Presented as a Society of Tribologists and Lubrication Engineers paper at the World Tribology Congress in London, United Kingdom, September 8-12,1997
Final manuscript approved April 14,1997 1 a PETER G. BALL (Member, STLE) Machine Reliability Services (A' Asia) Pty Ltd. Toowoomba, Queensland 4350, Australia
An effective machine wear control program integrates the use of lubricant, vibration, thermal and performance analyses. This is because it is economically unacceptable to shut down expensive production equipment due to tribological failures. There is considerable concern that many industries appear to be uncertain regarding these issues.
The strategy for integrated machine wear control should be to use only those tests that are relevant. Acquisition of reliable and trendable data is vital to this concept and its management. Machine wear control incorporating spectrometry, fluid condition, diagnostic ferrography, lubricant contaminant monitoring, vibration, thermography and performance analysis allows a rational approach to failure prevention. When incorporated with failure potential analysis, it can often provide a safe interchange from scheduled to on condition maintenance.
This paper outlines the most significant analysis methods and provides basic guidelines for their integration. It also explains how this approach can provide maximum benefits, if undertaken correctly.
KEY WORDS
Condition Monitoring, Ferrography, Friction, Failure Analysis, Lubricant Analysis, Optical Microscopy, Predictive Maintenance, Thermography, Tribology, Vibration Analysis, Wear Particles
INTRODUCTION
Managers generally have a clear notion of the term reliable. It is associated with the behaviour of employees and their dependability. The employee is expected to do the right thing at all times, work consistently at all times, complete tasks on time and produce accurate and timely reports.
Managerial outlook toward machinery is somewhat similar, and a reliable machine or process is considered to be basically sound if it is to be able to meet its design specification and give trouble-free performance in a given environment.
In the real world, managers (engineers included) must clearly understand that nothing is perfect, that all machine plant life is finite, and that the very best equipment will eventually fail. There is a tendency for management to become preoccupied with machine availability calculations and neglect the fact that reliability provides availability.
The British Standards Institution definition for mechanical reliability is the probability that a component, device, or system will perform its prescribed duty with out failure for a given time when operated correctly in a specified environment.
There is a general abhorrence of the term "breakdown maintenance." The concept of planned, preventive maintenance is well supported because it complies with logical management practices and can be easily budgeted.
Predictive maintenance and/or condition-based management policies are still subject to misunderstanding. The recent innovation of proactive maintenance appears to be achieving some measure of acceptance; however, operations management continues to remain easy prey for the vendor-initiated philosophy of scheduled component replacement.
The proactive approach aims at avoiding the underlying, conditions that lead to machine faults and system degradation. Unlike predictive/ preventive maintenance, proactive maintenance initiates actions directed at failure root causes, and not just symptoms.
TRIBOLOGY
Data released by the Strategic Industry Foundation indicates annual maintenance costs in Australia are between 30- 50% higher than European and Japanese companies and 15% higher than their North American counterparts. This fact could be roughly assessed as being due to a lack of basic understanding of machinery wear characteristics, more simply described by the term tribology, the science and technology of interacting surfaces in relative motion, and of related subjects and practices (2).
The field of tribology is well positioned to provide the technology for management of maintenance and, at the same time, coordinate the prime requirements of affordable reliability and high equipment availability.
The costs of wear involve not only replacement parts but also lost production and opportunity through unscheduled downtime. During the mid-1980s, these costs to the Australian economy were conservatively estimated at around six percent of the Gross National Product, and by the year 2000 it is estimated that good tribological practice will save Australian industry close to $3 billion per annum.
In the current economic climate, Australian management will, by the use of condition-based machine reliability practices, access these potential savings. An improved national fiscal bottom line will reflect this pragmatic approach.
PROBLEM AREAS AND FAILURE MODES
There are two distinct classes of engineering failure: intermittent and permanent. The permanent type can be subdivided into complete failure and partial failure, with further classification into sudden failure and gradual failure, which combined can be either catastrophic (sudden or complete) or degradation (partial and/or gradual).
In an attempt to conduct an investigation into system or component failure it is often found that there are as many explanations for the failure as there are commentators.
A typical review of machinery failures would reveal that most failures are due to fracture, excessive deformation and surface failure, particularly corrosion deterioration. These can be categorized as corrosion, contamination, fatigue, overheating, overstressing, seizure and wear (3).
A realistic definition of failure analysis would be that it is the logical, systematic examination of an item or its design, to identify and analyse the probability, causes, and consequences of real, or potential malfunction.
The techniques employed by condition-based monitoring for the purpose of failure analysis are many and varied. They include failure modes, effects, and criticality analysis, fault tree analysis, image analysis, probability density charting and statistical analyses such as Pareto, Weibull, Pearson, etc. Each has its function and, on many occasions, none of them would be employed, experience and common sense being more appropriate to the situation.
