Advanced Technologies and Financial Benefits
A focus on monitoring critical asset health provides the basis to decrease life cycle costs of assets. The objective is to use reliability centered maintenance (RCM) methodologies to not only find the defects, but to eliminate as many defects as possible. A dynamic defect elimination program will analyze the failures and attempt to eliminate the defects. For example, if an asset is shutdown to fix problems that are detected and they are occurring three times per year, the root cause of the defect has to be eliminated. The possibilities to eliminate defects are many, but resources and advanced technologies are available. Today with mill cost ranging between $10,000 and $50,000 per hour, it is necessary to use the best tools to truly improve the process.
Life cycle costs can be greatly reduced if there is a focus on monitoring critical assets when new equipment is installed, but it is never too late to begin using these practices to improve an existing asset or process. Figure 1 shows the difference between organizations which do not effectively eliminate defects (Fig. 1a) and those which have succeeded in defect elimination (Fig. 1b). Late efforts represent ‘lost’ opportunities and ‘higher’ costs (red), while early efforts, often started prior to equipment installation, dramatically reduce overall cost.
Figure 1. (a) Defect elimination started late, (b) defect elimination employed early
It is not just about maintenance; maintenance organizations must work closely with operations and other departments to avoid down time and high asset costs. While unplanned down time is an easy metric is to talk about, there are other negative factors that occur to operations and steel mills in general. Rolling weeks are declared for product ranges to allow customers to order material in any manner to sustain their operations; when a mill incurs down time there is a good chance that the mill’s delivery performance will not be good. Today a mill needs a delivery performance very close to 100%. When a customer’s order is not received on time there is a very good chance that the customer will utilize competitive suppliers. Missing deliveries and becoming an unreliable producer must be avoided.
Furthermore, it is apparent that overcapacity in the global steel industry is a trend likely to continue for the foreseeable future. In order to retain and grow market share, companies will need to become more efficient with their resources and utilize the proper technologies and innovations to alleviate the need for capital spending. Transitioning to a reliability focus can provide companies with the means to accomplish this and it is not unusual to realize a 50 percent or greater reduction in maintenance spending. These reductions can result from the elimination of unnecessary maintenance tasks which may otherwise introduce defects into the equipment, and the application of condition-monitoring techniques which provide time for necessary corrective action to be taken while the scope of the repair is still small and collateral damage has not yet occurred.
The utilization of condition monitoring techniques also enables organizations to understand the operating condition of their equipment, providing enough advanced warning of problems to enable “just in time” parts procurement. Spare parts inventories can be reduced, releasing working capital for other uses and reducing carrying costs that can be 25-30 percent of the inventory value on an annual basis. It is not unusual to see organizations reduce their inventories by as much as 60 percent or more as they become highly reliable. Most importantly, organizations with a reliability focus typically realize higher asset efficiencies, meaning that the assets remain productive for a longer period of time, either between maintenance cycles or across the life cycle of the asset. Properly maintained (and properly operated) equipment will last significantly longer than equipment which is neglected or abused, and will allow for capital funds normally spent of replacements to be better utilized elsewhere.2 ASSET
Asset Management Strategies
Asset management programs can generally be categorized as the following; reactive (RM), preventive (PM), predictive (PdM), and proactive (PAM). In a reactive organization, equipment is expected to fail and resources (labor, material, spare parts, etc.) are assigned quickly to restore function. Preventive programs rely on discrete maintenance intervals, based on calendar days, cycles, runtime, or units produced; meaning maintenance activities can be performed regardless of the equipment condition. Predictive programs utilize condition based monitoring to determine when equipment requires maintenance to restore function. Proactive programs have the characteristics of a predictive program, but additionally seek to understand the root cause of equipment failures to ensure that future failures can be eliminated or the risks associated can be minimized.
A relative cost comparison of the different maintenance programs is shown in Figure 2a. Preventive maintenance costs are 29% less than reactive maintenance, predictive maintenance costs are 53% less than reactive maintenance and 33% less than preventive maintenance, and proactive costs are 77% less than RM, 66% less than PM and 50% less than PdM.3 Figure 1b shows a breakdown of the type of maintenance work performed that is characteristic of reliability focused organizations. Notably, 80% of the work being done is planned in advance, half of which is the result of findings from PM and PdM tasks. Reactive organizations will typically find the opposite to be true; that is only 20% of the work is able to be planned in advance. Also, reliability focused organizations only spend 15% of their time doing preventive maintenance tasks because unnecessary tasks (such as those based solely on OEM recommendations) are eliminated, and the use of condition based techniques negates the need to base maintenance intervals on time/cycles, etc.
