An Extension Neural Network and Genetic Algorithm for Bearing Fault Classification

A genetic algorithm enhanced extension neural network (GA-ENN) is presented which improves on the traditional ENN by including the automatic determination of the learning rate. The GA allows the best network that produces the lowest classification error to be obtained. The effectiveness of this new system is proven using the Iris dataset. The system is then applied to the problem of bearing condition monitoring, where vibration data from bearings are analysed, diagnosed as faulty or not and their severity classified. This system is found to be 100% accurate in detecting bearing faults with an accuracy of 95% in diagnosing the severity of the fault.