In the industrial field, a variety of machines are connected. In particular, a motor, that is composed of a bearing, a core, a winding and etc., one of most important component an actuator system. The life of motors depends on the condition of the bearings. Therefore, we developed an algorithm to diagnose bearing condition to predict the life of the motor. In this algorithm, a learning model is constructed by converting vibration data into learning data using a three-axis accelerometer and is composed of learning data that show the highest accuracy with low number of data. In this study, It decreases diagnosis time and load of interface board. Therefore, it is possible to reduce the cost of system.