News

Predictive Maintenance Management Techniques

• satyendra
• April 10, 2023
• 1 Comment
• MTTF, Oil analysis, predictive maintenance, preventive maintenance, Reactive analysis, Thermography, Tribology, ultrasonic, Vibratory analysis, visual inspection,

Predictive Maintenance Management Techniques

Maintenance cost is a major part of the total operating cost in all manufacturing plants. Depending on the specific industry, maintenance cost can represent between 15 % to 60 % of the cost of the goods produced. The dominant reason for the ineffective maintenance management is the lack of factual data to quantify the actual need for repair or maintenance of plant machinery, equipment, and systems. Maintenance scheduling has been, and in several cases, still is predicated on statistical trend data or on the actual failure of the plant equipment.

There are several different definitions of maintenance. Swedish standard SS-EN 13306 define it as a ‘Combination of all technical, administrative, and managerial actions during the life cycle of an item intended to retain it in, or restore it to, a state in which it can perform the required function’.

Main maintenance goals are to achieve maximum productivity and further optimization performance of the machine and to ensure safe operation of the device at least throughout its life expectancy. Increase in machine availability with minimal downtime, reducing downtime ideally to zero, and the process of continuous improvement are methods of securing these goals. Over time, the idea and the need to create a department which deal solely with the management of maintenance has evolved as an organizational philosophy. Like others, this department uses process management, planning, available information technology, and other leadership tools such as teamwork, etc. Maintenance according to terminology, and standards, has a direct link to reliability and hence to readiness, under which maintainability and maintenance is ensured.

As maintenance evolved, its role in the organization changed gradually. Earlier it used to be mostly about the need to repair damaged machines, but today, on the contrary, the need for prevention of disorders and accidents on the machinery is the most important aspect of maintenance, while ensuring the efficient, reliable, and safe operation of the machinery.

Maintenance is already associated with the time when people began to make tools. First, it was a necessity to repair tool which has gradually evolved into the need to work with clean and functional tools. Up to World War II, people did not place any emphasis on maintenance, as the machines were oversized and hence virtually trouble-free, and possible failures were handled operatively until they occurred. Until the end of the 1940s, people were talking about the first generation of the maintenance techniques characterized by so-called after-maintenance failure or reactive maintenance. With increasing demand and mechanization of production during World War II, the machines became more prone to failure and people began to monitor more closely and examine their lifespan, failure rate, and downtime. As the world entered the 1970s, the so-called second generation of maintenance techniques characterized by preventive maintenance with regular inspections and repairs appeared. John Moubray’s third generation of maintenance technique is the generation of predictive maintenance, when not only the need to plan and periodically check and repair the machine grew, but also to prevent machine failures by estimating its future status by tracking it or based on previous experience. It is the period of the greatest development of various failure analyses and risks.

Presently three maintenance strategies which are being followed in industry are (i) reactive or breakdown maintenance, (ii) preventive maintenance, and (iii) predictive maintenance. These three strategies are shown in Fig 1.

Fig 1 Types of maintenance

The types of maintenance have developed gradually depending on the circumstances. The oldest type is maintenance after failure, or reactive maintenance, whose advantage is the full use of the machine’s useful life. On the contrary, the disadvantage is unplanned breakdowns and possible long machine downtime. There are big financial losses and consequent costs of repair, purchase of parts, possibly safety risks associated with the disorder. This type is divided into immediate or delayed maintenance.

As the machines began to become larger and more complex, their operation became more expensive and a maintenance technique was developed with regular inspections and repairs, called preventive maintenance, whose goal is to prevent the failure. Regular inspections and revisions need to be defined, either by frequency of disorders expressed e.g., time to failure, etc., or are specified by the manufacturer. Also, basic perceptions such as hearing, sight, and touch by the machine operator are used, where one can guess from experience that there is something wrong with the machine.

With increasing effort to prevent unexpected failures and optimize maintenance costs new types of maintenance techniques have been developed such as predictive maintenance. The basic principles of predictive maintenance are diagnostics and condition monitoring of the equipment. By the long-term monitoring of certain machine parameters and diagnostics, a comprehensive evaluation of the condition of the machine is assessed, and based on this assessment maintenance of the machine is planned.

With reactive maintenance, the machine is used to its limit and repairs are performed only after the machine fails. Reactive maintenance approach is used for maintaining an inexpensive system.

Preventive maintenance approach is used for complex systems with some very expensive parts. This approach is taken since the risk of running the system to failure results into extremely costly repairs of the highly damaged parts. But, more importantly, it is also a safety issue. Preventive maintenance is done to prevent failure before it occurs by performing regular checks on the equipment. One big challenge with preventive maintenance is to determine when to do maintenance, since it is not known when failure is likely to occur. Hence preventive maintenance is based on conservative planning, especially when safety-critical equipment is there. But by scheduling maintenance very early, the machine life which is still usable gets wasted and this adds to the costs.

Predictive maintenance helps in estimating time-to-failure of a machine. Knowing the predicted failure time, helps in finding the optimum time to schedule the maintenance of the equipment. Predictive maintenance not only predicts a future failure, but also pinpoints problems in the complex machinery and helps in the identification of the parts which need to be fixed.

Traditional maintenance management techniques are (i) run-to-failure or break down maintenance, and (ii) preventive maintenance management. Run-to-failure is a reactive management technique which waits for machine or equipment failure before any maintenance action is taken. Preventive maintenance management programmes are time-driven. The statistical life of a machine-train is given by the mean-time-to-failure (MTTF) or bath-tub curve as shown in Fig 2. The bath tub curve indicates that a new machine has a high probability of failure because of installation problems during the first few weeks of operation. After this initial period, the probability of failure is relatively low for an extended period. After this normal machine life period, the probability of failure increases sharply with elapsed time. In preventive maintenance management, machine repairs or rebuilds are scheduled based on the MTTF statistics.

Fig 2 Typical bath-tub curve

Reliability is evaluated by various indicators, maintainability is described by diagnostics, reparability and ease of maintenance, including lubrication, cleaning and adjustment. Maintenance supportability is then more likely to be maintenance management’s ability to acquire and provide resources for the function of maintenance department and its activities.

In recent times, a variety of management methods, such as total productive maintenance (TPM) and reliability-centered maintenance (RCM), have been developed and hyped as the panacea for ineffective maintenance. TPM is the Japanese approach to effective maintenance management. It stresses absolute adherence to the basics, such as lubrication, visual inspections, and universal use of best practices in all aspects of maintenance. TPM is not a maintenance management programme. Majority of the activities are directed at the production function and assume that maintenance provides the basic tasks needed to maintain critical production assets.

