Intelligent Sensor Data Analysis Using A1 Based Techniques

Human operators, geological conditions, equipment changes, mining sequences, and roof conditions are some of the continually changing operating conditions that drive a dynamic mining environment. These dynamic characteristics make it difficult to automatically analyze mining data because attributes change as the mining environment changes. Mine monitoring systems could provide mine managers with much more useful information than they do presently, if automatic analysis were possible. Automatic analysis of sensor data is also key to autonomous mining systems. This paper describes artificial intelligence (AI) and sensor fusion based sensor data analysis techniques developed over the past four years. These techniques were used to classify machine operating events by continuously monitoring machine power usage in a continuous miner section at an under- ground coal mine. The result is a report similar to an industrial engineering time study. First, numerical pattern classification techniques broadly identify operating events. However, domain knowledge about inter-machine and intra-machine interactions is necessary to identify and refine dynamic event boundaries. By implementing AI and sensor fusion based techniques we are able to classify machine events more accurately and recognize more specific events. These techniques could easily be applied to many different mining sensor data applications