Sie sind auf Seite 1von 5

Cost-Optimized Scheduled Maintenance Interval for ReliabilityCentered Maintenance

William R. Wessels MEVATEC Corporation Huntsville


Key Words: Reliability-centered maintenance, Scheduled maintenance interval, Predictive maintenance, Reliability, Maintainability, Availability SUMMARY AND CONCLUSIONS Implementation of a reliability-centered maintenance program is based on performance of proactive maintenance actions that serves to preserve system functionality. Predictive maintenance is a recognized proactive maintenance action that has been underutilized due to quantitative complexity and to the lack of a constraint that defined an optimum scheduled maintenance interval. A cost-optimized scheduled maintenance interval is proposed that uses costs as the constraint and overcomes quantitative complexity by use of current computer and software technology. Combined with condition-based monitoring the cost-optimized scheduled maintenance interval enables an organization that owns and operates fleets of assets to implement a comprehensive reliability-centered maintenance program. NOTATION A.................. System availability t ................... The independent variable time ................... The scheduled maintenance interval in time units RSMI(t) ......... System reliability with a scheduled maintenance interval RSYS()System reliability at time equals the scheduled maintenance interval k .................. The number of scheduled maintenance intervals that have occurred since the implementation of an RCM program RSYS(t-k) .... System reliability at time greater than the kth scheduled maintenance interval .................. The mean-time-between-failure, MTBF/meantime-between-replacement, MTBR MMTP ......... The predicted mean maintenance time for the system FCP ............... The predicted corrective maintenance actions per thousand hours for the system MCPT ............ The mean predicted corrective maintenance time for the system FPP ............... The predicted predictive maintenance actions per thousand hours for the system MPPT ............ The mean predicted predictive maintenance time for the system CMP .............. The predicted total maintenance cost CCP............... The predicted mean corrective maintenance cost CPP ...............The predicted mean predictive maintenance cost INTRODUCTION In 1975 mine maintenance foremen wondered why machinery and equipment corrective repairs could not be scheduled rather than occur randomly, often at the worst possible time, or in the worst possible situations. This was certainly not the first occasion that those responsible for maintaining machinery and equipment pondered on the concept of predictive maintenance as a panacea to the disruptions in operations resulting from unplanned corrective maintenance actions. There is a strong likelihood that a predictive maintenance concept was articulated in the earliest days of the industrial revolution as the advent of mechanical and electrical machinery and equipment replaced manual labor, hand tools and simple animal drawn vehicles as the primary mechanisms for production of goods and services. Therefore we can observe that the predictive maintenance concept is not new. Yet predictive maintenance has remained an elusive concept to implement. RCM Theory In their books on reliability-centered maintenance Smith and Moubray identified proactive maintenance as a key component of a reliability-centered maintenance program, to include predictive maintenance, condition-based maintenance, and scheduled inspections. Both authors emphasize that reliability-centered maintenance is the analysis, design, and implementation of proactive maintenance procedures that serve to preserve system functionality [1][2]. System functionality can be measured in terms of system dependability, which is, in turn, characterized by system availability and system reliability. The availability of a system is the probability that a system is in an operable state when its functionality is required or scheduled. The reliability of a system is the probability that a system will function to its performance specifications during the time that it is expected to operate. The dependability of a system is the product of the availability and reliability. The parameter of reliability is the failure rate, or its inverse, the mean-time-betweenfailure/mean-time-between-replacement, MTBF/MTBR. The parameters of availability are the mean-time-betweenmaintenance-actions, MTBMA, and maintenance time. The