The progression from a reactive to predictive and proactive maintenance program (i.e. a reliability focused organization) can be challenging and typically takes several years. However, the benefits that can be realized from high asset reliability are compelling and affect many other key business drivers, including: lower unit cost by distributing costs over a higher volume of product, higher production volume from fewer production interruptions (equipment failures), increased capacity without costly capital investment, increased market share through more competitive pricing, ability to expand into new markets when transportation costs are offset by lower manufacturing costs, fewer environmental incidents, and fewer injuries when fewer workers are exposed to hazards caused by equipment failure.4
As defined by the NASA Reliability Centered Maintenance Guide, RCM integrates Preventive Maintenance (PM), Predictive Testing and Inspection (PT&I), Repair (also called reactive maintenance), and Proactive Maintenance to increase the probability that a machine or component will function in the required manner over its design life-cycle with a minimum amount of maintenance and downtime. These principal maintenance strategies, rather than being applied independently, are optimally integrated to take advantage of their respective strengths, and maximize facility and equipment reliability while minimizing life-cycle costs. The goal of this approach is to reduce the Life-Cycle Cost of a facility to a minimum while continuing to allow the facility to function as intended with required reliability and availability.5
Figure 2. (a) Comparative costs of maintenance programs and (b) the desired distribution of maintenance activities.
SMRP (Society for Maintenance and Reliability Professionals) has established two metrics to help facilities compare the expenditures for maintenance and the value of stocked maintenance inventory with other plants of varying size and value, as well as to benchmarks. These metrics are defined as the ‘Total Maintenance Cost as a Percent of Replacement Asset Value (RAV)’ and ‘Stocked MRO Inventory Value as a Percent of RAV’. The RAV is used in the denominator to normalize the measurement because plants vary in size and value. SMRP defines RAV as the dollar value that would be required to replace the production capability of the present assets in the plant; including production/process equipment, as well as utilities, facilities and related assets.6 Top performers in the steel industry are actually spending less on maintenance and inventory, while achieving the highest equipment availability. In such facilities, maintenance spending is less than 2% of the plant replacement asset value (RAV), while other organizations spend 7-20%.(7)
Top performers also have less MRO (maintenance, repair, operations) inventory because they have optimized their inventory needs and utilized techniques to better detect the onset of failure. This characteristic results in increased time to order replacement parts and also to perform corrective maintenance for a failure which has not yet become catastrophic and damaging to ancillary equipment/components.
Consider a facility spending $15MM USD on maintenance for a $100MM USD facility, equating to a 15% total maintenance cost per RAV. If the facility moves to a top quartile performance (2% $Maint/RAV), this will yield annual savings of $13MM USD. Additional financial benefits can be realized through the reduction of inventory once a sustainable program has been implemented. Note that in conjunction with the metrics defined above, top performance is also related to high equipment availability. Simply cutting maintenance and inventory expenditures will not result in a sustainable asset reliability program. The metrics for maintenance and inventory expenditures as a percent of RAV are lagging indicators for how organizations are maintaining their assets. Leading indicators common among top performers include the following(8):
- A high percentage (>90%) of assets are monitored with predictive technologies such as vibration, thermography, and oil/lubrication analysis, where applicable
- The number of preventive maintenance work orders has decreased significantly in comparison to lower quartile performance, as predictive technologies are responsible for much of the maintenance work generated.
- Bill of material records are populated for assets in the CMMS, enabling effective and accurate job planning
Resources (time, money, people, effort, materials, etc.) are not boundless and as such the use of rigorous reliability analyses, the application of condition monitoring techniques, or even performing preventive maintenance tasks are not justified for every asset in a facility. For some assets, running them to failure is a sound business decision. To develop effective maintenance strategies and allow for work order prioritization, a criticality ranking of the assets should be conducted. While certain assets are widely accepted as critical based on their value and function, the balance of assets may not be readily categorized.
Criticality ranking establishes categories (e.g. Safety, Environmental, Production, Maintenance, and Quality) and weighted criteria within each category as a means to differentiate assets on a numerical basis. Assets are ranked at the lowest functional location in the hierarchy so that a group of assets in a functional location are assigned a singular ranking. This methodology provides granularity in the ranking that will aid in work prioritization and identifies where resources should be allocated to provide maximum benefit. Note that an accurate master equipment list is a prerequisite for effectively conducting the ranking process. Once critical assets have been identified, organizations are able to better focus efforts at reliability improvements and defect elimination including: RCM analyses, failure modes and effects analyses, root cause analysis, operator and craft training, and the deployment of condition based technologies.
