Among other things, this obliges us to try to identify every failure mode that is reasonably likely to affect the functions of all the assets in our care, to understand the consequences of each failure mode, and to select the most cost-effective failure management policies.

In the absence of any comparable asset management strategy formulation processes, the only satisfactory way to do this for modern, complex industrial systems is to apply Reliability Centered Maintenance (RCM). The only truly responsible way to do so is to apply RCM correctly.

However, applying RCM correctly is time-consuming and expensive. My last article in this series (MT 4/98, pg 50) mentioned that this is leading some people to focus too heavily on the cost of strategy formulation processes like RCM rather than on what they achieve.

This search for “silver bullets” is leading to the development of shortcuts that all lead ultimately to dangerously superficial or incomplete maintenance strategies. The following paragraphs list the most common shortcuts, and reviews their main shortcomings in the light of parallels between financial and physical asset management:

• Applying RCM in reverse to existing maintenance schedules: This shortcut asks what failure modes are being prevented by existing maintenance schedules, and applies the RCM consequence assessment and task selection process only to those failure modes, without asking what other failure modes may have been over-looked by the existing schedules. This is like basing this year’s accounts solely on last year’s transactions.

• Using generic failure modes effects analyses (FMEA): this shortcut asks us to take FMEAs developed elsewhere and apply them to our own assets as part of the RCM process, on the premise that if machines are similar, then surely they will suffer from more-or-less the same failure modes. This is akin to borrowing a set of financial ledgers used by a different organization in the same line of business and using them to make our own financial decisions.

• Using generic RCM analyses: this shortcut asks us to acquire entire RCM analyses performed on similar assets used elsewhere, and apply them to our own assets. This is like basing our financial decisions on an entire set of accounts developed by another, similar business.

• Applying RCM to “critical” processes only: this is akin to asking our accountants to track only the 20 percent of our transactions which account for 80 percent of our expenses—an approach which would greatly reduce the costs of bookkeeping but which would also rapidly lead to financial chaos.

• Using computers to drive the RCM process: this shortcut suggests that it is possible to speed up the RCM decision process by computerizing it. This is akin to asking a computer to make all our investment decisions for us—a process that not even Wall Street has mastered yet. (Of course, computers are as helpful in storing and sorting the results of RCM analyses as they are for tracking financial transactions. They cannot, however, be used to make the decisions for us.)

• Using inadequately trained people to apply RCM: this shortcut entails using people with only 2 or 3 days of training—sometimes less—to lead the application of RCM to complex assets. This is usually done in the belief that any reasonably experienced maintenance person would be able to master RCM with minimal guidance. This is like suggesting that anyone with a reasonable grasp of arithmetic should be able to prepare a full set of accounts after attending a 3-day course on finance.

No sane accountant would allow shortcuts like these to be applied to financial assets. To do so would lead to chaos and eventually to ruin.

In the experience of the author, applying such shortcuts to the development of physical asset management strategies is aslo ruinous - suicidally so in some cases, if only because people develop a totally lase sense of security about assets to "a sort of" RCM.

Consultants and practitioners alike have written articles about workflow, data, information, and software, and how they must be congruent. Wherever software and work processes don’t agree, one or the other must be adjusted or the project will fall short of expectations. Considerable analysis and planning is always required. One of the biggest stresses comes from having to assign some of the best people to the project while trying to keep daily operations current.

As I listened to the stories and edited the articles, I could sympathize with the writers, but I couldn’t feel their pain—until now, but just a pinprick compared to their heart attack pain. I’m facing a microscopic desktop version of the choice between installing enterprise resources planning (ERP) software or a best of breed computerized maintenance management system (CMMS). I’m trying to decide whether to use Microsoft’s Outlook 98 (the ERP) or to upgrade my contact management software (the CMMS) so it can link with other best of breed software.

