Preventive Condition-Directed Maintenance (PCDM) is known to be the most efficient repair policy for equipment items (parts, units, assemblies, etc.) which are subject to stochastic deterioration with increasing failure rate and can be periodically inspected. The problem is how to predict an item's residual life (until the next failure) with the acceptable accuracy and consistency.
We consider a repairable and replaceable item or a single-unit system and suppose the future operating and environmental conditions to be similar to the ones when the learning statistics was obtained. Practically, this is a reasonable assumption. An adequate deterioration model was developed and a new conditional distribution of remaining lifetime was created on its base. Expected remaining life, optimal time interval to next inspection, and other necessary parameters may be obtained using this probability function. Rather high prediction accuracy with 14% mean relative error was approved by computation tests. Monte-Carlo Simulation of the developed PCDM policy showed, that total repair and downtime cost per unit time for this strategy was 9 to 12% less than for the repair-at-failure and periodic repair policies.