0-7803-7717-6/03/$17.00 2003 IEEE

412

2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium

MTBMA is a function of MTBF so it can be said that both reliability and availability are dependent upon the failure rate [3]. The term failure rate begs the question, failure of what? When a system is not capable of functioning when its is scheduled to function, or ceases functioning during its scheduled operation, the system experiences a downing event resulting from the failure of a Lowest Replaceable Unit, LRU. This point needs to be emphasized; systems do not fail, LRUs fail and only LRUs. An LRU is defined by the organization. An LRU is that design level of the system that is removed to either be replaced or repaired. Replaced and discarded parts are logically identified as an LRU. An example includes a hydraulic hose on a mine haul truck, or an electronic circuit board in a radar cabinet. Parts that are removed and repaired, or repaired in place, often are sources of identity confusion. Where a part is removed for repair, and the repair consists of removal and replacement of internal parts, the part removed from the system is still identified as an LRU, and the parts removed and replaced within it are identified as shop Replaceable Units, SRU. But if the same part is repaired in place the parts removed and replaced are the LRUs. An example is a hydraulic pump. An organization that responds to the failure of a hydraulic pump on the system by removing and replacing it will identify the hydraulic pump as an LRU. Then removing and replacing internal parts, SRUs, repairs the failed hydraulic pump. But an organization that responds to the failure of a hydraulic pump on the system by removing and replacing a failed internal part, or parts, leaving the hydraulic pump in place during the maintenance procedures will not identify the hydraulic pump as an LRU. The internal parts would be identified as LRUs [4][5]. The conclusion is drawn that predictive maintenance must be applied to those LRUs that have failure modes, effects and consequences that result in a system downing event. Smith and Moubray appear to differ on the practicality of predictive maintenance. Smith suggests that the quantitative analysis is too difficult. Moubray provides a theoretical discussion of the quantitative analysis with a brief coverage of mean-time-between-failure, MTBF, or mean-time-betweenreplacement, MTBR. It is indeed fitting that a book on reliability-centered maintenance would acknowledge the importance of a parameter of system reliability as one of its criteria. Moubray introduces the concept of the P-F curve as an analytical approach to the design of a reliability-centered maintenance program (Fig 1 P-F Curve) [2].
f(t)
P F

Moubray shows that an LRU has an output function, as for any process point, that has an expected value over time until failure of the LRU is imminent. At some point in time following the initiation of degradation towards failure the prediction of failure can be determined, P, on the P-F curve. The time after the prediction time, P, when the LRU output function drops below an acceptable level and a failure state occurs, is the failure point, F. Moubray observes that proactive maintenance by inspection or condition-based monitoring will preserve system functionality when the P-F duration is long enough to detect the degradation and repair the LRU prior to an unscheduled LRU failure. An example of condition-based monitoring that identifies the prediction point, P, prior to the failure point, F, is the water temperature gauge on an automobile. As the gauge shows increases the operator is alerted to the potential for overheating in adequate time to avert loss of system functionality of the car. Moubray agrees that predictive maintenance is required to preserve system functionality when the P-F duration is too brief to allow proactive maintenance to prevent the LRU failure. In such cases condition-based monitoring and inspection are not effective. Consider again the automobile. Most models include tachometers that show the engine RPM as one is driving. One can suggest that this gauge is a condition-based monitor for the engine, but should the tachometer show the RPM drop to zero as the car is being driven the operator is not provided sufficient time to preserve system functionality. In this example preservation of system functionality can only be achieved through predictive maintenance. SMI Theory Predictive maintenance is defined as the calculation of the duration that an LRU can be expected to operate absent common causes of failure and the replacement of the LRU at or prior to that duration of use. This maintenance action goes against the nature of maintenance managers and practitioners because it calls for the replacement of an LRU that has not yet failed; often an LRU that shows no evidence of degradation. This is the quantitative approach Smith eschewed and Moubray avoided; not so Dr. Alessandro Birolini. Dr. Birolini developed the quantitative approach to describe the impact of the scheduled maintenance interval, SMI, on the reliability, MTBF, and availability of an LRU and the system. The expression for the system reliability using Dr. Birolinis SMI is presented in equation 1 [6].

