Automatic Recognition Of Adhesion States Using An Extreme Learning Machine

In this study, the use of an extreme learning machine (ELM) for automatic identification of the adhesion state is investigated. The influence of different activation functions and the number of neurons in the hidden layers on the recognition performance is investigated. This study aims to select a better activation function and construct an approach for the adhesion state recognition by using an ELM. Monitoring data on the adhesion characteristics of a heavy-duty locomotive are used to fabricate the recognition model. Comparing with the backpropagation (BP) neural network and a BP optimized algorithm, the experimental results show that our ELM recognition method has a faster training speed and higher recognition accuracy than the other two approaches.