Determining AGE levels: Requirements-based methodology. (Inventory Leveling Within AMC).
Aerospace ground equipment (AGE) is used to service aircraft while on the ground. The aircraft maintainer's job is to ensure aircraft are serviced and repaired expeditiously, thereby maintaining high percentages of the aircraft fleet in mission-ready status. The desire to have these high aircraft mission-capable rates has resulted in keeping high inventory levels of everything imaginable necessary to sustain the aircraft. No maintainer wants to see aircraft mission losses due to a lack of functional AGE. Thus, to mitigate the impact of potentially unreliable AGE, excess AGE inventory is the norm.
The level of AGE (or any support equipment) at a given location is determined by that location's table of allowance authorization. Currently, the Air Mobility Command (AMC) queries subject matter experts (SME) to determine the table of allowance authorizations for AGE. This is done base by base, with unit type codes and the mission requirements of each base determining the final total allowance authorization. As the Air Force enters the 21st century, it must reduce excess assets in the most effective manner possible.
By closely examining what is actually required to support aircraft, the Air Force can identify excesses and shortfalls to obtain maximum utility from limited resources. Considering the reliability of AGE units is a major input into the number of units required due to potential reliability problems with either newly deployed or aging systems.
AMC is purchasing a new nitrogen system, the Self-Generating Nitrogen Servicing Cart (SGNSC). The method AMC currently uses to determine AGE levels seems to overstate need, and the purchase of a new nitrogen system for use throughout the Air Force, with AMC as the lead command, is an excellent opportunity to compare current practice with a more analytical approach to determining AGE levels. The intent, of course, is to reduce AGE levels without impacting aircraft mission-capable rates.
Scope of Research
The primary purpose of research for this article was to define and demonstrate a methodology for assessing AGE utilization in a given scenario while noting any impacts on mission capability. The goal was to use this quantitative methodology to size an AGE fleet to meet aircraft demand at a base of study and perform sensitivity analysis on any maintenance delay if the aircraft must wait for AGE assets.
To analyze the impact of SGNSC and limit confounding effects of possibly redundant variables, variables such as maintenance and fuel were modeled as unconstrained resources.
The initial SGNSC contract was a $20M effort with an estimated 570 units at $35K each, not including operations and maintenance costs.(1) While it did not compare with the B-1 program, it did have potential for cost reductions and was worth examining. Further, these were procurement costs only and did not include other costs such as reliability, maintainability, and mobility or deployment.
This article does not examine the impact of other AGE on aircraft availability. It only addresses the impact of SGNSC through comparison of distributions of different variables against each through a queuing simulation to determine range of utilization and size of SGNSC inventory to accommodate mission requirements.
AGE is used for servicing and maintenance of aircraft on the ground and is a relatively inexpensive way to maintain aircraft compared to using systems on board the aircraft. It may be readily replaced without impacting aircraft mission capability. AGE is a necessary part of the flight-line environment, and some form of AGE is almost always used in the performance of aircraft maintenance.
SGNSC is a self-contained, powered AGE unit that uses outside air to refill nitrogen storage cylinders. The cart takes outside air, builds pressure, and filters nitrogen through a membrane into the storage cylinder. The nitrogen is retained in the cylinder until discharged. The system is entirely self-contained and does not require refilling from outside sources, saving time and money, while increasing safety.
The Logistics Composite Model (LCOM) is a discrete event-queuing simulation. It was developed in the 1960s by the Air Force Logistics Command and the Rand Corporation to analyze maintenance processes. (2) In 1970, the Tactical Air Command used LCOM to determine maintenance manpower requirements for a squadron of F-4E aircraft. The end result, according to the final report--proof positive through the actual operational unit flying, a schedule developed through LCOM--was a valid model for determination of manpower requirements. (3) In 1992, the Aeronautical Systems Center (ASC) conducted a study for the F-15E Eagle Century Plus Radar Program. ASC combined this study with a validation effort for LCOM, which compared LCOM predictions with actual results. As can be seen from Table 1, the LCOM model conformed very closely to actual sortie rates and APG-70 actions, with the difference between the model and the real world close to or less than 1 percent.
LCOM results were also compared to F-15E operations at Luke AFB, Arizona, for a 56-day period with the results presented in Table 2. The results, again, were very close to the real world.
