Reliability engineers and their associated job functions often try hard to make their responsible supply-chain staff or purchasing departments understand that it is impossible for the maintenance department to plan, or accurately predict, spare parts demand. The reason is that a major proportion of equipment and systems fail randomly. A good example would be rolling element bearings in all industries worldwide.
According to bearing manufacturers statistics, 91% fail before they reach the end of their conservatively estimated design lives, and only 9% reach the end of design life. The millions that prematurely fail every year can all be slotted in one or more of the seven basic failure categories: 1) operator error, 2) design oversights, 3) maintenance mistake, 4) fabrication/processing defect, 5) assembly/installation error, 6)material defects and 7) operation under unintended conditions.
Experience clearly shows a preponderance of failures that could be avoided, but will continue to persist because of human error. That is also the reason for the randomness in which they occur. There is generally neither the investment in education, personnel, tools, time, or systematic grooming of a culture of knowledge and loyalty. The latter two would be needed to reduce failure frequencies and to improve the predictability of failures.
Large-scale investments have been made in predictive maintenance devices and tools that are then sitting idle or cannot be used because plant staff members have not been trained, mentored or taught to interpret reams of data. A good example would be the many computerized maintenance management systems (CMMS) which to this day, are being fed useless information such as bearing replaced, bearing failed and bearing repaired, instead of bearing failed because loose flinger ring abraded and brass chips contaminated the lube oil. Corrections made by upgrading to a clamped-on flinger disc. Of course, the prerequisite is for the facility to practice failure analysis and take remedial action, instead of merely replacing parts.
People in authority make frightfully wrong decisions. As an example, they legislate to standardize on just one type of lubricant, or to buy critically important machinery and components from the lowest bidder, or to procure replacement parts without linking them to a well thought-out specification, or to allow parts to be stocked without first thoroughly inspecting them for dimensional and material-related accuracy or specification compliance.
Random failures demand ready availability of good options.
Because of the randomness of failures in hydrocarbon processing industry (HPI) plants that are not adhering to true best-of-class concepts, it is necessary to have certain parts in stock. We know that some theoreticians extrapolate widely from data that apply to predictable wear-out failures. However, what transpires in HPI facilities is a function of numerous variables, most of which have lots to do with human error. And so, spare parts predictions determined from mathematical models are generally too far off to merit intelligent discourse. Instead, reasonable specifics are a function of equipment type, geographic location, skill levels of workforce members, etc. For decades, it has been understood that relevant answers require auditing or reviewing a particular local situation.
Some years ago, one of our books or articles gave the following statistics for oil and petrochemical plants:
a) 25% of all failures are preventable but not prevented
b) 15% of all failures are predictable but not predicted
c) 20% of all failures are predicted but not acted upon to undertake repair
d) 25% of all failures are predicted and machines stopped to do repairs
e) 15% of all failures are neither preventable nor predictable.
Taking into consideration that these statistics were generated about 15 years ago, and also realizing that certain advancements have been made since the early 1990s, we were recently asked if an update would be available. It was pointed out that new and more powerful predictive technology, perhaps more precise diagnostics, and more (or sometimes less) reliable equipment and parts might have caused a shift in these percentages. If so, what are the new numbers?
After reflecting on the issue, we believe some of todays probable statistics are still close to the ones published 15 years ago. However, some definitions or updated numbers may be helpful.
a) 25% of all failures are preventable but not prevented because of an arbitrary decision that is simply not rooted in knowledge and experience. Example: Use the cheap oil may overlook the fact that the cheap oil lacked demulsifiers or anti-foaming agents, etc.
b) 15% of all failures are predictable but not predicted. Example: The random appearance of black oil is attributable to O-ring degradation of a certain style of bearing protector seal. The bearings will soon fail, but nobody has read the books and articles that describe the occurrence. The occurrence should be linked to a certain risky design feature on a widely used product.1
c) 20% of all failures are predicted but not stopped to undertake repair. Chances are someone in authority overruled an expert who asked for a shutdown when vibration increased beyond a safe level.
d) 25% of all failures are predicted and equipment is shut down for repair. Good; everybody is happy. Unfortunately, the organizations energy is funneled into restorative maintenance efforts instead of proactive upgrade efforts that would prevent failures in the first place. We believe prevention is better than spending money for restoration.
e) Only 1% of all failures are neither preventable nor predictable. As of 2010, we changed our minds and no longer believe the old 15% figure was correct. Human beings make the decision to build cities in earthquake zoneseither with suitable building codes, or by disregarding such codes. Strong levies can be built or not built, maintained or not maintained. But, yes, some machines might fail because a neighboring pressure vessel exploded or because a fire in another unit spreads. These may indeed fit the neither preventable nor predictable 1% category. The remaining 14% belong in the other categories.
As to spare parts, our recommendation is to be informed on not only the supply capabilities of original equipment manufacturers (OEMs) but to also becoming familiar with the competence and response times offered by non-OEM vendors. The best of these both repair and systematically upgrade machines during a maintenance event. Being on their mailing list for relevant bulletins (Fig. 1) and understanding their reverse-engineering capabilities will prove even more helpful in cases where spare parts cannot be quickly obtained from the OEM. HP
| Fig. 1. Informative bulletins inform us of non-|
OEM capabilities for rapid repairs.
1 Bloch, H. P., Getting all the facts is more important than ever, Hydrocarbon Processing, May 2009, p. 9.
|The author |
Heinz P. Bloch is Hydrocarbon Processings Reliability/Equipment Editor. A practicing consulting engineer with close to 50 years of applicable experience, he advises process plants worldwide on failure analysis, reliability improvement and maintenance cost avoidance topics. He has authored or co-authored 18 textbooks on machinery reliability improvement and over 490 papers or articles dealing with related subjects.