Any item or system in a plant is usually designed to have a performance capability that is safely above that required to meet the duty demands. Consequently, there is some margin over which the performance can deteriorate before the process duty fails to be completed.
The deterioration may be relatively slow and uniform, say for fouling a heat exchanger or maybe precipitously fast, for example, in the case of machine bearings under overload conditions. Most equipment undergoes periods of higher and lower deterioration as the plant loadings change, but the critical issue for predictive maintenance is the detection accuracy of the monitoring method.
The combination of this accuracy and the rate of deterioration set the warning time for predicting the failure. The interval between monitoring events must be shorter than the warning time, or the loss will have occurred before it has been detected! In addition, the warning time must be adequate to allow considered action to be taken to prevent the failure, either by “correcting” the operating conditions or executing a controlled shut-down.