I'd do this with a stored procedure. Open a cursor (use a select statement which orders the results), read through these results, insert each record into a new temp table unless the values have not changed from the last insert, in which case ignore them. This will be more performant than using a distinct, if the results are ordered the only comparison that is required is the last record, not every record in the table. I can't help you with the syntax of the procedure I'm afraid (I have almost no experience with DB2) but that's the logic I'd use.
Maximilian ofers some very good advice:
Optional - add primary key to old table. I would highly recommend this as to prevent duplicate records from coming back.
I'd add to this by saying its not optional. An entity without a primary key is not valid relational data, so has no place in your database.