Each asset type will have different failure modes and it is important to understand how failures progress over time so that the correct measures can be enacted in order to eliminate failures or mitigate the associated risks. Traditionally, the probability of equipment failure is expected to increase as the equipment ages and wears out. However, studies have concluded that only 11% of failures are age-related. The reality of equipment failure is that the overwhelming majority (72%) of failures are the result of infant mortality.9 This indicates that most equipment failures are the result of installation errors, design issues, and most importantly, intrusive preventive maintenance tasks that can introduce faults into the equipment.
Once a failure has begun to occur (Figure 3a) there is a subsequent degradation in equipment health that can be predicted as a function of time. Early in the failure progression, there are notable changes in vibration, followed by an increase in wear debris in the lubrication. Further degradation results in heat being generated which is detectable using IR thermography and audible noise coming from the equipment. Just prior to functional failure, it may be possible (but not safe) to feel the heat being generated from the fault. The last point in the curve beyond the functional failure represents catastrophic failure. It is important to note that predictive technologies, when applied correctly, are capable of detecting failures early in the progression. If acted upon, repairs can be planned and scheduled with adequate time to gather any necessary replacement parts and other resources.
Figure 3. (a) P-F curve (b) example of bearing damage (early in failure progression) readily detectable with PeakVue.
There are numerous technologies that can be deployed to detect the onset of equipment failure, including vibration analysis,infrared imaging (thermography), ultrasonic testing (airborne & contact), motor testing, lubrication analysis/wear debris,thickness measurement/imaging, and process parameter monitoring. Vibration monitoring is widely regarded as one of the most effective methods to ascertain machine health.10 Vibration is the movement of a body about its reference position; for rotating equipment this is generally the centerline of the shaft. Movement, generated from an excitation force such as a defect, is detected with a transducer (accelerometer, velometer, or displacement probe). Generally, waveforms from the transducer can be used to quantify the severity of defect based on the amplitude; and spectral analysis using FFT (Fast Fourier transforms) can be used to identify the dominant frequencies within the waveform so that they can be correlated with known fault frequencies. Regardless of the technology used it is important to understand the limitations and ensure that potential failure modes are addressed using RCM methodologies.
Critical equipment in a steel mill such as rolling mills, shears, cranes, coilers, turrets, etc. present unique challenges for effective condition based maintenance. As previously noted, the vibration signatures of equipment can provide some of the most useful information regarding the health of the machinery. However, vibration monitoring in the traditional sense is not always best suited for several reasons, including: equipment size, accessibility, process variability, varying and low speeds, and intermittent loading.
Due to the large mass of the machinery and distance from the vibration source to the outer casing where the sensor is installed, it is difficult to detect vibration energy from defects early in their failure progression. Additionally, the equipment is often operated at a relatively low speed can add complications since a standard accelerometer is only linear down to 30
RPM (0.5 HZ), a low frequency accelerometer is only linear down to 12 RPM (0.2 HZ). Also, note that vibration amplitude is proportional to the cube of the speed, meaning that lower speeds produce significantly lower amplitudes.
In order to effectively detect machinery faults for this type of equipment, it is important to understand the difference between macroscopic vibration and stress waves. Macroscopic vibration is the result of a forcing function (such as a bearing defect) transferring energy through the machine, causing physical movement which detected by a case mounted vibration sensor. Typically, a defect would have to be relatively large and far enough into the failure progression to create translational movement of a large-mass machine. In those instances, audible noise from the fault may be detected and vibration monitoring may only serve to pinpoint the source.
Stress waves are short duration, high frequency events produced by metal to metal impacting which introduce ripples on the surface of machinery. These transient events can be the result of fatigue cracking, scuffing, and abrasive wear, and other common rotating machinery faults. In normal spectral analysis of vibration, the signal is processed with a high-order, low-pass filter to prevent aliasing. The use of an anti-aliasing filter will also remove any short term transient events, making this methodology ineffective at capturing stress waves.11
PeakVue is an advanced analysis methodology capable of detecting stress waves with the use of standard accelerometers. This methodology separates the stress waves from the macroscopic vibration using a high-pass filter, determines the absolute peak value of the signal over each time increment specified by the analysis bandwidth, rectifies and amplifies the data, and then uses FFT to provide spectral analysis. Studies have shown that PeakVue is effective at detecting machinery faults for equipment running at 0.5 to 10 RPM11. Some machinery faults are still readily detected with macroscopic vibration, so both can and should be utilized. Rains provides practical information on measurement setup, sensor selection and placement, the importance of trending, and severity assessment.12 Because PeakVue analyzes stress waves, it is effective at detecting failures very early in the progression (as evidenced in Figure 3b). Unlike other stress wave measurements, the maximum peak level can be trended as an indication of an impending problem.