RSMI (t ) = RSYS ( )k RSYS (t k )

(1)

Figure 1 - P - F Curve

The scheduled maintenance interval, , is the duration between repair or replacement maintenance actions. The number of intervals expected over the life of the system, k, is an integer value. The time since the last scheduled maintenance interval is the independent variable, time, t, minus the cumulative time preceding the last scheduled maintenance interval, k. The comparison between the reliability function and MTBF for a system without a

2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium

413

scheduled maintenance interval and the reliability function and MTBF for a system with a scheduled maintenance interval is shown in Figure 2 Reliability and MTBF [5].
1.0

R(t)

RSYS()kRSYS(t-k)

SMI = RSYS ( ) RSYS (t k )dt


RSYS(t)

RTF = RSYS (t )dt


0

Theoretically the availability of a system with a scheduled maintenance interval declines in a small magnitude from one scheduled maintenance activity to the next and as the system is repeatedly restored the availability returns to unity. The small magnitude of the decreases in availability over time between scheduled maintenance intervals is justified by the assumption that the decrease for any interval is comparable to the decrease from the condition of the system when new. The availability over time for a system that does not implement a scheduled maintenance interval shows a gradual decline in overall availability of the system. The increases in availability following each maintenance action does not reach unity because the system is not restored by the maintenance action. Therefore only the LRU that failed is removed and replaced while othe LRUs that are in a near failure state remain in place. Limitations Birolini did not show the approach to calculate the appropriate scheduled maintenance interval for a system. Instead Birolini allows the calculation of the impact on a system for a specified scheduled maintenance interval. But the question that demands an answer is, what is the optimum scheduled maintenance interval for a system? The reliability, maintainability and availability engineering and analysis disciplines of reliability-centered maintenance cannot alone provide the answer. The reason is that the reliability, maintainability and availability engineering and analysis for a system is not constrained. This fact is illustrated by an example. Assume a system is run to failure and LRUs are removed and replaced when each causes a downing event. Let that system have a mean-time-between-failure of 200 hours. If one introduces a scheduled maintenance interval at 160 hours the theory shows that the reliability will improve, resulting in an improvement in the system MTBF and availability. But if the scheduled maintenance interval was implemented at 120 hours the resulting system reliability, MTBF and availability would be more improved than for 160 hours. So too would a scheduled maintenance interval of 80 hours be more improved than for 120 hours. One could extend this iterative approach until the scheduled maintenance interval is 0 hours, that is, maintenance is performed continuously. The reliability and availability would each equal unity and the MTBF would be infinite. Unfortunately there would also be absolutely no utility derived from the system [4]. COST-OPTIMIZED SCHEDULED MAINTENANCE INTERVAL In a business application the system in the preceding example would have capital and operating costs and provide no productivity. Returning to the prime objective of reliability-centered maintenance, to preserve system functionality, the needed constraint to characterize an optimum scheduled maintenance interval can be determined. System functionaly serves the owner of a system with a measure of utility. In a business environment the utility of a

Figure 2 - Reliability and MTBF It is graphically evident that the reliability for a system that implements a scheduled maintenance interval in which LRUs are replaced and repaired prior to failure is significantly improved over the reliability for a system that is allowed to run to failure. Since the MTBF, , is the indefinite integral of the reliability function it is also evident that the MTBF for a system that implements a scheduled maintenance interval is significantly improved over the MTBF for a system which is allowed to run to failure. But the calculation of the MTBF for a system that implements a scheduled maintenance interval is impeded by the discontinuity of the reliability expression. Indeed solving for the MTBF tends to support Smiths contention that the quantitative approach to predictive maintenance is too difficult for practical application. But the advent of software programs that perform integral operations allows the calculation of the MTBF. The comparison between the availability function for a system without a scheduled maintenance interval and the availability function for a system with a scheduled maintenance interval is shown in Figure 3 Availability [5].
1.0