In case of RCM, a basic premise is that all machines fail and have a finite useful life, but neither of these assumptions is valid. If machinery and plant systems are properly designed, installed, operated, and maintained, they do not fail, and their useful life is almost infinite. Few, if any, catastrophic failures are random, and some outside influence, such as operator error or improper repair, causes all failures. With the exception of instantaneous failures caused by gross operator error or a totally abnormal outside influence, the operating dynamics analysis methodology can detect, isolate, and prevent system failures. Since RCM is predicated on the belief that all machines degrade and fail (P-F curve), most of the tasks, such as failure modes and effects analysis (FMEA) and Weibull distribution analysis, are used to anticipate when these failures are going to occur. Both of the theoretical methods are based on probability tables which assume proper design, installation, operation, and maintenance of plant machinery. Neither is able to adjust for abnormal deviations in any of these categories.

Fig 3 P-F curve

Predictive maintenance has several definitions. To some employees, predictive maintenance is monitoring the vibration of rotating machinery in an attempt to detect incipient problems and to prevent catastrophic failure. To others, it is monitoring the infrared image of electrical switchgear, motors, and other electrical equipment to detect developing problems. The common premise of predictive maintenance is that regular monitoring of the actual mechanical condition, operating efficiency, and other indicators of the operating condition of machine-trains and process systems provide the data needed to ensure the maximum interval between repairs and minimize the number and cost of unscheduled outages created by machine-train failures.

Predictive maintenance means of improving productivity, product quality, and overall effectiveness of the production plants. It is not vibration monitoring or thermal imaging or lubricating oil analysis or any of the other non-destructive testing techniques which are being considered as predictive maintenance tools. Predictive maintenance is a philosophy or attitude which, simply stated, uses the actual operating condition of plant equipment and systems to optimize total plant operation.

A comprehensive predictive maintenance management programme uses the most cost-effective tools (e.g., vibration monitoring, thermography, tribology) to get the actual operating condition of critical plant systems and based on this actual data all maintenance activities are scheduled on an as-needed basis. Inclusion of predictive maintenance in a comprehensive maintenance management programme optimizes the availability of process machinery and greatly reduces the cost of maintenance. It also improves the product quality, productivity, and profitability of the plants.

Predictive maintenance is a condition-driven preventive maintenance programme. Instead of relying on industrial or in-plant average-life statistics (i.e., mean-time-to-failure) to schedule maintenance activities, predictive maintenance uses direct monitoring of the mechanical condition, system efficiency, and other indicators to determine the actual mean-time-to-failure or loss of efficiency for each machine-train and system in the plant.

A comprehensive predictive maintenance programme provides factual data on the actual mechanical condition of each machine-train and the operating efficiency of each process system. This data provides the maintenance manager with actual data for scheduling maintenance activities. A predictive maintenance programme can minimize unscheduled breakdowns of all mechanical equipment in the plant and ensure that repaired equipment is having acceptable mechanical condition. The programme can also identify machine-train problems before they become serious.

Majority of the mechanical problems can be minimized if they are detected and repaired early. Normal mechanical failure modes degrade at a speed directly proportional to their severity. If the problem is detected early, major repairs can normally be prevented.

Predictive maintenance using vibration signature analysis is predicated on two basic facts namely (i) all common failure modes have distinct vibration frequency components which can be isolated and identified, and (ii) the amplitude of each distinct vibration component remains constant unless the operating dynamics of the machine train change. These facts, their impact on machinery, and methods which identifies and quantifies the root cause of failure modes need to be are developed.

Predictive maintenance using process efficiency, heat loss, or other non-destructive techniques can quantify the operating efficiency of non-mechanical plant equipment or systems. These techniques used in conjunction with vibration analysis can provide maintenance managers and plant engineers with information which enable them to achieve optimum reliability and availability from the plant equipments.

Five non-destructive techniques are normally used for predictive maintenance management. These are (i) vibration monitoring, (ii) thermography, (iii) tribology, (iv) visual inspection, (v) ultrasonic, and (vi) other techniques. Each technique has a unique data set which assists the maintenance manager in determining the actual need for maintenance.

Most comprehensive predictive maintenance programmes use vibration analysis as the primary tool. Since majority of the normal plant equipment is mechanical, vibration monitoring provides the best tool for routine monitoring and identification of incipient problems. However, vibration analysis does not provide the data needed on electrical equipment, areas of heat loss, condition of lubricating oil, or other parameters which is needed to be considered as part of the predictive maintenance programme.

A variety of technologies are used as part of a comprehensive predictive maintenance programme. Since mechanical systems or machines account for most plant equipment, vibration monitoring is normally the key component of most predictive maintenance programmes. However, vibration monitoring cannot provide all of the information needed for a successful predictive maintenance programme. This technique is limited to monitoring the mechanical condition and not the other critical parameters needed to maintain reliability and efficiency of machinery.

Vibration monitoring is a very limited tool for monitoring critical process and machinery efficiencies and other parameters which can severely limit productivity and product quality. Hence, a comprehensive predictive maintenance programme is needed to include other monitoring and diagnostic techniques. These techniques include vibration monitoring, thermography, tribology, process parameters, visual inspection, ultrasonics, and other non-destructive testing techniques.

Vibration monitoring – Since majority of plants consist of electro-mechanical systems, vibration monitoring is the primary predictive maintenance tool. Over the past few years, majority of these programmes have adopted the use of microprocessor-based, single-channel data collectors and Windows-based software to acquire, manage, trend, and evaluate the vibration energy created by these electro-mechanical systems. Although this approach is a valuable predictive maintenance methodology, these systems’ limitations can restrict potential benefits.

Computer-based systems have several limitations. In addition, some system characteristics, particularly simplified data acquisition and analysis, provide both advantages and disadvantages. While providing several advantages, simplified data acquisition and analysis can also be a liability. If the database is improperly configured, the automated capabilities of these analyzers yield faulty diagnostics which can allow catastrophic failure of critical plant machinery. Since operator involvement is reduced to a minimum, the normal tendency is to use untrained or partially trained personnel for this repetitive function. Unfortunately, the lack of training results in less awareness and knowledge of visual and audible clues which can, and are to be, an integral part of the monitoring programme.

Majority of the microprocessor-based vibration-monitoring systems collect single channel, steady-state data which cannot be used for all applications. Single-channel data are limited to the analysis of simple machinery which operates at relatively constant speed. Although majority of the microprocessor-based instruments are limited to a single input channel, in some cases, a second channel is incorporated in the analyzer, however, this second channel normally is limited to input from a tachometer, or a once-per-revolution input signal. This second channel cannot be used for vibration data capture. This limitation prohibits the use of most microprocessor-based vibration analyzers for complex machinery or machines with variable speeds.

Single-channel data acquisition technology assumes the vibration profile generated by a machine-train remains constant throughout the data acquisition process. This is normally true in applications where machine speed remains relatively constant i.e., within 5 rpm (revolutions per minute) to 10 rpm. In this case, its use does not severely limit diagnostic accuracy and can be effectively used in a predictive maintenance programme.