Intermittent loading is another common challenge for steel mill equipment because the vibrations measured on a machine will change depending on whether or not it is loaded. Rotational forces alone, despite the equipment mass, generally do not provide enough excitation to effectively capture machinery faults with vibration analysis. Figure 4 shows comparative data from the same collection point prior to and during loading of the mill stand. While the stand is unloaded the vibration energy remains very low as shown in the time waveform (Fig. 4a) and reduces the peaks shown in the spectral analysis (Fig 4b). Under loading, the gear mesh harmonics (indicating gear misalignment) are readily identified.
Figure 4. Comparison of vibration (a) waveform and (b) spectrum in loaded (purple) and unloaded state (blue).
Measuring vibration on intermittently loaded equipment also requires events such as the head-end impact of material entering the work rolls is accounted for, as large impacts can make spectral analysis difficult. In Figure 5a, the time waveform of the recorded data shows the resultant vibration present on the high speed pinion bearing of the reduction gearbox as a slab impacts the work rolls. The magnitude of vibration during impact is approximately 50 times higher than during normal operation (i.e. steady-state rolling conditions), and is well above the established alert and alarm limits for the application. However, this is not necessarily an indication of a machinery fault and is only the result of how the equipment operates. The FFT data used for spectral analysis, shown in Figure 5b, takes on a “ski-slope” shape and any dominant frequencies within the waveform are blurred and indiscernible due to the impact.
Figure 5. (a) Time waveform and (b) spectrum of vibration recorded during head-end impact of slab.
In online (continuous) monitoring systems, variables such as torque, motor current, speed, or another external source may be used to trigger the collection of vibration data thus distinguishing between loaded and unloaded conditions. The time which the equipment is loaded also needs to be considered as it is beneficial to record the loaded condition over several revolutions of the shaft for effective analysis. To accurately identify closely spaced fault frequencies and to avoid smearing of the spectral data, measurement parameters need to be configured to provide adequate resolution.
Process parameters such as grade, temperature, width, head-end geometry, scale, etc. can affect equipment health. Abnormalities in these variables or the operation of equipment outside of its design envelope can produce excessive stresses, causing mechanical fatigue or sudden yielding of components. Understanding how these various characteristics affect the equipment will aid in the selection and use of condition monitoring techniques. As shown in Figure 6a, the rolling of hard material at the maximum design width created excessive vibration in the bull gear thrust bearing, in excess of the prescribed alarm levels. While these vibrations are not correlated with specific fault frequencies in the spectral data, it is a sign of distress in the equipment. Production of harder grades and wider slabs, subjected the equipment to high loads which in some cases can tend to create a binding effect on the mechanical components, producing large axial forces, as detected in the thrust bearing. Continued loading in this manner may likely initiate bearing failures. As noted previously, it is important to not only find failures after they have developed, but to take a proactive approach in identifying conditions/events which are the root cause of the failure. Misaligned couplings (tooth wear pattern shown in Figure 6b) are often capable of producing these types of axial forces under loading. In this case, addressing the misalignment problem, rather than the high bearing vibration, provides an effective method for eliminating defects.
Figure 6. (a) bull gear thrust bearing vibration trend, (b) gear misalignment responsible for “binding” effect.
Varying equipment speed as required for certain products typically necessitates the use of technique called order tracking. Otherwise, the peaks which correlate with different faults will become smeared across several frequencies, making it difficult to distinguish between different faults. This technique automatically adjusts the data collection to account for changes in shaft speed and requires a tachometer signal. As opposed to normal averaging, order tracking will convert the frequency values to orders before they are averaged.
Speed changes can also result in resonant conditions, which is when a forcing function coincides with a natural frequency. The source of the forcing function can be misalignment, imbalance, electrical line frequency, gear meshes, etc. Resonance is capable of producing high amplitude vibrations and can occur in the radial, axial, or angular (torsional) orientation depending on the specifics of the equipment dynamics. As shown in Figure 7a, large torsional vibrations were produced during the rolling of certain materials as speed increased during entry of the strip into the mill stand, and excited the first torsional natural frequency. Note that even for short duration excitation of the fundamental natural frequency, high amplitude vibration (torsional in this case) can result. These events were determined to be the root cause of fractured shaft keys in the drivetrain.(13)
Figure 7. (a) high amplitude torsional oscillations due to resonant conditions, and (b) torsional oscillation in finishing mill.
