MYTH: A well-trained technician can predict within a window of a few hours when a machine will fail.
TRUTH: This myth is common throughout industry and poses a danger to machinery. The myth often leads to operating machinery with known defects for periods longer than is safe for the equipment. Let’s explore the elements that brought about this myth and why it remains so prevalent.
One reason for the birth of this myth is the curve fitting plots found in many PDM software programs. Figure 1 shows such plot. Those not trained in the use of the software may think that the time projected for the condition to reach the alarm level indicates “time-to-failure”. In reality, the projected time is “time-to-alarm”. The machine may run quite sometime after exceeding an alarm. These plots are very valuable if used as intended.
Another reason for this pervasive myth is that published PF curves are almost always shown as downward exponential curves. Figure 2 shows a typical curve. If PF curves really are this shape, finding the “time-to-failure” would simply be a mathematical process. However, PF curves are seldom this shape because many variables may influence the shape of the curve. Figure 3 shows how a real curve may look and also presents some reasons why the shape is erratic. PF curves may have hundreds or thousands of shapes.
Even the best-trained technicians, using the best tools can’t possibly know all variables that may lead to component or machine failure. Load, speed, temperature, and environment are only a few of the many variables that may affect “time-to-failure” for a given defect. A well trained technician can detect defects and may even be able to state with a fair amount of confidence that a machine will operate or will not operate until the very first opportunity to take it out of service for repairs. Doing otherwise and operating with known defects is tantamount to rolling the dice and gambling with your machinery.