Majority of the microprocessor-based instruments are designed to handle steady-state vibration data. Few have the ability to reliably capture transient events such as rapid speed or load changes. As a result, their use is limited in situations where these changes occur. In addition, vibration data collected with a microprocessor-based analyzer are filtered and conditioned to eliminate non-recurring events and their associated vibration profiles. Anti-aliasing filters are incorporated into the analyzers specifically to remove spurious signals such as impacts or transients. Although the intent behind the use of anti-aliasing filters is valid, their use can distort a machine’s vibration profile.

Since vibration data are dynamic and the amplitudes constantly change, as shown in Fig 4, majority of predictive maintenance system suppliers strongly recommend averaging the data. They typically recommend acquiring 3 samples to 12 samples of the vibration profile and averaging the individual profiles into a composite signature. This approach eliminates the variation in vibration amplitude of the individual frequency components which make up the machine’s signature, however, these variations, referred to as beats, can be a valuable diagnostic tool. Unfortunately, they are not available from microprocessor-based instruments because of averaging and other system limitations.

Fig 4 Vibrational analysis data

The most serious limitations created by averaging and the anti-aliasing filters are the inability to detect and record impacts which frequently occur within machinery. These impacts normally are indications of abnormal behaviour and are frequently the key to detecting and identifying incipient problems.

Majority of the predictive maintenance programmes rely almost exclusively on frequency-domain vibration data. The microprocessor-based analyzers gather time-domain data and automatically convert it using Fast Fourier Transform (FFT) to frequency-domain data. A frequency-domain signature shows the machine’s individual frequency components, or peaks. While frequency-domain data analysis is much easier to learn than time-domain data analysis, it cannot isolate and identify all incipient issues within the machine or its installed system. Because of this limitation, additional techniques (e.g., time-domain, multi-channel, and real-time analysis) are to be used in conjunction with frequency domain data analysis to get a complete diagnostic picture.

Several of the microprocessor-based vibration-monitoring analyzers cannot capture accurate data from low-speed machinery or machinery which generates low frequency vibration. Specifically, some of the commercially available analyzers cannot be used where frequency components are below 600 cycles per minute (CPM) or 10 Hz (Hertz). Two major issues restricting the ability to acquire accurate vibration data at low frequencies are electronic noise and the response characteristics of the transducer. The electronic noise of the monitored machine and the ‘noise floor’ of the electronics within the vibration analyzer tend to override the actual vibration components found in low-speed machinery. Analyzers especially equipped to handle noise are needed for the majority of the industrial applications. Some of the commercially available microprocessor-based analyzers are capable of acquiring data below 600 CPM. These systems use special filters and data acquisition techniques to separate real vibration frequencies from electronic noise. In addition, transducers with the needed low-frequency response are to be used.

All machine-trains are subject to random, non-recurring vibrations as well as periodic vibrations. Hence, it is advisable to acquire several sets of data and average them to eliminate the spurious signals. Averaging also improves the repeatability of the data since only the continuous signals are retained. Typically, a minimum of three samples are to be collected for an average, however, the factor which determines the actual number is time. One sample takes 3 seconds to 5 seconds, a four-sample average takes 12 seconds to 20 seconds, and a 1,000-sample average takes 50 minutes to 80 minutes to acquire. Hence, the final determination is the quantity of time which can be spent at each measurement point. In general, three to four samples are acceptable for good statistical averaging and keeping the time needed per measurement point within reasonable limits. Exceptions to this recommendation include low-speed machinery, transient-event capture, and synchronous averaging.

Several of the microprocessor-based vibration-monitoring systems offer the ability to increase their data acquisition speed. This option is referred to as overlap averaging. Although this approach increases speed, it is normally not recommended for vibration analysis. Overlap averaging reduces the data accuracy and is to be used with caution. Its use is to be avoided except where fast transients or other unique machine-train characteristics need an artificial means of reducing the data acquisition and processing time.

When sampling time is limited, a better approach is to reduce or eliminate averaging altogether in favour of acquiring a single data block, or sample. This reduces the acquisition time to its absolute minimum. In majority of the cases, the single-sample time interval is less than the minimum time needed to get two or more data blocks using the maximum overlap-averaging sampling technique. In addition, single-sample data are more accurate. Tab 1 describes overlap-averaging options. The approach described in this table assumes that the vibration profile of monitored machines is constant.

Perhaps the most serious diagnostic error made by typical vibration-monitoring programmes is the exclusive use of vibration-based failure modes as the diagnostic logic. As an example, majority of the logic trees state that when the dominant energy contained in a vibration signature is at the fundamental running speed, then a state of unbalance exists. Although some forms of unbalance create this profile, the rules of machine dynamics clearly indicate that all failure modes on a rotating machine increases the amplitude of the fundamental or actual running speed.

Without a thorough understanding of machine dynamics, it is virtually impossible to accurately diagnose the operating condition of critical plant production systems. As an example, gear manufacturers do not finish the backside (i.e., non-drive side) of gear teeth. Hence, any vibration acquired from a gear set when it is braking is of an order of magnitude higher than when it is operating on the power side of the gear.

Another example is even more common. Majority of the analysts ignore the effect of load on a rotating machine. If one is to acquire a vibration reading from a centrifugal compressor when it is operating at full load, it can generate an overall level of 2.5 mms-peak (millimetre per second). The same measurement point generates a reading in excess of 10 mms-peak when the compressor is operating at 50 % load. The difference is the spring constant which is being applied to the rotating element. The spring constant or stiffness at 100 % load is twice that of when operating at 50 %, however, spring constant is a quadratic function. A reduction of 50 % in the spring constant increases the vibration level by a factor of four. For achieving maximum benefits from vibration monitoring, the analyst is required to understand the limitations of the instrumentation and the basic operating dynamics of machinery. Without this knowledge, the benefits are dramatically reduced.

The highest mistake made by traditional application of vibration monitoring is in its application. Majority of the programmes limit the use of this predictive maintenance technology to simple rotating machinery and not to the critical production systems which constitute the plant’s capacity. As a result, the auxiliary equipment is kept in good operating condition, but the plant’s throughput is unaffected.

Vibration monitoring is not limited to simple rotating equipment. The microprocessor-based systems used for vibration analysis can be used effectively on all electro-mechanical equipment, no matter how complex or what form the mechanical motion can take. As an example, it can be used to analyze hydraulic and pneumatic cylinders which are purely linear motion. For accomplishing this type of analysis, the analyst is required to use the time-domain function which is built into these instruments. Proper operation of cylinders is determined by the time it takes for the cylinder to finish one complete motion. The time needed for the cylinder to extend is shorter than its return stroke. This is a function of the piston area and inlet pressure. By timing the transient from fully retracted or extended to the opposite position, the analyst can detect packing leakage, scored cylinder walls, and other failure modes.