LCOM has been selected by numerous system program offices (SPO)--including but not limited to the B-2, F-22, joint strike fighter (JSF), and C-17 SPOs--for use in determining supportability requirements. (6) LCOM was formally accredited by the JSF SPO as a satisfactory supportability model to analyze sortie generation rate, manpower, support equipment and facilities, spares, prognostics and health management, cannibalization, and resource constraints. (7)
Current AGE BOI and Utilization
Interviews with AMC AGE personnel revealed the AGE basis of issue (BOI) is currently determined by subject matter experts with field experience. The SPO for the weapon system meets with the command headquarters AGE representatives and the AGE management agency from Robins AFB, Georgia. They review AGE usage at the bases where the weapon system is maintained and then negotiate the AGE table of allowances based on estimated future usage.
Currently, all AGE utilization is very low. Metered hours per cart pointed to an overabundance of AGE, possibly even an overabundance for surge situations, which is a worst-case scenario for flight-line operations and aircraft maintenance.
LCOM, a simulation model, with stochastic inputs from several sources, drove demand for SGNSC and determined capacity and utilization. Standard flight schedules determined the potential population of aircraft requiring SGNSC support. Work unit codes for each aircraft type were used to address variance in demand characteristics and differences in SGNSC utilization by airframe.
Sensitivity analysis was performed on the reliability of the SGNSC system. Between telephone interviews and correspondence with the San Antonio Air Logistics Center engineer, the mean time between failures (MTBF) for SGNSC, as a new system, was estimated to be about 500 hours. The mean time to repair (MTTR) was estimated to be 2 hours. SGNSC failures were modeled using an exponential distribution with a MTBF of 50, 100, and 500 hours. LCOM repair times were modeled using a lognormal distribution with a standard deviation of 29 percent of the mean.
First Lieutenant Jeff Havlicek raised the importance of addressing travel time in an AGE study and suggested the variability of travel times could have a statistically and practically significant effect on mission effectiveness. (8) He used two constant travel times of 15 and 45 minutes. (9)
To model travel times, a delivery delay was incorporated into the LCOM model. Travis AFB, California, tracks AGE delivery times, and according to the latest information available, 80 percent of AGE deliveries were within 10 minutes, and 99 percent of AGE deliveries were within 20 minutes. A minimum delivery time was unavailable as was the exact distribution. An assumption was made that 100 percent of the time maintenance would call for SGNSC support 10 minutes prior to actually needing the SGNSC. The travel time was modeled in LCOM with a notional minimum travel time of 5 minutes and another point at 10 minutes. Eighty percent of the delivery times were linearly interpolated between 5 and 10 minutes. The remaining 20 percent of the delivery times linearly were interpolated between 10 and 20 minutes, with the upper bound set at 20 minutes. AGE delivery drivers were assumed to be available when needed.
In terms of modeling users of the SGNSC resource, aircraft were the SGNSC users or calling population. Transient aircraft use very little nitrogen and could be adequately served with one primary SGNSC and one spare, for a total of two SGNSC. Transient aircraft typically experience temporary repairs or failures until they can get to home station for a permanent fix. Assuming transient aircraft can be adequately serviced with two SGNSC, one primary and one spare, this study excludes transient aircraft and concentrated on the demands of Travis' C-S and KC-10 aircraft. This simplified the model and facilitated extensibility of the methodology to other bases, aircraft, and AGE. This methodology extensibility was a primary consideration for this research effort.
Aircraft preflight status was given a higher priority for nitrogen than all other tasks requiring nitrogen, allowing the preflight aircraft to preempt other tasks that require nitrogen, similar to what would happen during a red streak or short-notice, high-priority maintenance on a flight line if there were not enough resources to go around. If this happened on an actual flight line, the lower priority task would be preempted to service the flyer. The LCOM model accurately reflected this situation. (10)
G081 (Gee-oh-eighty-one) was the maintenance data-collection database for heavy aircraft and was key to the success of this effort. A page-by-page review of all applicable technical orders for each airframe was beyond the scope of this study. As a result, subject matter experts familiar with the airframe were interviewed to determine the work unit codes requiring SGNSC support. If the work unit code (WUC) required SGNSC support, WUCs were used and matched against SGNSC requirements. Data from G081 were gathered by aircraft type. One issue with G081 was the time necessary to complete the maintenance task. This includes all maintenance, not just the time necessary for nitrogen servicing. Only the total time is collected in the maintenance data system. For those WUCs, field interviews were used to determine appropriate nitrogen service times.