Torque monitoring is becoming more prevalent in the steel industry as users need to quantify the effects of rolling advanced high strength steels and other materials that the equipment may not have originally been designed to produce. Strain gage telemetry systems typically offer users the most cost-effective solution for temporary or continuous duty, and can be powered inductively or with batteries (6 month lifetime readily achievable).
Torque measurements can provide valuable quantifiable data regarding the effects of varying process conditions on the mechanical equipment in a drivetrain. Certain conditions can produce high magnitude torque spikes, or torsional vibrations, both of which can be damaging and are difficult to detect with any other methods, including monitoring of motor armature current. Sudden torque spikes are often the result of improper steel temperature, abnormal head-end geometry, or operator/setup error and can create stresses in excess of the material yield strength or endurance strength.
Vibratory torque, as shown in Figure 7b can be damaging to the product quality and drivetrain components. In this case, torque sensors had been installed on the gear spindles between the work rolls and pinion stand in an effort to determine why fasteners securing power transmission components were loosening in operation and causing component fracture. Just as with linear vibration, FFT analysis can be performed on torsional vibrations to identify the dominant frequencies in the waveform, although there are usually only one or two dominant frequencies present which typically are less than 50 Hz for rolling mill applications. The dominant frequency of the torsional vibration was found to be the gear mesh frequency of the pinion stand gearing and excessive wear/backlash was creating these oscillations which were in turn responsible for loosening fasteners on the gear spindles.
Variable frequency drives are another common source of torsional oscillations. Often, the electrical torque calculated from the motor current does not provide a good representation of the mechanical torque on the drivetrain components. This results in the misconception that the drivetrain is operating without problems because high magnitude torsional oscillations are undetectable by the electrical torque values alone.14
For critical assets that are intermittently loaded, operate at varying and low speeds, and are subjected to varying process conditions, online monitoring systems provide the most effective way to monitor machinery health. Online systems are also well suited for these applications because they eliminate the need to send personnel into the field to access monitoring points with a handheld analyzer that are not safely accessible.
One such system installed on the roughing mill stands was designed so that numerous machinery conditions and faults (previously attributed to significant downtime) including fatigue failures, poor lubrication, gear wear, and bearing failures could be detected early, allowing repairs to be completed during scheduled outages. To accomplish this, accelerometers were installed at each bearing location in the radial and axial orientation. In addition to detecting macroscopic vibration, PeakVue technology was also utilized because the low speed section of the drivetrain operates at less than 20 RPM. Tachometers used to measure rotational speed were installed at the motor coupling shafts and torque sensors (wireless strain gage telemetry systems) were installed on the gear spindles operating between the work rolls and pinion stand (Figure 8a). Field wiring from the sensors was brought into a data processing unit (Figure 8b), which analyzes the vibration, torque, and speed signals. The data processors are connected to a server over a local area connection, which allowed plant-wide and remote access to the real-time data.
Figure 8. (a) sensor placement on drivetrain, (b) data processor installation in motor room.
Torque spikes generated at the work rolls were found to be responsible for several failures of the gear spindles. Continuous monitoring of torque needed to be employed because spikes will only occur during operational abnormalities and are not typically repeatable. Furthermore, the resultant damage (typically initiation and propagation of cracks) is difficult to detect because that energy would have to be transmitted a long distance from the gear spindles to the nearest accelerometer. A continuous torque trend and magnified view of several torque spikes at head-end impact is shown in Figure 9. Alerts from these events prompted the mill personnel to remove the spindles for inspection, as it was suspected that damage had occurred. Following the removal of the gear spindles during a scheduled outage several weeks later, the results of a magnetic particle inspection are shown in Figure 10a. As highlighted, several cracks had initiated in the roll sleeve bore surface. While salvaging the components provides some financial benefit, the greatest benefit is realized from avoiding unplanned downtime. In certain market conditions, these failures can cause up to 8 hours of downtime, and depending on market conditions can equate to $250 to $1100 per minute. More severe torque spikes can create further damage to the gear spindle components, rendering them useless (Figure 10b).
Figure 9. (a) spindle torque trend, and (b) magnified view of spikes recorded at head-end impact
Figure 10. (a) small subsurface cracks initiated in gear spindle roll sleeve, (b) severe crack propagation.
