In this paper, we will develop prediction system of accident of the system by using radius basis function network (RBFN). In monitering the machine, although it is connected on line, it is sometimes difficult to store or transfer all data that it corrects. In stochastic way of doing diagnosis, they need continuous data to find mean value and standard deviation. Beside that it has no meaning when data has two steady states. However, in our system, we can express situation of steady states even if they have more than two steady states. By using actucal data, we will show the effectiveness of the proposed method.
