oracle never accepts the inconsistency of the database. so when oracle goes down due to power outage or aborted manually, next time once the database base is opened it goes mainly to the three states nomount,mount,and open when oracle try to open the database it makes check of SCN in the controlfile and each database datafile header if the check is not correct it will try to perform crash recovery and apply the online redo logs to bring the database in consistent state. By any means if it could not apply then it will try to locate the archive logs to apply and asks for media recovery.