With Outlook, I get an outstanding integration of a variety of functions, but at a substantial investment of time learning the software and customizing it. With the contact manager, I get some slick features that can really speed some of my work, but I have to supplement it with other software. As I try to decide between the new and powerful and the familiar and speedy, I toy with a third option—using the application development wizards in my database manager to build my own custom solution. But first I have to look at my current job and separate the must do and should do tasks from what I used to do and what would be nice to do.

Perhaps it really doesn’t matter, for me personally or for most maintenance organizations. From what I hear, few if any maintenance departments use anywhere near the full potential of their current maintenance information software. I know my contact manager has a lot more to offer than I'm using.

Unless you have the discipline to use the software, it's not going to do much for you. On the other hand, if you have the discipline, the software becomes less important. MT

Historically, machine condition monitoring and predictive maintenance activities have been very sectarian in nature. Researchers and practicing engineers alike have maintained a one-dimensional approach to the business of machine condition assessment.

In industrial applications, especially in the power generation and petrochemical industries, vibration analysis has been the technique of choice. Conversely, in the fleet industries, oil analysis has been the technique of choice because of the preponderance of diesel engines. In general industrial applications such as primary metals, pulp and paper, etc., both have been used, but oil analysis has failed to reach its full potential.

In the industrial (nonfleet) applications of machinery condition monitoring, we have experienced in the past a split between oil analysis and other modes of condition assessment or predictive maintenance, with respect to both the responsible department and the purpose for acquiring the information. Vibration analysis and several other technologies typically have resided within the maintenance department where decisions about ensuring machinery reliability are made. Oil analysis has resided within the lubrication group or the chemistry department where decisions about making oil changes are made.

This is changing. Many plants have found significant synergistic opportunities to increase reliability and uptime when oil analysis is administered by the reliability organization and coordinated with other condition assessment activities. The following discussion explains the unique contributions of oil analysis. The mutually reinforcing roles of oil analysis and vibration analysis in assuring machinery reliability will be covered in a future article.

Oil analysis in machine condition monitoring
The role of oil analysis has a varied and inconsistent history. In the petrochemical and power generation industries, oil analysis has been conducted primarily to determine when and/or if an oil change is required. In hydraulic applications, it has been used to control the contamination that jams servo-valves and abrades components, leading to premature component failure. In fleet applications, oil analysis has been applied to determine when additives have depleted, soot is building up in the oil, fuel and/or coolant is contaminating the oil, or abnormal component wear is occurring. Each application is valid and each application provides information to support important, but varied, decisions. In sum, there are three distinct categories, or dimensions, of oil analysis:

1. Fluid health analysis—Oil analysis reveals the general health of oil. The oil’s physical, chemical,and additive properties can be measured and trended to guide decisions about if and when an oil should be changed or regenerated with an additive package. Oil analysis also identifies when the incorrect oil has been added to a system. When oil is degrading abnormally, oil analysis often can determine if the degradation is oxidative, hydrolytic, or from another root cause. In addition to simple oil change decisions, oil analysis supports decisions to change oil base-stock or additive formulation or control the environment in which the oil operates. Machines cannot run healthfully without healthy lubrication, making these decisions imperative to the reliability effort.

2. Contamination monitoring—Contamination is a leading cause of machine degradation and failure. Abrasive particles and moisture combined lead to the generation of the majority of wear in various industrial applications. Also, particles and moisture contamination strip the oil of its additives and exacerbate lubricant degradation. Contamination monitoring enables the reliability organization to make effective decisions to control this important cause of machine failure.

3. Wear debris detection and analysis—When a machine is ailing, it generates particles. The detection and analysis of wear debris assists in scheduling maintenance actions and in determining the root cause of a problem.
An effective program of oil analysis should include a focus on each of the three distinct dimensions of oil analysis. Relying upon oil analysis information simply to guide oil change decisions leaves a tremendous amount of information value on the table about the machine’s health and the interface between the machine and its environment.

