Determination of Hidden Unit Number in a BP Neural Network

Abstract

BP neural network is one of the most widely used methods in fault diagnosis.As the number of hidden units in a BP neural network is difficult to determine,this paper uses comparative analysis to analyze the existing empirical formulas that determine the number of hidden units,and proposes an improved method.The method is applied to wind turbine gearbox fault diagnosis.The result shows that the improved method can effectively speed up the search for the optimal number of hidden units.