A lot of repeated work existed in the design of manufacturing process for similar parts, and many kinds of process knowledge which were not reused effectively contained in the historical process data. In these conditions, in order to reuse the typical process effectively and enhance design efficiency, typical process discovery was studied. The problem of typical process discovery was described firstly, the attributed directed graph was used to model the process. And the similarity between processes could be measured with the similarity between process cells and the similarity between process routes by the process model. Based on the process similarity, affinity propagation method was used to cluster the process. For the purpose of getting the effective clustering results and making sure the best number of clusters, the Silhouette index and the in-group proportion index were separately adopted in the clustering analysis. Finally, we have experimented the validity of the discovery algorithm by clustering typical process of machining process for satellite plate.