The WUC was the initial data flag. Each job included the WUC and time taken for the repair job. The time taken to complete the job determined the mean time for the job length. An assumption of unconstrained maintenance availability was necessary to focus on analysis of the changes to the SGNSC quantities. Data collected from G081 were the actual number of occasions that systems requiring nitrogen were serviced. Distributions were based on these maintenance intervals. It was assumed maintenance was available according to the same priority schedule and nitrogen would be required in a similar manner. This assumption may or may not hold in a wartime environment; however, it was necessary as data for wartime consumption were not available. Data were aggregated to the fleet for an overall distribution.
Failure data were extracted from G081 by WUC and aggregated to include the number of failures, MTTR, and the mean time to service, as nitrogen consumption is not necessarily required for the entire task time. This was an acceptable assumption, as it reflected reality on the flight line; technicians do not call for the nitrogen cart until they require it. A majority of the components that require nitrogen servicing are part of the aircraft landing gear system, and failures are more accurately reflected if defined by number of landings as opposed to the standard number of flying hours. Modifications to the database accommodated this failure pattern. Historical aircraft arrivals at Travis were compared to the number of failures recorded in G081 for the same period to arrive at the number of failures per number of landings. An exponential distribution was used to model the failure rates of these nitrogen systems. The failure rates, as determined by system, with task and service times are given in Tables 4 and 5. Basic postflight (BPO) and preflight (PRE) service intervals were interpreted through interviews with flight-line personnel, as BPO/PRE nitrogen servicing is often undocumented.
The percentage and number of canceled missions are a more immediate, readily identifiable reflection of AGE availability on mission effectiveness than flight sortie effectiveness . If an aircraft mission is canceled, then there is a very real penalty for not having AGE available. All other resources were assumed to be unconstrained to isolate SGNSC and allow analysis of SGNSC effectiveness. Flight sortie effectiveness or mission capability is not as closely related to AGE availability. It is the author's opinion that mission capability can suffer some, but the cost of AGE is not comparable to the cost of a lost mission. As a result, the number or percentage of canceled missions was examined for statistical and practical significance.
Data on utilization of AGE were collected to give users an expectation of usage. The proposition of an overabundance of AGE was addressed examining utilization and AGE wait time. At issue was not necessarily utilization, although this gives the decision makers an idea of usage, but the ability of AGE to meet mission requirements was at issue. However, focus on utilization does not consider the impact of multiple requests. The capacity to handle periods of high demand was expected to be the main driver of AGE and a natural means for sizing an AGE force such as SGNSC.
A variety of scenarios were defined to examine two factors of interest: SGNSC inventory levels and SGNSC reliability. AMC has projected 18 SGNSC units for Travis, the base of study. The transient aircraft mission at Travis required SGNSC; however, this mission was neither a focus of this study nor a significant user of local SGNSC. Two SGNSC were detailed to support the transient mission to account for this real concern. Three SGNSC inventory levels were examined: 5, 10, and 15. For each inventory level, a SGNSC MTBF of 50, 100, and 500 hours was modeled.
Travis AFB operations were modeled for a 5-year period. As aircraft complete missions, failures occur. Failures requiring SGNSC were modeled using peacetime and surge flying schedules.
Data collected from this 5-year simulation represented steadystate data. As with most steady-state simulations, the initial period of the simulation, called the transient or warmup period, was not indicative of steady-state conditions. Including transient data in steady-state calculations introduced bias. The transient period, conservatively determined to be the first 6 months of the simulated timeframe, was removed." Final statistics were based on 30 replications, each with the initial transient removed. Scenarios were compared based on 95-percent confidence intervals. As noted in the results below, various confirmatory simulations were conducted as dictated by the initial analysis of the simulation data. The primary data examined were SGNSC utilization, mission effectiveness, and time spent waiting for SGNSC assets to become available.
