Skip to Main Content
The main limitations of neural network (NN) methods for fault diagnostics applications are training data and data memory requirements, and computational complexity. Generally, a NN is trained offline with all the data obtained prior to commissioning, which is not possible in a practical situation. In this paper, three novel and self-commissioning training algorithms are proposed for online training of a feedforward NN to effectively address the aforesaid shortcomings. Experimental results are provided for an induction machine stator winding turn-fault detection scheme, to illustrate the feasibility of the proposed online training algorithms for implementation in a commercial product.