Upon commissioning of the system, it was apparent that one of the bearings installed in the pinion stand was running roughly (Figure 11). Vibrations were recorded above the prescribed alarm limits, however a baseline had not been established yet due to the limited data recorded (less than 1 month) in the condition monitoring system. It should be noted that the macroscopic vibration measurements showed little to no increase in the trend as a result of the low vibration energy generated from the bearing fault, as seen in Figure 11b. Only PeakVue measurements (stress waves) were able to clearly indicate a problem with the bearing. During the next scheduled outage, the bearing covers were removed so that a visual inspection could be completed, revealing a cracked roller (Figure 12a). Upon removal of the bearing, the fracture surface was inspected and indicated that fatigue cracks had initiated at the inner diameter of the roller and progressed outward (Figure 12b).
Figure 11. Peak-Peak and overall trends of (a) PeakVue and (b) normal velocity (macroscopic)
Figure 12. (a) cracked roller found during inspection, (b) fatigue cracks originated at inner diameter of roller.
The use of other technologies can help to validate the results and ensure that the failure has been properly identified. For rotating equipment, infrared (thermographic) inspection can help to quickly identify poor lubrication and installation issues which generate heat from increased friction (Figure 13). Lubrication analysis can identify and quantify the amount of wear particles and debris present in lubrication, eliminating the need to access lubrication cavities during a PM which can serve to introduce failures into the components.
Figure 13. Thermographic image of (a) universal driveshaft to pinpoint incorrectly installed components and (b) gearbox
When predictive technologies are applied properly to equipment impending failures can be identified earlier in their progression. However, it is important to understand that each technique and technology has limitations and there is no “silver bullet” for assessing machinery health. Many times, multiple technologies need to be utilized to detect the numerous potential faults, and to verify the findings of other technologies. Also, it’s important to realize that the inherent reliability of a system is a function of design; a poorly designed system will never be reliable, and no amount of maintenance investment will improve it. Maintenance actions can either preserve this inherent reliability or destroy it.2
Advanced technologies aimed at detecting potential failures early in their progression should be utilized for critical assets and applied based on RCM methodologies. The deployment of these technologies within a facility is typically an easy task given the vast resources available today but alone do not provide for an effective and sustainable asset management strategy. It’s been seen that nearly 50% of programs are not achieving the anticipated return on investment. Many times this is due to the fact that reliability initiatives are not properly viewed as a business strategy, and are not integrated throughout the organization. Too often, the maintenance organization is responsible for reliability initiatives, the IT department is responsible for the CMMS, etc.8 Success and sustainability of a reliability program largely depends on an organizations ability to integrate these condition monitoring technologies into their business. Only limited success and financial benefit will be realized if these initiatives are viewed solely as a maintenance initiative.
- DuPont-Ledet (DuPont Corporation), Study on of Detect Elimination and Life Cycle Cost, 1994.
- Emerson Process Management, The Financial Benefits of Reliability, White paper, June 2015.
- EPRI Power Generation Study on Maintenance Costs.
- Al Poling., The Business Case for Asset Reliability, Maintenance Technology Magazine, March 2015, pp 16-21.
- National Aeronautics and Space Administration, Reliability Centered Maintenance Guide for Facilities and Collateral Equipment, September 2008.
- Society for Maintenance and Reliability Professionals, Business and Management Metrics 1.4 and 1.5, 2012.
- Internal Benchmarking Study across 14 Industries, Management Resources Group, Inc.
- John Schultz, Robert DiStefano, The Business Case for Reliability, AISTech 2004.
- United States Navy, Reliability-Centered Maintenance (RCM) Handbook S9081-AB-GIB-010, April 2007.
- Ramesh Gulati, Maintenance and Reliability Best Practices, Industrial Press Inc., New York, 2009.
- David Stobbe, Marc Phillips, James Robinson, Capture and analysis of stress waves provides significant improvement in condition monitoring of critical rotating machinery, Iron and Steel Engineer, July 1999.
- Daniel Rains, Practical Approach to PeakVue, Vibration Institute Training Conference, June 2013.
- Dan Phillips, Dennis Craig, Understanding TAFs and Their Effect on Drivetrain Maintenance and Reliability,AISTech 2014.
- Dan Phillips, Julian DelCampo, Chandra Deshpande, Transient Torque Measurement, Modeling, and Effect on Drivetrain Reliability, METEC Dusseldorf, 2015.