Vibration monitoring is needed to be focused on the critical production systems. Each of these systems is to be evaluated as a single machine and not as individual components. As an example, a machine, annealing line, or any other production system is to be analyzed as a complete machine, not as individual gearboxes, rolls, or other components. This methodology permits the analyst to detect abnormal operation within the complex system. Issues such as tracking, tension, and product-quality deviations can be easily detected and corrected using this method.

When properly used, vibration monitoring and analysis is the most powerful predictive maintenance tool available. It is required to be focused on critical production systems, not simple rotating machinery. Diagnostic logic is to be driven by the operating dynamics of machinery, not simplified vibration failure modes. The proof is in the results. When properly used, vibration-based predictive maintenance can generate return on investment of 100:1 or better.

Thermography – Thermography is a predictive maintenance technique which can be used to monitor the condition of plant machinery, structures, and systems, not just electrical equipment. It uses instrumentation designed to monitor the emission of infrared energy (i.e., surface temperature) to determine operating condition. By detecting thermal anomalies (i.e., areas which are hotter or colder than they are required to be), an experienced person can locate and define a multitude of incipient issues within the plant.

Infrared technology is predicated on the fact that all objects having a temperature above absolute zero emit energy or radiation. Infrared radiation is one form of this emitted energy. Infrared emissions, or below red, are the shortest wavelengths of all radiated energy and are invisible without special instrumentation. The intensity of infrared radiation from an object is a function of its surface temperature. However, temperature measurement using infrared methods is complicated since three sources of thermal energy can be detected from any object namely (i) energy emitted from the object itself, (ii) energy reflected from the object, and (iii) energy transmitted by the object. Only the emitted energy is important in a predictive maintenance programme.

Reflected and transmitted energies distort raw infrared data. Hence, the reflected and transmitted energies are to be filtered out of the acquired data before a meaningful analysis can be completed. Variations in surface condition, paint, or other protective coatings, and several other variables can affect the actual emissivity factor for plant equipment. In addition to reflected and transmitted energy, the user of thermographic techniques is also to consider the atmosphere between the object and the measurement instrument. Water vapour and other gases absorb infrared radiation. Airborne dust, some lighting, and other variables in the surrounding atmosphere can distort measured infrared radiation. Since the atmospheric environment is constantly changing, using thermographic techniques needs extreme care each time infrared data are acquired.

Majority of infrared-monitoring systems or instruments suppliers provide filters which can be used to avoid the negative effects of atmospheric attenuation of infrared data. However, the user is to recognize the specific factors which affect the accuracy of the infrared data and apply the correct filters or other signal conditioning needed to negate those specific attenuating factor or factors. Collecting optics, radiation detectors, and some forms of indicator are the basic elements of an industrial infrared instrument. The optical system collects radiant energy and focuses it on a detector, which converts it into an electrical signal. The instrument’s electronics amplifies the output signal and processes it into a form which can be displayed. There are three types of thermographic instruments which are normally used as part of an effective predictive maintenance programme. These are (i) infrared thermometers, (ii) line scanners, and (iii) infrared imaging systems.

Infrared thermometers or spot radiometers are designed to provide the actual surface temperature at a single, relatively small point on a machine or surface. Within a predictive maintenance programme, the point-of-use infrared thermometer can be used in conjunction with several of the microprocessor-based vibration instruments for the monitoring of the temperature at critical points on plant machinery or equipment. This technique is typically used for the monitoring of the bearing cap temperatures, motor winding temperatures, spot checks of process piping temperatures, and similar applications. It is limited in that the temperature represents a single point on the machine or structure. However, when used in conjunction with vibration data, point-of-use infrared data can be a valuable tool.

Line scanners are the type of infrared instrument which provides a one-dimensional scan or line of comparative radiation. Although this type of instrument provides a somewhat larger field of view (i.e., area of machine surface), it is limited in predictive maintenance applications.

Unlike other infrared techniques, thermal or infrared imaging provides the means to scan the infrared emissions of complete machines, process, or equipment in a very short time. Majority of the imaging systems function much like a video camera. The user can view the thermal emission profile of a wide area by simply looking through the instrument’s optics. A variety of thermal imaging instruments are available, ranging from relatively inexpensive, black-and-white scanners to full-colour, microprocessor-based systems. Several of the less expensive units are designed strictly as scanners and cannot store and recall thermal images. This inability to store and recall previous thermal data limits a long-term predictive maintenance programme. Point-of-use infrared thermometers are commercially available and are relatively inexpensive.

Training is critical with any of the imaging systems. The variables which can destroy the accuracy and repeatability of thermal data are to be compensated for each time infrared data are acquired. In addition, interpretation of infrared data needs extensive training and experience. Inclusion of thermography into a predictive maintenance programme enables a person to monitor the thermal efficiency of critical process systems which rely on heat transfer or retention, electrical equipment, and other parameters which improve both the reliability and efficiency of plant systems.

Infrared techniques can be used to detect problems in a variety of plant systems and equipment, including electrical switchgear, gearboxes, electrical substations, transmissions, circuit breaker panels, motors, building envelopes, bearings, steam lines, and process systems which rely on heat retention or transfer.

Safety is important in infrared thermography. Equipment included in an infrared thermography inspection is normally energized, and hence, a lot of attention is to be given to safety. The basic rules for safety which are to be followed while performing an infrared inspection are described below.

Plant safety rules are to be followed at all times. A safety person is to be available at all the times, since proper use of infrared imaging systems needs the person to use a viewfinder, similar to a video camera, to view the machinery to be scanned, and the person is blind to the surrounding environment. Hence, a safety person is needed for ensuring safe completion. It is necessary to notify area personnel before entering the area for scanning. A qualified electrician from the area is to be assigned to open and close all electrical panels. Where it is safe and possible, all equipment to be scanned are to be online and under normal load with a clear line of sight to the item. Equipment whose covers are interlocked without an interlock defect mechanism are to be shut down when allowable. If safe, their control covers are to be opened and equipment restarted.

When used correctly, thermography is a valuable predictive maintenance and / or reliability tool. However, the derived benefits are directly proportional to how it is used. If it is limited to annual surveys of roofs and / or quarterly inspections of electrical systems, the resultant benefits are limited. When used to regularly monitor all critical process or production systems where surface temperature or temperature distribution indicates reliability or operating condition, thermography can yield substantial benefits. For getting the maximum benefits from the investment in infrared systems, one is to use its full power. The thermography programme is to be concentrated on those critical systems which generate capacity in the plant.