Initial results were impressive. At an inventory of five SGNSC with a 50-hour MTBF, aircraft sorties did not suffer at all. A subsequent confirmatory run reducing the inventory to three still did not affect the flying schedule. SGNSC utilization was only 29 percent, which included travel time. LCOM limitations necessitated including travel time in the utilization rate. However, wait time increased dramatically, from an acceptable average 4.4 hours per month with five SGNSCs, to a likely unacceptable 69.2 hours per month with three SGNSCs. This confirmed nitrogen utilization was not very high.
People are the most valuable resource on the flight line, and if they are waiting for equipment, they cannot work. Greater coordination between the AGE shops and maintenance holds promise in leveling out demand by forecasting nitrogen requirements. The ability to plan AGE consumption is merely held out as an opportunity for future improvement, especially regarding deployments. The current demands for attention on maintenance forced this study to focus on the most efficient and effective utilization of AGE within existing command structures and maintenance concepts.
Therefore, the focus changed from one of aircraft ability to meet the schedule to one of reducing wait time to an acceptable level of pain. General goals in the service sector are an 80 percent utilization rate for resources. Some sectors cannot and probably should not try to attain this kind of utilization. A more appropriate comparison would be with emergency services. An emergency ambulance has a utilization of about 30 percent. (12) However, if someone must wait for an ambulance, the family may not be comforted knowing an ambulance fleet was reduced to increase overall utilization. The flight line presents a somewhat similar scenario; we do not want to wait for support equipment when trying to restore aircraft to a mission-capable status. The consequences of waiting for AGE on the flight line outweigh the advantages gained by higher utilization of AGE.
The failure rates of SGNSC were manipulated to determine the sensitivity of demand. MTBF times of 50, 100, and 500 hours were used. The differences were very small as illustrated in Figure 1.
SGNSC was not very sensitive to changes in reliability as Figure 1 shows. It was much more sensitive to the quantity of SGNSC. An additional run with an inventory of four was included in Figure 1. Wait times did not begin until an inventory dropped and a quantity of five SGNSCs was reached. Wait time increased very quickly after that, as Table 5 shows.
A comparison of confidence intervals by SGNSC MTBF in Table 6 shows that an inventory of five SGNSCs or more resulted in no statistical difference in wait time with 95 percent confidence. Even when there was a statistical difference, the practical differences were minor until SGNSC was constrained to three units.
While the peacetime results were illuminating, they did not address the ability to meet maximum demand. The military, by nature, requires excess capacity. The ability to respond quickly and with force during wartime is necessary. An unfortunate side effect of this capability is the apparent lack of utilization of capacity during a peacetime posture. Using an LCOM surge template, the model was shifted into a fly-when-ready mode. SGNSC quantities of 5, 10, and 15 were used again to examine sensitivities. Additional confirmatory runs with quantities of 11 and 12 SGNSC were added to further clarify wait times and utilization. MTBF times were initially 500 hours, but additional runs with 50-hour MTBF times were conducted to verify SGNSC availability under maximum-usage scenarios at quantities of 11 and 12. The results of the comparison of 50- and 500-hour MTBF times under a surge scenario were very similar to the peacetime results. While Table 7 shows statistical differences at a 95-percent confidence, the practic al differences are again minor at these inventory levels.
The effect of varying reliability of the SGNSC carts was minor compared to varying the quantity of SGNSC. The wait time knee in the curve occurred when SGNSC inventory fell to 12 carts. Reduced further to 11 and then 10 units, wait times increased dramatically. An inventory of five SGNSCs gave an impressive 94-percent utilization. However, just as we do not want to wait for an ambulance, we cannot accept the waiting time associated with this tremendous utilization. Utilization and wait times for the various quantities of SGNSC are listed in Table 8.
The effect of changing to a fly-when-ready mode of operation exposed SGNSC to a much higher demand rate. What was apparently a vastly underutilized fleet of ten units with a dismal peacetime utilization of 9 percent exploded during surge to 51 percent, with an unacceptably low, overall average wait time of 22 hours per month.
SGNSC is currently being fielded. Unit reliability is uncertain, but historical AGE data and modular aircraft support system research yielded reasonable bounds for MTBF data. This study failed to judge MTBF as a prime driver for SGNSC BOI.
