PSO-BASED ONLINE SEQUENTIAL EXTREME LEARNING MACHINE FOR CLASSIFICATION AND ENGINEERING APPLICATION

摘要

Online sequential extreme learning machine (OS-ELM) is a method of online learning, which can handle the problem of data arriving or chunk-by-chunk with varying chunk size. However, the input weights (connections between the input and hidden layers) and the hidden layer biases of OS-ELM are randomly determined, which will cause an inaccurate classification result, that varied a lot, and in the field of fault diagnosis, the diagnostic mode is mainly divided into two types: online and offline. At present, many offline learning and diagnostic methods have been proposed by researchers, but there are few online sequential learning and monitoring methods. Therefore, in this paper, a new classification method PSO-OSELM (Particle Swarm Optimization and Online Sequential Extreme Learning Machine) is proposed for classification and engineering application. In the PSO-OSELM method, PSO is used to determine the optimal input weights and thresholds of OS-ELM, which minimizing the classification error of OS-ELM. Contrastive experiments between PSO-OSELM and OS-ELM have been conducted on wine datasets. The results show that PSO-OSELM has higher classification accuracy than OS-ELM, and PSO-OSELM is further applied to diagnose the faults of roller bearing. It can be seen from the training and testing results that the fault classification accuracy of PSO-OSELM is better than OS-ELM.