Tribology – Tribology is the general term which refers to design and operating dynamics of the bearing-lubrication-rotor support structure of machinery. Two primary techniques are being used for predictive maintenance namely (i) lubricating oil analysis, and (ii) wear particle analysis. Lubricating oil analysis, as the name implies, is an analysis technique which determines the condition of the lubricating oils used in mechanical and electrical equipment. It is not a tool for determining the operating condition of machinery or detecting potential failure modes. A large number of plants are attempting to accomplish the latter and are disappointed in the benefits which are derived. Simply stated, lubricating oil analysis is to be limited to a proactive programme to conserve and extend the useful life of the lubricants.

Although some forms of lubricating oil analysis can provide an accurate quantitative breakdown of individual chemical elements (both oil additive and contaminants contained in the oil), the technology cannot be used to identify the specific failure mode or root-cause of incipient problems within the machines serviced by the lubricating oil system. The primary applications for the lubricating oil analysis are quality control, reduction of lubricating oil inventories, and determination of the most cost-effective interval for oil change.

Lubricating, hydraulic, and dielectric oils can be periodically analyzed using these techniques to determine their condition. The results of this analysis can be used to determine if the oil meets the lubricating requirements of the machine or application. Based on the results of the analysis, lubricants can be changed or upgraded to meet the specific operating requirements. In addition, detailed analysis of the chemical and physical properties of different oils used in the plant can, in some cases, allow consolidation or reduction of the number and types of lubricants needed to maintain plant equipment.

Elimination of unnecessary duplication can reduce the needed inventory levels and hence the maintenance costs. As a predictive maintenance tool, lubricating oil analysis can be used to schedule oil change intervals based on the actual condition of the oil. In mid-size to large plants, a reduction in the number of oil changes can amount to a considerable annual reduction in maintenance costs. Relatively inexpensive sampling and testing can show when the oil in a machine has reached a point which warrants change.

Wear particle analysis is related to oil analysis only in that the particles to be studied are collected by drawing a sample of lubricating oil. Whereas lubricating oil analysis determines the actual condition of the oil sample, wear particle analysis provides direct information about the wearing condition of the machine-train. Particles in the lubricating oil of a machine can provide considerable information about the machine’s condition. This information is derived from the study of particle shape, composition, size, and quantity.

For the analysis of particulate matter, two methods are used to prepare samples of wear particles. The first method, called spectroscopy or spectrographic analysis, uses graduated filters to separate solids into sizes. Normal spectrographic analysis is limited to particulate contamination with a size of 10 micrometres or less. Larger contaminants are ignored. This fact can limit the benefits which can be derived from the technique. The second method, called ferro-graphic analysis, separates wear particles using a magnet. Obviously, the limitation to this approach is that only magnetic particles are removed for analysis. Non-magnetic materials, such as copper, aluminum, and so on which make up several of the wear materials in typical machinery are hence excluded from the sample.

Wear particle analysis is an excellent failure analysis tool and can be used to understand the root-cause of catastrophic failures. The unique wear patterns observed on failed parts, as well as those contained in the oil reservoir, provide a positive means of isolating the failure mode.

There are some limitations of the tribology technique. Three major limitations associated with using tribology analysis in a predictive maintenance programme are (i) equipment costs, (ii) acquiring accurate oil samples, and (iii) interpretation of data. The capital cost of spectrographic analysis instrumentation is normally very high to justify in-plant testing. Because of this, majority of the predictive maintenance programmes rely on third-party analysis of oil samples. In addition to the employee cost associated with regular gathering of oil and grease samples, there is a cost involved for a simple lubricating oil analysis by a testing laboratory.

Standard analysis normally includes viscosity, flash point, total insolubles, total acid number (TAN), total base number (TBN), fuel content, and water content. More detailed analysis, using spectrographic, ferro-graphic, or wear particle techniques which include metal scans, particle distribution (size), and other data cost more per sample.

A more severe limiting factor with any method of oil analysis is acquiring accurate samples of the true lubricating oil inventory in a machine. Sampling is not a matter of opening a port somewhere in the oil line and catching a half litre sample. Extreme care is to be taken to acquire samples which truly represent the lubricant and which passes through the machine’s bearings.

One recent example is an attempt to acquire oil samples from a bull gear compressor. The lubricating oil filter had a sample port on the clean (i.e., downstream) side, however, comparison of samples taken at this point and one taken directly from the compressor’s oil reservoir has indicated that more contaminants have existed downstream from the filter than in the reservoir. These locations actually have not represented the oil’s condition. Neither sample has truly representative of the oil’s condition. The oil filter has removed majority of the suspended solids (i.e., metals and other insolubles) and hence the sample has not been representative of the actual condition. The reservoir sample is also not representative since majority of the suspended solids has settled out in the sump.

Proper methods and frequency of sampling lubricating oil are critical to all predictive maintenance techniques which use lubricant samples. Sample points which are consistent with the objective of detecting large particles are to be chosen. In a recirculating system, samples are to be drawn as the lubricating oil returns to the reservoir and before any filtration occurs. Oil is not to be drawn from the bottom of a sump where large quantities of material build up over time. Return lines are preferable to reservoir as the sample source, but good reservoir samples can be taken if careful, consistent practices are used. Even equipment with high levels of filtration can be effectively monitored as long as samples are drawn before oil enters the filters.

Sampling techniques involve taking samples under uniform operating conditions. Samples are not to be taken more than 30 minutes after the equipment has been shut down. Sample frequency is a function of the mean-time-to-failure (MTTF) from the onset of an abnormal wear mode to catastrophic failure. For machines in critical service, sampling every 25 hours of operation is appropriate. For majority of the industrial equipment in continuous service, however, monthly sampling is adequate. The exception to monthly sampling is machines with extreme loads. In this case, weekly sampling is desired.

Understanding the meaning of analysis results is perhaps the most serious limiting factor. Results are normally expressed in terms which are totally unfamiliar to plant engineers or technicians. Hence, it is difficult for them to understand the true meaning, in terms of oil or machine condition. A good background in quantitative and qualitative chemistry is beneficial. At a minimum, plant personnel need training in basic chemistry and specific instruction on interpreting tribology results.

Visual inspection – Visual inspection is the first method used for predictive maintenance. Almost from the beginning of the Industrial Revolution, maintenance personnel performed daily ‘walkdowns’ of critical production and manufacturing systems in an attempt to identify potential failures or maintenance-related issues which can impact reliability, product quality, and production costs. A visual inspection is still a viable predictive maintenance tool and is to be included in all total-plant maintenance management programmes.

Ultrasonics – Ultrasonics, like vibration analysis, is a subset of noise analysis. The only difference in the two techniques is the frequency band which they monitor. In the case of vibration analysis, the monitored range is between 1 Hz to 30,000 Hz, while ultrasonics monitor noise frequencies higher than 30,000 Hz. These higher frequencies are useful for select applications, such as detecting leaks which normally create high-frequency noise caused by the expansion or compression of air, gases, or liquids as they flow through the orifice, or a leak in either pressure or vacuum vessels. These higher frequencies are also useful in measuring the ambient noise levels in different areas of the plant.