Ao

Figure 3 Availability

414

2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium

system can be measured by either productivity measured in marginal revenue or costs measured in marginal expenses. In a not-for-profit environment the utillity of a system can be measured by cost avoidance or costs minimalization. Since reliability-centered maintenance describes a business activity that is characterized as a cost center it is logical to define the cost function for a system as the constraint needed to characterize a scheduled maintenance interval for a system. The cost function for a system is dependent on direct and indirect expenses and capital costs. Direct costs include direct labor expenses, direct materials expenses, and direct overhead expenses allocated to the implementation of all maintenance activities performed to preserve system functionality. Indirect costs include indirect labor expenses, indirect materials expenses, and indirect overhead expenses allocated to the sustainment of a maintenance infrastructure. Capital costs include the depreciation expenses for the capital cost of the system and the cost of money for the capital investmant amount. There is another cost associated with the loss of functionality of a system lost opportunity. When a system is not functioning when it is scheduled to operate it accumulates lost productivity that can be characterized as unrealized marginal revenue. Lost opportunity is marginal revenue that is lost by the inability to operate the system because of a failure of an LRU. Combining reliability, maintainability and availability parameters of reliability-centered maintenance with a systems cost functions yields the cost-optimized scheduled maintenance interval. The theoretical primary feature of the cost-optimized schedule maintinance interval is that an organization will realize the optimum costs of operation of the system that correspond to the optimum marginal revenue from operation of the system yielding the optimum marginal profit. This objective serves as the motivation for maintenance managers and practitioners to change the way they think about maintenance procedures and practices and to embrace reliability-centered maintenance. Specifically, the cost-optimized scheduled maintenance interval should encourage maintenance managers and practitioners to remove and replace LRUs that have yet to fail in order to prevent random failures of the system. The costoptimized scheduled maintenance interval allows the maintenance foreman of thirty years ago to plan system breakdowns. Characterization of the cost-optimized scheduled maintenance interval is achieved by calculation and association of the costs of the parameters of maintainability that are realized by predictive maintenance scenarios. As the scheduled maintenance interval is allowed to vary from the baseline MTBF (calculated for a run-to-failure maintenance policy) to zero the cost-optimized scheduled maintenance interval is the global solution of the system cost function. Each value for the schedule maintenance interval determines the predicted mean maintenance time which is calculated as a function of the predicted number of corrective maintenance actions per thousand hours, Fcp, the predicted mean corrective maintenance time, Mcpt, the predicted number of predictive maintenance actions per thousand hours, Fpp, and the predicted mean predictive maintenance time,

Mptp. The equation for the calculation of the predicted mean maintenance time is the weighted average of the mean maintenance time, shown in equation 2.

MMTp =

Fcp M cpt + Fpp M ptp Fcp + Fpp

(2)

The logical objective for a reliability-centered maintenance program is to minimize the occurrences of corrective maintenance time. In an ideal reliability-centered maintenance environment the mean corrective maintenance time will approach 0 hours and all maintenance time will be predictive maintenance activities. The predicted maintenance cost, Cmp, is calculated as a function of the predicted number of corrective maintenance actions per thousand hours, Fcp, the predicted mean cost of corrective maintenance actions, Ccp, the predicted number predictive maintenance actions per thousand hours, Fpp, and the predicted mean cost of the predictive maintenance actions, Cpp. The expression for the predicted maintenance costs is shown in equation 3.