Proactive control of machine health
Avoiding machinery failure should be the prime directive of the reliability organization. Once a machine is specified, designed, manufactured, installed, and deployed, there exists a finite number of variables to ensure health. Production management defines load via production schedules. The reliability and maintenance department has control over the following failure root causes that are known to lead to machine degradation:

• Machine alignment
• Machine balance
• High operating temperatures
• Lubricant health
• Lubricant contamination

Precision alignment and balance programs have proven to reduce the
occurrence of failure. Likewise, controlling lubricant quality and contamination has proven to be very effective in extending the life of mechanical components and systems. In a study by the Canadian National Research Council, contamination was found to be the leading cause of wear in a variety of industries investigated. In fact, 82 percent of all wear was found to be particle induced (see accompanying table).

The effects of contamination are slow and usually imperceptible until the late stages of failure. The result, however, is very predictable. Particle contamination or water contamination can reduce the life of mechanical components by orders of magnitude. In laboratory studies at Oklahoma State University, the British Hydromechanics Research Association, and others, component life was found to be a predictable function of lubricant contamination.

Nippon Steel reduced bearing failures by 50 percent through aggressive contamination control. International Paper’s Pine Bluff mill reported a 90 percent reduction in bearing failures through aggressive contamination control.

In another study, Alumax of South Carolina reported a per-machine reduction in component replacement costs from $15,000 per year to under $500 per year, more than a 96 percent reduction. The Alumax figures do not include the softer labor and downtime costs that always accompany a failure.

The point is that aggressive contamination control improves the reliability of mechanical equipment. But, while the proactive maintenance activities of alignment and balance monitoring assurance are typically the domain of the reliability division, the related activities of contamination control are often left out of check.

The SKF Bearing Co. states that the three “silent assumptions” of bearing life are proper alignment, proper temperatures, and contamination control. The monitoring and control of contamination should take its rightful place in the reliability assurance process and organization.

Improving decision effectiveness with oil analysis
The other principal objective of a condition-monitoring program is to improve the quality of maintenance and operations decisions. These decisions are primarily machine oriented rather than lubricant oriented. The oil carries important information about the machine in the form of wear debris. Wear debris represents the reciprocal of the machine’s surface. By analyzing the metallurgy, morphology, size, color, and relative population of different wear mechanisms, the skilled analyst can often reach the following conclusions:

• What components are wearing?
  By assessing particle metallurgy, the relative concentrations of various metals, and particle shape, or morphology, the analyst often can identify which component(s) is (are) wearing. With this information, more precise actions can be scheduled, reducing repair costs, repair time, and the occurrence of repairing healthy components.
• How severe is the situation?
  Wear particle size, shape, discoloration, and other factors lead the analyst to conclude a relative situation severity. The primary question, of course, is does the situation warrant immediate action to avoid catastrophic failure, or can it wait until a scheduled outage or downtime? This is a critical question in the operations domain. While failure prognosis is tricky at best, condition monitoring certainly gets the severity estimate into the appropriate order of magnitude to support a scheduling decision.
• What is the root cause of the problem?
  If the root cause is not identified, maintenance activities tend to resemble a broken record … the same song plays over and over again. Because wear is in fact the mirror image of the component surface, no better method exists for pinpointing the wear mechanism. With lubricant analysis and contamination analysis, the root cause for mechanical wear can be assessed with remarkable accuracy.
  

When proactive and decision support objectives are combined, oil analysis yields tremendous value through the extension of machine life and improved operations and maintenance decisions.