As it is being applied as part of a predictive maintenance programme, several organizations are attempting to replace what is perceived as an expensive tool (i.e., vibration analysis) with ultrasonics. For example, several plants are using ultrasonic meters to monitor the health of rolling-element bearings in the belief that this technology provides accurate results. Unfortunately, this perception is not valid.

Since this technology is limited to a broadband (i.e., 30 kHz to 1 MHz), ultrasonics does not provide the ability to diagnosis incipient bearing or machine issues. It certainly cannot define the root-cause of abnormal noise levels generated by either bearings or other machine-train components. As part of a comprehensive predictive maintenance programme, ultrasonics are to be limited to the detection of abnormally high ambient noise levels and leaks. Attempting to replace vibration monitoring with ultrasonics simply does not work.

Other techniques – A number of other non-destructive techniques can be used to identify incipient issues in plant equipment or systems. However, these techniques either do not provide a broad enough application or are too expensive to support a predictive maintenance programme. Hence, these techniques are used as the means of confirming failure modes identified by the predictive maintenance techniques described earlier.

Traditional electrical testing methods are used in conjunction with vibration analysis to prevent premature failure of electric motors. These tests include (i) resistance testing, (ii) Megger testing, (iii) HiPot testing, (iv) impedance testing, and (v) other techniques.

In case of resistance testing, resistance is measured by using an ohmmeter. In reality, an ohmmeter does not directly measure resistance, instead it measures current. The scale of the meter is calibrated in ohms, but the meter movement responds to current. The quantity of current supplied by the meter is very low, typically in the range of 20 micro-amperes to 50 micro-amperes. The meter functions by applying its terminal voltage to the test subject and measuring the current in the circuit. For practical purposes, although resistance testing is of limited value, some useful tests can be performed.

A resistance test indicates an open or closed circuit. This can tell a person whether there is a break in a circuit or if there is a dead short to ground. It is important to remember that inductive and capacitive elements in the circuit distort the resistance measurements. Capacitive elements appear initially as a short circuit and begin to open as they charge. They appear as open circuits when they are fully charged. Inductive elements appear initially as open circuits, and the resistance decreases as they charge. In both cases, the actual charging time is tied to the actual resistance, capacitance, and inductance in the circuit in question. It still needs five-time constants to charge capacitors and inductors.

It is also important to remember that when disconnecting the meter from the circuit where there are now charged capacitive and inductive elements present, due caution is to be observed when disconnecting the test equipment. Resistance testing is of limited value for testing coils. It detects an open coil, or a coil shorted to ground. Resistance testing very frequently does not detect windings which are shorted together or weak insulation.

Megger testing is used to measure high resistances. For this testing, a device known as a mega-ohmmeter can be used. This instrument differs from a normal ohmmeter in that instead of measuring current to determine resistance, it measures voltage. This mode of testing involves applying relatively high voltage (500 volts to 2,500 volts, depending on the unit) to the circuit and verifying that no breakdown is present. Normally, this is considered a non-destructive test, depending on the applied voltage and the rating of the insulation. This method of testing is used primarily to test the integrity of insulation. It does not detect shorts between windings, but it can detect higher-voltage–related issues with respect to ground.

HiPot (high potential) testing is a potentially destructive test used to determine the integrity of insulation. Voltage levels used in this type of test are twice the rated voltage plus 1,000 volts. This method is used primarily by some equipment manufacturers and rebuilding facilities as a quality assurance tool. It is important to note that HiPot testing does some damage to insulation every time it is performed. HiPot testing can destroy insulation which is still serviceable, so this test is normally not desired in field use.

Impedance has two components namely (i) a real (or resistive) component, and (ii) a reactive (inductive or capacitive) component. Impedance testing is useful since it can detect significant shorting in coils, either between turns or to ground. No other non-intrusive method exists to detect a coil which is shorted between turns.

Other techniques which can support predictive maintenance include acoustic emissions, eddy-current, magnetic particle, residual stress, and majority of the traditional non-destructive methods.

Optimizing predictive maintenance

A large number the predictive maintenance programmes which have been implemented have failed to generate measurable benefits. These failures have not been caused by technology limitation, but rather by the failure to make the necessary changes in the workplace which permits maximum utilization of these predictive tools. As a minimum, the following proactive steps can eliminate these restrictions and as a result help gain maximum benefits from the predictive maintenance programmes.

Culture change – The first change which is needed to take place is to change the perception that predictive technologies are exclusively a maintenance management or breakdown prevention tool. This change is to take place at the higher management level and permeate throughout the plant. This task can sound simple, but changing management attitude toward or perception of maintenance and predictive maintenance is difficult. Since majority of the higher-level managers have little or no knowledge or understanding of maintenance, or even the need for maintenance, convincing them that a broader use of predictive technologies is necessary is extremely difficult. In their myopic view, breakdowns and unscheduled delays are solely a maintenance issue. They cannot understand that majority of these failures are the result of non-maintenance issues.

From studies of equipment reliability issues conducted over a 30 years period, it is seen that maintenance is responsible for around 17 % of production interruptions and quality problems. The remaining 83 % are totally outside of the traditional maintenance function’s responsibility. Inappropriate operating practices, poor design, non-specification parts, and a countless of other non-maintenance reasons are the primary contributors to production and product-quality issues, not the maintenance.

Predictive technologies are to be used as a plant or process optimization tool. In this broader scope, these technologies are used to detect, isolate, and provide solutions for all deviations from acceptable performance which result in lost capacity, poor quality, abnormal costs, or a threat to employee safety. These technologies have the power to fill this critical role, but that power is simply not being used. For accomplishing this new role, the use of predictive technologies is to be shifted from the maintenance department to a reliability group which is charged with the responsibility and is accountable for plant optimization.

This group is required to have the authority to cross all functional boundaries and to implement changes which correct issues uncovered by their evaluations. This approach is a radical departure from the traditional organization found in the majority of the plants. As a result, resistance is met from all levels of the organization. With the exception of those few employees who understand the absolute need for a change to better, more effective practices, majority of the employees does not openly embrace or voluntarily accept this new functional group. However, the formation of a dedicated group of professionals which is absolutely and solely responsible for reliability improvement and optimization of all facets of plant operation is necessary. It is the only way a plant can achieve and sustain world-class performance.

Employees selection for this new group is not easy. The team needs to have a thorough knowledge of machine and process design, and be able to implement best practices in both operation and maintenance of all critical production systems in the plant. In addition, the team members are to fully understand procurement and plant engineering methods which provides best life-cycle cost for these systems. Finally, the team members are to understand the proper use of predictive technologies.

