Intelligent monitoring of a pneumatic robot leg system by neural networks that can automatically adapt

This paper describes the selection of key parameters to charactersise a pneumatic robot system and the use of a neural network to classify changes to each key parameter. As systems can be characterized by parameters that can be monitored over time, once initialised with random experimental data then changes to the network reflect changes in each key parameter and therefore changes to the monitored system. A decision tree is then used to diagnose problems by analysing the results for the monitored parameters.