The detection of incipient failure with the results and predictive analysis trends from ferrography, vibration and performance monitoring, together with oil elemental and condition analysis plays an important role. Considerable use is made, for illustrative purposes, of the well- known "bath tub" hazard rate curve, which relates failure rate and time in accordance with the three classical stages of failure: early, random, and wear-out.
RISK, ROOT CAUSE AND FAILURE POTENTIAL ANALYSIS
Some failure analysts concerned with reliability, risk and root cause analysis would insist that the most opportune time to eliminate the potential for machine failure is on the drawing board.
There is a school of thought that suggests that too much effort is wasted in prevention methods such as oil filtration and that the time and effort would be better utilized in influencing the machine manufacturer to construct to suit the environment. As admirable as these approaches may be, there is a real world out there and we live in it.
It is wise to use failure potential analysis whenever experience and gut feeling, say that something could go wrong in the future and the cost will be considerable.
Four activity techniques (4) that will get one started are as follow.
- Identification of vulnerable areas within one's sphere of responsibilities.
- Identification of specific potential problems that could have sufficient negative effect on the operation to merit taking action now.
- Identification of the likely causes of these potential problems and identification of actions to prevent them from occurring, or reoccurring.
- Identification of contingent actions that can be taken if preventive actions fall, or where no preventive action is possible.
In the current economic climate one cannot afford to have any failure potential. Problems must be found before they happen. Failure potential analysis is one of the finest tools available for bringing into focus today the best thinking of an informed management team properly concerned with the future (5).
MEASUREMENT TECHNIQUES
The principal techniques under consideration for integrated machine wear analysis are vibration analysis, performance monitoring, thermography, and machine lubricant analysis.
Vibration Analysis
Many an industrial predictive program relies heavily upon vibration monitoring which, even by itself, is an extremely powerful diagnostic tool. In some cases, an abnormal vibration spectrum will be the first indication of problems in a piece of equipment. In other cases, oil analysis will give an earlier indication of abnormality.
For example, a rotational imbalance would first be observed by vibration analysis. Only after a period of time in operation would the excess stress on the bearings result in a greater quantity of wear debris being present in the lubricating oil. On the other hand, an abrasive wear situation, in which contaminant particles have infiltrated the lubricant would first be detected by oil analysis.
Only after a significant amount of wear has occurred would a problem be indicated by vibration analysis.
One of the great problems with vibration is that it will not contain itself in a nice tidy area for precise analysis. It will insist on running away through foundations, conduits, pipelines and structures, etc. All machines vibrate: it is a natural phenomenon. Fundamental to using machine vibration as a measure of machine health condition is a need to understand it in relation to certain machine faults. These faults may be caused by poor installation, wear or damage.
Machine vibration may also be the result of misalignment, unbalance (imbalance), mechanical looseness, bearing or gear defects, vane/blade pass problems, electric motor defects or just simply uneven loading of the machine.
It is necessary to know what is normal, what is abnormal and be able to measure its behaviour and trend it. The two most useful criteria to capture are vibration amplitude and frequency, which can be effectively processed by digital fast Fourier transform (FFT) and displayed as a spectrum for further analysis. Frequency of vibration gives a clue to machine problems and their characteristic signature can recognize many problems.
The skill in successful troubleshooting is to be able to pick out the exceptions and not to be led to false conclusions. For instance, problems with gears almost always show up as a tooth meshing vibration modulated by a 1 x rpm vibration caused by eccentricity or a broken tooth. This appears on the frequency spectrum as a peak at the gear meshing frequency with sidebands on either side, separated by the gear running speed.
Performance Monitoring Performance monitoring refers to observing and trending a number of different operating parameters which, in many cases, are quite easy and inexpensive to obtain, such as temperatures, pressures, flow rates, oil consumption, etc. The operator is best placed to act as the monitor in humanized predictive maintenance (6). Without operator input, any attempt at assessing the potential for failure is close to guess- work.
Thermography
Thermography (infrared thermal imaging) is a technique for detection of hot spots in electrical switchboards, boilers, furnaces, mill drive systems, power transmission lines, etc. This capability can identify emerging defects with sufficient lead-time to allow for the planning of shutdown maintenance and the minimization of further damage.
Machine Lubricant Analysis
There are three basic methods for consideration under this heading: ferrography, elemental and oil condition analysis, and lubricant contamination monitoring.
Ferrography
Ferrography is a technique, which magnetically separate classifies and counts wear particles from lubricating oils, hydraulic fluids and greases. Optical microscopy examination of this wear fingerprint on a filter membrane (filtergram) can often diagnose the causes of abnormal wear within the machine compartment sampled.
