The focus of this study is the development of a credible fault detection system of the cylindrical plunge grinding process. The acoustic emission (AE) signals generated during machining were analyzed to determine the relationship between grinding-related faults and characteristics of changes in signals. Furthermore, a neural network, which has excellent ability in pattern classification, was applied to the diagnosis system. The neural network was optimized with a momentum coefficient, a learning rate, and a structure of the hidden layer in the iterative learning process. The success rates of fault detection were verified.