Utilization and wait time were inversely related. High utilization should not become a factor for SGNSC BOI as it comes with a high a cost to the maintainer.
The BOI driver seemed to be the unit surge mission. While still yielding excess peacetime capacity, the resulting inventory levels were a fairly nice reduction in planned inventory levels (28 percent in this case).
AGE utilization was very low, and demands for AGE resources overstated. The current overabundance of AGE on the flight line is unaffordable in today's Air Force. The methodology yielded a useful, objective basis in determining AGE levels for new and existing programs and should be used in conjunction with current methods for more insight into AGE inventory levels.
The model promotes a reduction of AGE to at least an inventory of 12 plus 1 for transient aircraft. MTBF effects are minimal, and it is postulated that a spare for the transient support is unnecessary, provided transient support may borrow an SGNSC from the home-station AGE shop. This would mean an inventory of 13 SGNSC vice the current 18 programmed for Travis by AMC. The current contract for SGNSC, at $20M for 570 carts, is about $35K per cart. A reduction of five SGNSC would mean a reduction of about $175K in acquisition costs. If the model could be extended, the possibility of a 28-percent reduction in SGNSC acquisition costs would amount to about $5.6M over the life of the contract. These reductions in AGE levels Air Force-wide would also have the benefit of cost avoidance in operations and maintenance costs.
While the results are positive, this study only attempted to estimate actual requirements. The results did not incorporate war reserve materiel, deployment, or other potential demands or outside limiting factors, only demands anticipated at Travis AFB. It must be remembered these were estimates only and should be taken into consideration with other factors and experience before applying any results to the field. However, the results gave a reasonable estimation of the potential cost savings in reduced procurement costs.
One of the issues in optimizing a certain part (SGNSC) of an interrelated system is the effect on other parts of the system or the flight line. Reducing SGNSC may increase utilization, but AGE drivers may still be insufficient. Waiving reliability requirements may not have a serious effect on wait time, but AGE shop manpower may need to be increased. This study only examined the effects of reducing AGE levels to meet expected mission requirements. When a resource pool is reduced, other issues may arise.
An important issue discovered in the analysis of AGE-inventory sizing was the wait time for AGE. A queuing simulation was ideally suited to the fluid environment of the flight line, and WUCs were the most accurate indicator available to derive AGE consumption. Adjusting AGE inventory to minimize wait time or keep it down to an acceptable level was the prime measure of AGE mission effectiveness.
This study is not a mathematical formula to quantify the number of SGNSC carts needed on the flight line. The research was a more objectively oriented approach to identify those aspects of actual AGE needs on flight-line operations that have the greatest impact and the relative consequences of adjusting AGE inventory levels.
[Figure 1 omitted]
[Figure 2 omitted]
