With the proliferating of Web services, how to efficiently and correctly recommend the appropriate service for process designer have attracted wide attentions from both industry and academia. The traditional service recommendation with the service interface semantic matching may not work properly, that is, the recommended service will be unusable or unavailable. In order to solve this problem, a new service recommendation approach with case-based reasoning (CBR) is proposed for the service composition system in this paper. Instead of acquiring complete ontology and semantic information, this method mined service processes designed before to obtain frequent process sequences for the service recommendation. Furthermore, the architecture of the service recommendation system with CBR is presented and the experiment is conducted. The experimental result shows the effectiveness of our approach.