are selected to be the time series data on one operation shift of the machine. Connections between the time series were found on the basis of similarity of temporal sequences. The similarity between temporal sequences was estimated using the Dynamic Time Warping (DTW). The networks were constructed using a threshold model, with a threshold set as the time-series similarity. It is found that some networks respond to a machine failure on the day of its occasion, the day before and the day after it. The results prove applicability of the network analysis for the timely detection of mining machine failures.