System maintenance for reliable running of key machinery is critical to many industries, where condition monitoring and fault diagnosis is important supporting technology. This paper selects a typical component in rotating machinery, the gearbox, as the target to study a proper monitoring and fault diagnosis method to prevent malfunction and failure. The failure is divided into two levels. One is at the component level that includes various gear faults, and another is at system level that studies machinery statuses include looseness, misalignment and unbalance. A prototype system is built for experiment. Two intelligent methods include artificial neural network (ANN) and genetic algorithms (GAs) are combined to carry out signal classification and analysis. ANNs are one of the common machine learning technologies that used for detecting and diagnosing faults in rotating machinery. To look for a feasible combined solution, this paper tests the effect of back-propagation (BP) network and GAs are used in this paper for selecting the significant input features in a large set of possible features in machine condition monitoring with vibration signals. Considering the performance of machine learning system are hard to predict, and the quality of input signal is a major factor affecting the performance of training and learning of the system itself. Signal preprocessing is executed through feature extraction by wavelet packet transforms (WPT) technology and time domains statistical analysis to generate statistic variables for analysis. With an aim to identify a proper diagnosis approach, the effect of BP network and GAs are verified with case studies. © 2011 IEEE.