Abstract: For the ships equipment fault diagnosis problems with lack of applicability and accuracy during ships sailing, this paper designs a radical basis function neural network method for ships equipment fault diagnosis. An improved artificial bee colony (IABC) algorithm combining opposite learning initialization strategy and auto-adapted search strategy is designed, which builds higher quality initial solution space through opposite learning initialization strategy and adapts its local search length automatically to improve the ability of convergence and local optimization searching. IABC algorithm is used in parameter optimization of radical basis function neural network (RBFNN) for constructing a better performed classifier. The results show that the IABC-RBFNN framework can improve the accuracy and usability of ship fault diagnosis, and satisfy the real-time requirement of ships equipment fault diagnosis.

Key words: ship equipment fault diagnosis, radical basis function neural network, artificial bee colony algorithm, opposite learning strategy, auto-adapted strategy