Few plants have existing employees who have all of the fundamental requirements. This issue can be resolved in two ways. The first approach is to select personnel who have mastered one or more of these knowledge requirements. For example, the group can consist of the best operations, maintenance, engineering, and predictive personnel available from the present human resource of the plant. Care is to be taken to ensure that group members have a real knowledge of their specialty area.

One common issue which plagues plants is that the superstars in the organization do not have a real, in-depth knowledge of their perceived specialty. In other words, the best operators can in fact be the worst contributor to reliability or performance issues. Although they can get more capacity through the unit than anyone else, the practices used can be the root-cause of the chronic issues. If this approach is followed, training for the reliability team is to be the first priority. Few existing personnel have all of the knowledge and skills needed by this function, especially regarding application of predictive technologies. Hence, the organization is required to provide sufficient training to ensure maximum return on its investment.

The training is to focus on process or operating dynamics for each of the critical production systems in the plant. It is to include comprehensive process design, operating envelope, operating methods, and process diagnostics training which can form the foundation for the reliability group’s ability to optimize performance.

The second approach is to recruit professional reliability engineers. This approach can sound easier, but it is not since there are very few fully qualified reliability professionals available, and they are very, very expensive. Majority of these professionals prefer to offer their services as short-term consultants rather than become a long-term employee. If the organization tries to recruit rather than posting the employees internally, use of extreme caution is needed.

Proper use of predictive technologies system components, such as pumps, gearboxes, and so on, are an integral part of the system and are to operate within their design envelope before the system can meet its designed performance levels. Then, why the majority of the predictive programmes treat these components as isolated machine-trains and not as part of an integrated system. Instead of evaluating a centrifugal pump or gearbox as part of the total machine, majority of the predictive analysts limit technology use to simple diagnostics of the mechanical condition of that individual component. As a result, no effort is made to determine the influence of system variables, like load, speed, product, or instability on the individual component. These variations in process variables are frequently the root-cause of the observed mechanical issue in the pump or gearbox.

Unless analysts consider these variables, they are not able to determine the true root-cause. Instead, they make recommendations to correct the symptom (e.g., damaged bearing, mis-alignment), rather than the real issue. The converse is also true. When diagnostics are limited to individual components, system issues cannot be detected, isolated, and resolved. The system, not the individual components of that system, generates capacity, revenue, and bottom-line profit for the organization. Hence, the system is to be the primary focus of analysis.

When one thinks of predictive maintenance, vibration monitoring, thermography, or tribology is the normal vision. These are powerful tools, but they are not the panacea for plant problems. Used individually or in combination, these three cornerstones of predictive technologies cannot provide all of the diagnostics needed to achieve and sustain world-class performance levels. For gaining maximum benefit from predictive technologies, some changes are needed. Process parameters, such as flow rates, retention time, temperatures, and others, are absolute requirements in all predictive maintenance and process optimization programmes. These parameters define the operating envelope of the process and are essential requirements for system operation.

In several cases, the above data are readily available. On systems which use computer-based or programmable logic controller (PLC), the parameters or variables which define their operating envelopes are automatically acquired and then used by the control logic to operate the system. The type and number of variables vary from system to system but are based on the actual design and mode of operation for that specific type of production system. It is a relatively simple matter to acquire these data from the Level 1 control system and use it as part of the predictive diagnostic logic.

In several cases, these data combined with traditional predictive technologies provide all of the data which are needed by the analyst to fully understand the system’s performance. Manually operated systems are not to be ignored. Although the process data is more difficult to get, the reliability or predictive analyst can normally acquire enough data to permit full diagnostics of the system’s performance or operating condition. Analog gauges, thermocouples, strip chart recorders, and other traditional plant instrumentation can be used. If plant instrumentation includes an analog or digital output, most microprocessor-based vibration meters can be used for direct data acquisition. These instruments can directly acquire most proportional signal outputs and automate the data acquisition and management which is needed for this expanded scope of predictive technology.

Since majority of equipment used in manufacturing, production, and process plants consists of electro-mechanical systems, the discussion begins with the best methods for this classification of equipment. Depending on the plant, these systems can range from simple machine-trains, such as drive couple pumps and electric motors, to complex continuous process lines. Regardless of the complexity, the methods which are to be used are similar. In all programmes, the primary focus of the predictive maintenance programme is to be on the critical process systems or machine-trains which constitute the primary production activities of the plant.

Although auxiliary equipment is important, the programme is to first address those systems on which the plant relies to produce revenue. In several cases, this approach is a radical departure from the presently used methods in traditional applications of predictive maintenance. In these programmes, the focus is on simple rotating machinery and excludes the primary production processes.

Predictive maintenance for all electro-mechanical systems, regardless of their complexity, are to use a combination of vibration monitoring, operating dynamics analysis, and infrared technologies. This combination is needed to ensure the ability to accurately determine the operating condition, to identify any deviation from acceptable operations, and to isolate the root-cause of these deviations.

In the case of vibration analysis, single-channel vibration analysis, using microprocessor-based, portable instruments, is acceptable for routine monitoring of these critical production systems. However, the methods used is to provide an accurate representation of the operating condition of the machine or system. The biggest change which is to be made is in the parameters which are used to acquire vibration data.

When the first microprocessor-based vibration meter was developed in the early 1980s, the ability to acquire multiple blocks of raw data and then calculate an average vibration value was incorporated to eliminate the potential for spurious signals or bad data resulting from impacts or other transients which can distort the vibration signature. Normally, one to three blocks of data are adequate to acquire an accurate vibration signature. Today, majority of the programmes are set up to acquire 8 blocks to 12 blocks of data from each measurement point. These data are then averaged and stored for analysis.

The above methodology poses two issues. First, this approach distorts the data which is ultimately to be used to determine whether corrective maintenance actions are necessary. When multiple blocks of data are used to create an average, transient events, such as impacts and periodic changes in the vibration profile, are excluded from the stored average which is the basis for analysis. As a result, the analyst is unable to evaluate the impact on operating condition which these transients can cause.

The second issue is time. Each block of data, depending on the speed of the machine, needs between 5 seconds to 60 seconds of acquisition time. As a result, the time needed for data acquisition is increased by orders of magnitude. For example, a data set, using 3 blocks, can take 15 seconds. The same data set using 12 blocks then takes 60 seconds. The difference of 45 seconds do not sound like much until a person multiply it by the 400 measure points which are acquired in a typical day (5 man hours per day) or 8,000 points in a typical month (100 man hours per month).

Single-channel vibration instruments cannot provide all of the functions needed to evaluate the operating condition of critical production systems. Since these instruments are limited to steady-state analysis techniques, a successful predictive maintenance programme is also to include the ability to acquire and analyze both multichannel and transient vibration data. The ideal solution to this requirement is to include a multi-channel real-time analyzer. These instruments are designed to acquire, store, and display real-time vibration data from multiple data points on the machine-train.

