Abstract: A radial basis process neural networks model was proposed, which is a kind of three-layer forward structure constituted of input layer, radial basis function hidden layer and output layer. The transformation from input layer to hidden layer is nonlinear and that from hidden layer to output layer is linear. The neurons of hidden layer perform the pattern matching of process input information and aggregation operation of time and respond to the input patterns. Through inducting function orthogonal basis into input space, input function can be expanded under the orthogonal basis and aggregation operation process can be simplified by using the orthogonality of basis function. The corresponding learning algorithms were given and the effectiveness of this method was proved by rotation machinery fault diagnoses.