Definition of Predictive Maintenance - Predictive maintenance (PDM) compares the trend of measured physical parameters against known engineering limits for the purpose of detecting, analyzing, and correcting problems before failure occurs. A predictive approach can be applied to any equipment problem if, first, a physical parameter like vibration, temperature, pressure, voltage, current, or resistance can be measured. An engineering limit for the measured physical parameter must be established so a problem can be detected during routine monitoring. Also, the limit should be low enough to detect the problem before excessive damage occurs. Correcting of the root problem is the key to most predictive efforts.
The PDM cycle
Once a new piece of critical equipment has been added to the program and baselined, it enters the PDM cycle, figure 1.
The established parameters are measured periodically (weekly, biweekly, monthly, etc.). If the measurement exceeds the established engineering limit, it must be analyzed further. Analysis can take many forms. For example, a vibration signature can be taken on rotating equipment. A trained analyst may review the signature for common problems, such as misalignment and imbalance, as well as for not so- common problems, like resonance.
Once the source of the problem is determined, the best repair activity can be chosen. If the engineering limit is set low enough, there will still be plenty of time to correct the problem before further damage occurs. A work request is usually written to start the repair process. Correction of the root problem allows the equipment to reenter the periodic monitoring program.
The spectrum of PDM
There has been a historical misconception that equipment failures cannot be predicted. However, with predictive technology, a vast number of equipment failures can be predicted. Vibration measurement on rotating equipment is probably the best known of current predictive applications, but other categories of industrial equipment also benefit from a predictive approach. (See Table 1, "Spectrum of predictive maintenance.")
The mortality of machinery
Researchers into the reliability of equipment recognize that within a collection of machines there is a definite pattern of life spans. In
practice, this pattern manifests itself when a collection of machinery is subjected to rigorous operation. The plot of typical life spans is shown in the so-called bathtub curve, Figure 2.
Among collections of equipment, there is a rather high incidence of early failures, called infant mortalities. Most equipment that survives infancy will continue to perform with few failures occurring. In time, however, the failures begin to increase until the last of the group succumbs.
Finding the parameters
The failures that form the latter part of the curve are caused by identifiable physical phenomena. Depending upon the complexity of the machine, there may be several aging processes at work in a single piece of equipment, any of which may cause the ultimate failure. These processes are usually related to the basic physics of the materials and how the machine is used. Knowledge of the physical properties of materials comes from either theoretically or empirically derived conclusions. To understand how failures can be predicted, the mortality of machinery and the finding of parameters need to be understood.