These data provide the means for analysts to evaluate the dynamics of the machine and greatly improve their ability to detect incipient issues long before they become a potential issue.

Real-time analyzers are expensive, and some programmes in smaller plants are not be able to justify the additional cost. Although not as accurate as using a real-time analyzer, these programmes can purchase a multi-channel, digital tape recorder which can be used for real-time data acquisition. Several eight-channel digital recorders have the dynamic range needed for accurate data acquisition. The tape-recorded data can be played back through majority of commercially available single-channel vibration instruments for analysis. Care is to be taken to ensure that each channel of data is synchronized, but this methodology can be used effectively.

Vibration data is never to be used in a vacuum. Since the dynamic forces within the monitored machine and the system that it is a part of generate the vibration profile that is acquired and stored for analysis, both the data acquisition and analysis processes are always include all of the process variables, such as incoming materials, pressures, speeds, temperatures, and so on, that define the operating envelope of the system being evaluated.

Majority of the microprocessor instruments which are used for vibration analysis are actually data loggers. They are capable of either directly acquiring a variety of process inputs, such as pressure, temperature, flow, and so on, or permitting manual input by the operator. These data are essential for accurate analysis of the resultant vibration signature. Unless analysts recognize the process variations, they cannot accurately evaluate the vibration profile.

A simple example of this approach is a centrifugal compressor. If the load changes from 100 % to 50 % between the data sets, the resultant vibration is increased by a factor of four. This is caused by a change in the spring constant of the rotor system. By design, the load on the compressor acts as a stabilizing force on the rotating element. At 100 % load, the rotor is forced to turn at or near its true centre-line. When the load is reduced to 50 %, the stabilizing force is reduced by one-half, however, spring constant is a quadratic function, so a 50 % reduction of the spring constant or stiffness results in an increase of vibration amplitude of 400 %.

Heat and / or heat distribution is also an essential tool which is to be used for all electromechanical systems. In simple machine-trains, it can be limited to infrared thermometers which are used to acquire the temperature related process variables needed to determine the machine or system’s operating envelope. In more complex systems, full infrared scanning techniques can be needed to quantify the heat distribution of the production system. In the former technique, non-contact, infrared thermometers are used in conjunction with the vibration meter or data logger to acquire needed temperatures, such as bearings, liquids being transferred, and so on. In the latter method, fully functional infrared cameras are needed to scan boilers, furnaces, electric motors, and a variety of other process systems where surface heat distribution indicates the system’s operating condition.

The combination of the three technologies or methods is the minimum needed for an effective predictive maintenance programme. In some cases, other techniques, such as ultrasonics, lubricating oil analysis, Meggering, and so on, can be needed to help analysts fully understand the operating dynamics of critical machines or systems within the plant. None of these technologies can provide all of the data needed for accurate evaluation of machine or system condition. However, when used in combination and further augmented with a practical knowledge of machine and system dynamics, these techniques provide a predictive maintenance programme which virtually eliminates catastrophic failures and the need for corrective maintenance. These methods also extend the useful life and minimize the life cycle cost of critical production systems.

Predictive maintenance is more than maintenance. Traditionally, predictive maintenance is used solely as a maintenance management tool. In majority of the cases, this use is limited to preventing unscheduled downtime and / or catastrophic failures. Although this function is important, predictive maintenance can provide substantially more benefits by expanding the scope or mission of the programme. As a maintenance management tool, predictive maintenance can be used as a maintenance optimization tool. The programme’s focus is to be on eliminating unnecessary downtime, both scheduled and unscheduled, eliminating unnecessary preventive and corrective maintenance tasks, extending the useful life of critical systems, and reducing the total life-cycle cost of these systems.

Predictive maintenance technologies can provide even more benefit when used as a plant optimization tool. As an example, these technologies can be used to establish the best production procedures and practices for all critical production systems within a plant. Few of today’s plants are operating within the original design limits of their production systems. Over time, the products which these lines produce have changed. Competitive and market pressure have demanded increasingly higher production rates. As a result, the operating procedures which have been appropriate for the as-designed systems are no longer valid. Predictive technologies can be used to map the actual operating conditions of these critical systems and to provide the data needed to establish valid procedures which meet the demand for higher production rates without a corresponding increase in maintenance cost and reduced useful life.

Simply stated, these technologies permit plant personnel to quantify the cause-and-effect relationship of various modes of operation. This ability to actually measure the effect of different operating modes on the reliability and resultant maintenance costs provides the means to make sound decisions.

As a reliability improvement tool, predictive maintenance technologies cannot be beaten. The ability to measure even slight deviations from normal operating parameters permits appropriate plant personnel (e.g., reliability engineers, maintenance planners) to plan and schedule minor adjustments which prevent degradation of the machine or system, thereby eliminating the need for major rebuilds and associated downtime. Predictive maintenance technologies are not limited to simple electro-mechanical machines. These technologies can be used effectively on almost every critical system or component within a typical plant.

For example, time-domain vibration can be used to quantify the response characteristics of valves, cylinders, linear-motion machines, and complex systems, such as oscillators on continuous casters. In effect, this type of predictive maintenance can be used on any machine where timing is critical. The same is true for thermography. In addition to its traditional use as a tool to survey roofs and building structures for leaks or heat loss, this tool can be used for a variety of reliability-related applications. It is ideal for any system where surface temperature indicates the system’s operating condition. The applications are almost endless, but few plants even attempt to use infrared as a reliability tool.

Other than the mission or intent of how predictive maintenance is used in a plant, the real difference between the limited benefits of a traditional predictive maintenance programme and the maximum benefits which these technologies can provide is the diagnostic logic used. In traditional predictive maintenance applications, analysts typically receive between 5 days and 15 days of formal instruction. This training is always limited to the particular technique (e.g., vibration, thermography) and excludes all other knowledge which can help them understand the true operating condition of the machine, equipment, or system they are attempting to analyze. The obvious fallacy in this approach is that none of the predictive technologies can be used as stand-alone tools to accurately evaluate the operating condition of critical production systems. Hence, analysts are to use a variety of technologies to achieve anything more than simple prevention of catastrophic failures.

At a minimum, analysts need to have a practical knowledge of machine design, operating dynamics, and the use of at least the three major predictive technologies (i.e., vibration, thermography, and tribology). Without this minimum knowledge, they cannot be expected to provide accurate evaluations or cost-effective corrective actions. In summary, there are two fundamental needs of a truly successful predictive maintenance programme consisting of (i) a mission which focuses the programme on total-plant optimization and (ii) proper training for technicians and analysts. The mission or scope of the programme is to be driven by life-cycle cost, maximum reliability, and best practices from all functional organizations within the plant. If the programme is properly structured, the second requirement is to give the personnel responsible for the programme the tools and skills needed for proper execution.