Process mining is to discover structured process description from real execution data. It helps the discovery and design of business process, and improves the existent ones through delta analysis. One of the challenging problems in process mining is how to deal with duplicate tasks. This paper provides a duplicate tasks treatment stage before the real execution of mining algorithm, which method is well compatible with existent process mining algorithms and helps them dealing with duplicate tasks. In addition, this paper designs a distance measure to transfer the difference of event context into numerical form, and take advantage of such distance to distinguish duplicate tasks through clustering technology. The method in this paper is proved by experiments on typical process model having duplicate tasks.