Cmp = FcpCcp + Fpp C pp

(3)

As with the mean maintenance time the objective for a reliability-centered maintenance program is to replace the costs for corrective maintenance with the idealized costs of predictive maintenance. As the scheduled maintenance interval decreases the corrective maintenance actions per thousand hours, Fcp, is expected to decrease and the predictive maintenance actions per thousand hours, Fpp, increases. An assumption of reliability-centered maintenance is that corrective maintenance costs are more than predictive maintenance costs. An understanding of the components of these two cost categories intuitively indicates this difference. Corrective maintenance costs include hard failure of LRUs often including destruction of the LRU and collateral damage to proximate LRUs, inconvenient location of the failure away from maintenance resources, spontaneous demand for maintenance and spare parts takes more time to perform, and lost opportunity costs. Predictive maintenance costs occur prior to hard failure preventing collateral damage to proximate LRUs. Often replaced LRUs can be repaired using rebuild kits that are less expensive than new LRUs. Scheduled maintenance activities allow coordinated scheduling of maintenance resources minimizing total maintenance costs. Spare parts can be stocked in time for the scheduled maintenance actions. Predictive maintenance does not incur lost opportunity costs. SUMMARY Reliability-centered maintenance is an innovation that was developed with the Boeing 474 airplane. It specifies the implementation of proactive maintenance condition-

2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium

415

based monitoring and predictive maintenance to preserve system functionality. While condition-based monitor technologies have flourished over the past three decades predictive maintenance has not. Initially, the technology did not exist to allow maintenance managers and practitioners to accumulate the data and perform the calculations essential to the characterization of a cost-optimized scheduled maintenance interval for a system. Although the data has always been there the capability to mine that data and organize the data in computational databases was absent until the emergence of the personal computer and database software programs. There has also been an emergence on analytical software that can perform analysis of nonlinear, discontinuous algorithms, to include a global evolutionary solution for the cost-optimized scheduled maintenance interval. The economic competitiveness of the global marketplace demands that organizations utilize all business practices that serve to optimize marginal costs. The cost-optimized scheduled maintenance interval can serve that demand. BIBLIOGRAPHY 1. Smith, Anthony M. (1993), Reliability-Centered Maintenance, McGraw-Hill, Inc., New York, NY. 2. Moubray, John (1997), Reliability-centered Maintenance, 2nd Ed., Industrial Press, Inc., New York, NY. 3. Kapur, K.C. and Lamberson, L.R. (1977), Reliability in Engineering Design, John Wiley & Sons, New York, NY. 4. Wessels, William R. [1998], Seeking an Optimal Scheduled Maintenance Interval: An Analytical Characterization of the Impact of Scheduled Maintenance Intervals on the System Reliability Model, ASQ 52nd Annual Quality Congress Proceedings, pp. 484-489, Philadelphia, PA. 5. Lawler, Patrick, 1993. Unpublished research and lecture notes, U.S. Army Missile Command, Redstone Arsenal, Huntsville, AL.

6. Birolini, Alessandro, 1994. Quality and Reliability of Technical Systems, Zurich, Switzerland: Swiss Federal Institute of Technology. BIOGRAPHY William R. Wessels, PhD, PE, CRE, CQE MEVATCE Corporation 4940 Research Drive Huntsville, AL 35805 USA wwessels@peo.mevatec.com Bill is a Principal Engineer employed by MEVATEC Corp., in Huntsville, Alabama. He has a PhD in Industrial and Systems Engineering from the University of Alabama in Huntsville, an MBA in Decision Science from the University of Alabama, and a BS in Engineering from the United States Military Academy at West Point. He is a registered Profession Engineer in Mechanical Engineering by the Commonwealth of Pennsylvania, and an ASQ Certified Reliability Engineer and Certified Quality Engineer. The author has researched and published his findings for the prediction of failures for mining machinery, biomedical devices, army aviation missile weapons systems and radars, and electronics process equipment since 1975. Bill and his wife, Tudor, live on a small farm with seven dogs, ever changing numbers of horses, donkeys, goats, geese, turkeys, chickens, and who knows how many barn cats. The beasts ignore the findings of statistically rigorous algorithms, defy any statistical level of significance, behave in a random walk manner, and ridicule his expertise in analytical methods applied to them. Tudor finds any attempt to predict their behavior to be a lack of common sense and a waste of time. But I continue to try.

416

2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium

Das könnte Ihnen auch gefallen