Ferrographic procedures can cost effectively indicate the real wear conditions of many types of machinery without the necessity of disassembly. It is most important to remember that while direct reading (DR) ferrography is synonymous with wear debris analysis, the multitude of wear debris analysers available are not capable of ferrography if they provide only a total wear concentration index.
One such instrument worthy of note is the ferrous debris monitor. It is found to be useful as a screening device prior to DR ferrograph testing. The results obtained correlate well with those obtained on the DR ferrograph, and it is especially useful in assessing serial dilutions necessary when samples contain higher than normal levels of ferrous metal.
This test can often indicate the presence of non-Fe particles, especially when there is a low Fe level. Ferrographic oil analysis is applied to detect abnormal wear, diagnose the mode and severity of wear and provide some basis for prediction of the residual life of the wearing compartment. Size and distribution of particles can be measured and observed as concentrations of large particles, concentration of small particles and an index of wear severity calculated.
Ferrography is generally not suited to analysis of nonferrous wear debris particles. Nonferrous materials can be observed and reported with spectrometric analysis or other analysis methods. It is considered that each technique compliments the other and both should be used extensively.
Since the Ferrographic method is the most subjective, it would be used in the most appropriate applications, i.e., where the failure risk is the greatest and most costly.
Elemental Analysis by Spectroscopy
Spectrometric oil analysis (SOA) is the most widely applied technique for low-level debris monitoring. It gives a quantitative, multi-elemental analysis of wear debris in lubricating oil.
The serious limitation of the technique is its insensitivity to particles larger than approximately eight micrometers. Since incipient failure is usually manifested by a marked increase in the generation of large >15 micrometer particles, the time lag between the appearance of large particles and the change in the level of the elemental composition detected by SOA may lead in many circumstances to machine failure before maintenance action can be prescribed.
Spectrometric oil analysis is the method most favoured by machine vendors in Australia, especially when there are warranty implications.
Lubricant Condition Analysis
Most oil analysis laboratories perform one or more tests to determine oil's condition. These are mostly physical in nature although some tests are chemical and include viscosity, total base number (TAN), total acid number (TBN), pH, water, fuel dilution and Fourier transform infrared spectroscopy
(FTIR).
Lubricant Contamination Monitoring
Lubricant Contamination Monitoring is often overlooked even though most hydraulic component failures are due to the presence of abrasive contaminants in the fluid. Clean hydraulic fluid is the key to reliability and effective contamination control is the way to achieve and maintain this.
Insistence on care and attention to cleanliness during manufacture, installation and maintenance of the machine, and in the supply of new lubricants, will significantly reduce the risk of damage or failure. The most important factor governing the operational cleanliness of a system is the filter and how it is performing. The only way to confirm whether the correct filter has been selected is to monitor the dirt level in the lubricating fluid.
The preferred method of quoting the level of silt and solid contaminant particles in an oil sample is the use of International Standard Code ISO 4406. The code is constructed from the combination of two range numbers, i.e., 17/13, which indicates that there are between 640 and 1300 particles larger than 5 µm and between 40 and 80 particles larger than 15 µm. Both numbers relate to a 1 ml sample of oil.
WHICH TESTS TO USE AND WHY - DECISIONS, DECISIONS
In broad terms, diesel engines, transmissions, hydraulic systems, bearings and light duty gear systems do not need full diagnostic ferrography until spectrometric oil analysis, and/or ferrous debris monitor tests, flag the need. Differentials, final drives, and heavy-duty gear systems are different. They normally generate wear debris sizes, which are beyond the detection limits of a spectrometer: this is where ferrography is most beneficial.
It is important to keep an eye on the oil condition with viscosity, pH and water being the most important TAN should only be tested if the oil is expected to be in a compartment for a very prolonged period as it is often an unreliable test. pH is far more beneficial as it gives an immediate indication of the acidity of the lubricant.
These types of machine pre-screen the sample for ferrous debris concentration, and if the index number is greater than 100, then DR ferrograph readings are taken, rather than spectrometric elements. When reading from the ferrous debris monitor is below 100, then elemental analysis is proven to be realistic.
Hydraulic systems should always be tested for contamination level particle size distribution according to 150 4406. Many hydraulic system problems can be attributed to dirt. Specify ISO 4406 Code 17113 as a maximum level for the standard of cleanliness of new lubricants when issuing tenders for supply contracts.
For diesel engines it is important to monitor the total base number (TBN) if the operator wishes to extend oil drain periods. For 250-hour schedule changes it is a waste of time.
Dielectric constant is a far more useful test as it provides an indication of high soot levels and dilution from water or fuel.
