First, there is the (NO) maintenance approach. This is essentially "run until done" style. Kind of like never changing the oil in the car engine or replacing tires and expecting infinite life. If lucky, the damage that occurs is mild and repairable (at relatively high cost) - if unlucky, complete replacement of machinery required (EXTREMELY expensive proposition, since it also has to include potentially lengthy delay in acquiring new equipment and unanticipated loss-of-production)
Second, there is the (SCHEDULED) approach. Here, there is a set time frame for doing what is believed to be required to look after the equipment. May not cover all the bases in terms of actual preventive requirements, and may not happen on a cycle that catches trouble before it becomes one of the first cases. Loss-of-production can be minimized since the maintenance event occurs at a known moment.
Third case is the (PLANNED) maintenance. This is when something is known to be an issue, but other factors have to be considered to find the time / effort to achieve the goal. For example - considerations should include minimization of loss-of-production, minimization of in/out costs, having parts readily available before the unit is taken offline, having qualified personnel available to do the work. As a contrast - the "NO" maintenance approach could also be titled "UNPLANNED", since there's never a convenient time for a catastrophic event.
Fourth option is (PREDICTIVE) maintenance. This is usually taken to mean an examination of the machine - both in operation and during shutdown periods. Specific non-intrusive and non-destructive tests can be performed that lead to a trending capability to determine overall equipment health, and thus the likelihood of trouble before the next PLANNED or SCHEDULED maintenance event.