Oil analysis is a reliability function
Having identified the strategic importance of oil analysis for machine condition monitoring, the need to integrate it within the reliability organization, and the importance of on-site oil analysis, a tactical plan is required. While it is not feasible in many instances to install a fully capable on-site oil analysis laboratory, it is possible to implement an economic, streamlined on-site oil analysis program. As earlier stated, there are three distinct objectives in oil analysis: ensure the lubricant is fit for continued service, maintain contamination at acceptable levels, and detect and analyze abnormal wear. These objectives can be met with the following simple field tests:

• Particle count—A particle counter quantifies the amount of abrasive debris in the system. Research is conclusive that particle count and machine life are inversely related. Controlling contamination makes reliability problems disappear. Also, any generation of debris will be detected quickly by increasing particle counts, as the counter is very sensitive to change. Be sure the system reports in recognizable units (i.e., ISO 4406 Cleanliness Codes) and calibrates to known standards (i.e., ISO 4402). Also, be sure the technique is field friendly and applicable to the full range of fluids requiring analysis.
• Wear particle count—Used only as an exception tool, the ferrous particle counter quickly determines if debris is ingested (dirt) or generated (wear). Once this is determined, secondary port testing may be applied to localize the source of the debris. Rate of change analysis from frequent interval testing will help determine the severity of the situation.
• Moisture screen—Water is the scourge of hydraulic and lubricating systems. A simple hot plate “crackle” or “sputter” test can be used to determine if free or emulsified water is present. This is the test used by most labs to screen samples. If the test is negative, no chemical titration to quantify the water is performed. It is inexpensive, easy, and reliable.
• Viscosity test—Viscosity is the single most important property of the oil. It is the property that determines the fluid film thickness and the degree to which component surfaces are separated. Also, it signals any advanced stage of lubricant degradation and a “wrong oil” situation that can put the machine at significant risk.
  

These simple tests are sufficient to support the needs of the machine condition monitoring team in most instances. Further analysis may be required on an exception basis. For instance, if an increasing wear rate is occurring and the root cause cannot be effectively deduced with the simple on-site tests, the samples may be submitted to a wear debris analysis laboratory for further inspection. Additionally, occasional analysis of the fluid’s chemistry is suggested to estimate the remaining useful life of a fluid and to schedule oil changes and reconditioning. The accompanying flow diagram maps out the strategy for integrating on-site tests, the samples may be submitted to a wear debris analysis laboratory for further inspection. Additionally, occasional analysis of the fluid’s chemistry is suggested to estimate the remaining useful life of a fluid and to schedule oil changes and reconditioning. The accompanying flow diagram maps out the strategy for integrating on-site screening tests with the services provided by a full oil analysis laboratory.

When the three objectives of oil analysis are properly combined, and the program is conceived to provide the right information at the right place and time, oil analysis will improve the life of mechanical equipment and improve the quality of operations and maintenance decisions. The information generated from all condition assessment programs must be effectively combined to optimize the decision process. Analyzing machine condition is often a matter of reading between the lines. This is especially true when hunting for the root cause of a problem. Just as a carpenter goes to the job site with all the necessary tools to complete the job, the reliability technician must carry a full toolbox in his business of making and supporting effective decisions. It is clear that oil analysis is a natural ally of other machinery condition assessment technologies in the pursuit of machine reliability.

The effective integration of oil analysis with vibration analysis and other assessment technologies will be discussed in a future article. MT

Design and fabrication are critical steps in providing reliable equipment. Many frustrations could be avoided if a “crystal ball” were available to see, before acceptance, how the system will operate 6 months after bringing it on line. If it fails to meet expectations, the new equipment’s performance may be a critical parameter impacting total factory output. This is the reason to take strong action in making the right results happen.

All projects, structured or unstructured, have a defining moment when the equipment is configured and turned on for its first production run. Unstructured approaches often have many surprises and struggles to meet expectations. Even structured methods need to address the step of “Start Up—Functionality Demonstration.” A good tool to help in this situation is a testing formula involving mean time between failures (MTBF). The Reliability Quantification Test (RQT) formula came from Military Standard 781. The formula was championed throughout INTEL Corp. by Dave Troness. The company has been proactive in applying this tool in setting performance expectations with its suppliers.

