In distributed database systems materialized views are useful to solve many problems caused by data replication. This paper deals with the problem of updating materialized views incluing join operations. We use a differential update which uses logs that record the change of base tables in certain periods. This method avoids locking of the base tables so that it makes the system more efficient. We update materialized views periodically to save the updating cost envolved. A screen test is applied to differential files to eliminate tuples that are irrelevant to any of the views being updated. Using these methods, a detailed procedure is addressed to refresh materialized views. Then we show that our update procedure performs better than a semijoin approach.