Table 1 Desert Storm (1/16-2/28 1990) vs Modeled Statistics (4) Actual Model Sorties Flown 2,185 2,209.0 (within 1.1%) Flying Hours 7,360 7,379.6 (within 0.2%) APG-70 LRU Pulls 224 226.6 (within 1.1%) APG-70 CND 115+ 118.7 (within 3%) Table 2 Luke AFB vs Modeleed Statistics (5) Actual Model Sorties lown 1,040-1,120 1,111.2 Flying Hours 1,640 1,633.2 APG-70 LRU Pulls 105 105.1 Table 3 KC-10 Task/[N.sub.2] Service Times and Number of Landings per Action KC-10 Task Service Landings/ WUC System Time-Hrs Time Action 13DAB MLG 2.75 0.35 8.82 13DBB NLG 2.75 0.35 16.17 03200 BPO 4.67 0.38 2.00 03100 PRE 0.77 2.00 45ABH Accumulator 0.87 200.00 13AAO MLG Strut 1.12 12.13 13BAO NLG sturt 1.12 12.13 13AEO Centerline landing gear 1.12 19.40 46GJO Boom pneumatic disconnect 1.25 7.46 Table 4 C-5 Task/[N.sub.2] Service Times and Number of Landings per Action Task Service C-5 Time- Time- Landings/ WUC System Hrs Hrs Action 3100 Preflight 0.77 2.00 3200 Throughflight 0.50 2.00 3210 BPO 2.00 2.00 13AAA Shock Strut Assembly 2.8 0.75 16.44 13FCN Ldg Gr Strg Actuator 2.70 0.75 411.00 13LA (*) MLG Tire 2.00 0.35 .83 13LC (*) NLG Tire 2.00 0.35 6.42 24ALP APU Accumulator 3.95 0.88 206.00 91AAF Slide bottles 1.35 206.00 11LCH Crew Entry door accumulator 2.80 0.88 206.00 11LCK Crew Entry-door accumulator 2.80 0.88 250.00 Table 5 Effect of SGNSC Quantity on Wait Time and Utilization (Peacetime) Average total wait SGNSC (hrs/month) (hrs/month) Unit ((hrs/month) ((hours)time/ Quantity month (hrs) Utilization 3 59.6 28.9% 4 14.4 21.6% 5 4.4 17.9% 8 0 11.2% 10 0 9.0% 15 0 6.0% Table 6 Difference in Wait Time at 50 and 500 Hour MTBF (Peacetime) 95% CI 3/50 3/500 4/50 4/500 5/50 5/500 8/50 8/500 Lower 68.25 58.83 15.15 14.13 4.40 4.32 0.07 0.07 Upper 69.89 60.23 15.84 14.59 4.66 4.61 0.10 0.09 Table 7 Difference in wait time at 50-and 500-Hour MTBF (Surge) 95% CI 11/50 11/500 12/50 12/500 Lower 8.80 7.88 2.96 2.73 Upper 9.25 8.20 3.18 2.90 Table 8 Effect of SGNSC Quantity on Wait Time and Utilization (Surge) Average Wait Time SGNSC Quantity Per Month (Hrs) Utilization 5 2,860.0 94% 10 22.0 51% 11 8.0 46% 12 2.8 42% 15 0 34%
(1.) Defense LINK News, 1998.
(2.) L-COM Final Report, F4-E, Vol 1, Tactical Air Command, Langley AFB, Virginia, 15 Aug 73.
(3.) L-COM Final Report, 1-6.
(4.) JSF JIRD III Accreditation Report (Draft), Joint Accreditation Support Activity, Naval Air Warfare Center, Weapons Division, China Lake, Georgia, Sep 98. 4-17.
(6.) Interview with Alan J. Wallace, Aeronautical Systems Center, Wright-Patterson AFB, Ohio, 18 Dec 00.
(7.) Draft JSF JIRD III Accreditation Report, 4-7, 4-8.
(8.) 1st Lt Jeff Havlicek, "Aerospace Ground Equipment's Impact on Aircraft Availability and Deployment," master's thesis, School of Engineering, Air Force Institute of Technology, Wright-Patterson AFB, Ohio, Sep 97, 83.
(9.) Havlicek, 52.
(10.) Dick Cronk and Alan Wallace, LCOM User Manual, Version 94B, 58.
(11.) A. M. Law and W. D. Kelton, Simulation Modeling and Analysis, [3.sup.d] ed, Boston: McGraw-Hill 2000, 499-501.
(12.) James A. and Mona J. Fitzimmons, Service Management: Operations, Strategy, and Information Technology, [2.sup.d] ed, Boston: Irwin/McGraw Hill, 1998, 517.
Captain MacKenna is Chief, Maintenance Reengineering Section, Maintenance and Munitions Division, Air Force Logistics Management Agency, Maxwell AFB, Gunter Annex, Alabama. At the time of the writing of this article, he was a student at the Air Force Institute of Technology, Wright-Patterson AFB, Ohio. Dr Cunningham is the Professor of Logistics Management, Air Force Institute of Technology, and a frequent contributor to the Air Force Journal of Logistics. Lt Col Raymond R. Hill, Jr, is Associate Professor of Operations Research, Department of Operational Sciences, Management, Air Force Institute of Technology.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Aerospace ground equipment; Air Mobility Command|
|Author:||MacKenna, James; Cunningham, William A., III; Hill, Raymond R., Jr.|
|Publication:||Air Force Journal of Logistics|
|Article Type:||Statistical Data Included|
|Date:||Dec 22, 2001|
|Previous Article:||A critical look at PPBS: the unfunded priority list. (Logistics Today).|
|Next Article:||Defense Working Capital Fund: In the simplest of terms, the DWCF was established to allow the federal government purchase and repair activities to...|