It is often useful to remember that ferrous wear metals provide the best indicator to serious problems. Nonferrous components are often designed to be sacrificial and although there is a need to know what is happening inside, most breakdowns are the result of ferrous wear conditions. The three major contributors to poor machine condition are incorrect oil viscosity, dirt and water ingress, and operational overloading/abuse.
To accurately monitor an item of machinery using vibration, performance and condition analysis, it is essential to establish, as soon as possible, alarm or out of control levels, above which some form of remedial action is required to prevent machine breakdown. Alarm levels in the initial stages of monitoring should be set from experience.
A useful rule of thumb is that when a machine's vibration amplitude doubles over the time it was last read then there is a problem. If the vibration amplitude triples, then the machine should be shut down immediately.
This same logic can be applied to lubricant analysis also. Often, a single sample of oil is analysed when it can reveal interesting information regarding the condition of the machine or component part from which it was drawn. However, to obtain maximum benefit from an analysis, the results should, in most instances, be graphically represented and trended.
Spurious data, especially from incorrectly drawn oil samples, can often lead to unappreciated stoppages in production, only to discover that there was no defect condition at all. The laboratory analyst can only report on what is contained in the sample bottle. The real onus should be on the "sample taker" to be better.
There is nothing quite so humiliating as to be responsible for shutting down production equipment due to a decision based on an incorrectly drawn oil sample.
BENEFITS
Integrated machine wear control does require an investment in equipment and operational resources before returns from the strategy are realized as "bottom line savings." Cost benefit can be considered in terms of cost reduction, where:
Cost benefit (saving) = Cost of existing scheme - Cost of proposed scheme, and
Maintenance saving = Current maintenance strategy costs - Proposed maintenance strategy costs
The following statement appeared in the North Broken Hill Peko Limited Annual Report, Review of Operations for 1990: "New maintenance techniques introduced by the engineering technical services department at Ranger Mine produced cost efficiencies and savings of $1 million." The technologies referred to included vibration analysis, DR ferrography and filtergrams, thermography and lubricant contaminant monitoring. SOA procedures were already in use, and results were often disputed.
A strategy of integrated machine wear control can produce beneficial bottom line results, as indicated by the following survey facts covering 500 plants in America, Canada, Europe and Australia. The survey included plants from power generation, pulp and paper, metals food processing and textiles, and all participants had been operating the condition based maintenance program for three or more years (7).
- 50-80% reductions in repair costs
- 30% increase in revenue
- 50-80% reduction in maintenance costs
- Spares inventories reduced by more than 30%
- Overall profitability of plants increased by 20-60%
CONCLUSIONS
1. An effective machine wear control program integrates the use of lubricant, vibration, thermal and performance analyses. Do them on the same machine, at the same time, for maximum benefit.
2. The strategy for integrated machine wear control should be to use only those tests that aye relevant.
3. Acquisition of reliable and trendable data is vital to this concept and its management. Machine wear controlling incorporating spectrometry, fluid condition, diagnostic ferrography, lubricant contaminant monitoring, vibration, thermography and performance analysis allows a rational approach to failure prevention. When incorporated with failure potential analysis, it can often provide a safe interchange from scheduled to on-condition maintenance.
4. A properly integrated machine wear control program can produce increases in overall plant profitability between 20 to 60 percent.
REFERENCES
(1) Davidson, J., The Reliability of Mechanical Systems, 1MechE Guides for the Process Industries, Mechanical Engineering Publications Limited for the Institution of Mechanical Engineers, London, SP-6, (1994).
(2) Czichos, H., Tribology - A Systems Approach to the Science and Technology of Friction, Lubrication and Wear, Elsevier Scientific Publishing Company, Amsterdam, SP-11.
(3) Ball, P. G., "Improving Profitability Through a Reliability Centred Maintenance Strategy." in Proc. ICOMS-94, Maintenance- Repairing The Bottom Line, SP-2, Maintenance Engineering Society of Australia, Sydney, Paper No. 18, pp 1-6, (1994).
(4) Kepner, C. H. and Tregoe, B. B., The New Rational Manager, Princeton Research Press, NJ, SP- 142. (198 1).
(5) Ball, P. G., "Machine Reliability and the Understanding of Failure Potential," in Proc. (CM)2 Forum 1995, Centre for Machine Condition Monitoring, Monash University, Melbourne, pp l59-169, SP-160, (1995). -
(6) Sherwin, D. J., "Improving the Quality of Maintenance," in Proc. 1992- 2nd Queensland Maintenance Conf., Session Paper No. 2, SP-19, pp 15-21, (1992).
(7) "Cost Benefit Analysis Methods for Condition Monitoring, @ Solartron Instruments, Technical Report No. 27, in Potential Benefits for Condition Monitoring, SP-12, (1994).