RQT is my first choice in setting general acceptance standards for future suppliers of typical production equipment. Obviously, for highly critical, dangerous, or life-dependent systems, where greater than 80 percent confidence levels are required, more extensive, designed experiments should be carried out.

RQT can be used for new installations and for new subsystems proposed for existing equipment lines, which is a more-frequent occurrence for maintenance organizations. For proposed subsystems, one of the first steps is to quantify existing performance using on-going data collection. A functional block diagram of the operation is needed that has at least one block outlining the specific subsystem. MTBF information should be generated for each block. This information quantifies the current status and can help set standards for the new subsystem.

The expectations for the proposed subsystem should be part of the requirements document, and suppliers need to understand the assumptions and share the same goals. The purchase order should address the potential outcomes of the acceptance test, and the project timeline needs to account for the testing time, and possible re-testing.

The RQT is an objective way of determining, with high confidence, that the new system is acceptable, or that it needs further improvement, before impacting current operations. It is an appropriate test for the supplier to do before shipping but after “burn in.” Or, for demonstration testing, it can be done in line as long as the ability is provided to immediately go back to the previous configuration if the test fails. For multiple lines of identical equipment, RQT can determine an objective test plan for the first upgrade before migrating the changes to the rest of the lines.

The RQT equation is useful in solving for MTBF or in determining the required test time given an expected MTBF. The formula is:

MTBF = 2T/X² a2r+2
where:
T = time
X² = chi-squared

Chi-squared is a numerical value obtained from a standard chi-squared table, and can be found easily using MS Excel on a personal computer. The two parameters needed to find the proper X²value are alpha, a, and the Degrees of Freedom, DF. The recommendation is to use 80 percent as the desired confidence level (for typical production equipment); therefore a will be 0.2. DF is determined by solving the equation:

Degrees of Freedom = 2r + 2
where
r = number of failures.

Therefore, if the number of failures is 0, 1, 2, 3, ..., then the DF is 2, 4, 6, 8,...etc.
With these two parameters, start Excel and click on the function wizard (fx) button. Choose Statistical, and under the function name, select CHIINV. Click OK. The function wizard will ask for probability and Degrees of Freedom. The chi-squared value will show in the value box as soon as the a (0.2 for typical RQT) and the DF are filled in.

Two ways to use RQT
The RQT formula can provide a MTBF value from known data of the number of failures for a given test time. Suppose an existing subsystem was operated for 480 hours and it failed 7 times during that period of time. By solving the RQT formula for MTBF we would be able to say with 80 percent confidence that the MTBF for the existing system is at least 47 hours (MTBF = 2(480)/ 20.47 = 47 hours). You could then require the proposed subsystem to be that level or better so it would not significantly impact factory flow.

By properly evaluating the functional block diagram of the overall system, the influence of the subsystem could be quantified. If the new system were to be improved, the benefit for the overall system also could be estimated.

The second way to use the formula is for a test. Assume that the new subsystem is required to perform with 120 hours MTBF. To determine the testing time and failure expectations, we would first solve the RQT formula for time (T). An acceptance test table can now be developed by solving for T for various numbers of failures.

To make the test relevant, actual factory conditions need to be reproduced, and this may take creativity. Materials, environment, utilities, performance rates, waste rates, etc. need to be replicated as if the system were in production. These conditions may be difficult to replicate, especially at the supplier’s site. They need to be prioritized relative to the critical nature of the proposed system. Projects have had serious difficulty because testing was done with substitute materials and under ideal conditions. Do not discount any factor without carefully weighing the impact of that condition.

Resolve not to accept or go forward with compromised equipment. The project’s business case performance levels could be unattainable if the original design, fabrication, or assembly is not capable.

Equipment changes are part of every operating area and need to be recognized as opportunities for improvement. The RQT formula is a simple way of bringing some objectivity to bear on making the right things happen for successful transition to new equipment systems. If demonstration RQT is done with creativity, successful transitions will result. MT