MSU RETURN'I NG MATERIALS: P1ace in book drop to LJBRAfiJES remove this checkout from 4...:3-nnn. your record. FINES will be charged if book is returned after the date stamped below. -. 4:7 m * I} g‘ r v: f: .G 1: a E :1; I F i I" . 9‘ s. 9 , i A q a q g T. A“)? kl ii 5 i " . {so “' a w? 4%“ ‘ 4... CORROSION IN LARGE AIRCRAFT: MANAGEMENT OF THE PROBLEM By NINA SAMSAMI A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Metallurgy, Mechanics, and Materials Science 1983 ABSTRACT AIRCRAFT CORROSION: MANAGEMENT OF THE PROBLEM by Nina Samsami Field-level corrosion maintenance on 8-52 airplanes was compared from one airbase to another and it was found that variations can be related to environmental differences in weather and atmospheric pollutants. The extent and nature of such maintenance is similar to that of another large airplane system previously analyzed, the C-l41A, despite significant differences in age, mission, and utilization. It can be expected that all large mfllitary airplanes will be similar, with respect to the effect of environment on . maintenance. Considering the environmental variations, sound statistical-. principles then were applied to the problem of sampling a large airplane fleet for special in-depth discovery inspections.' It is shown that a smaller sample than used currently is sufficient, and large savings are possible. To my Mother and Father whose love and sacrifices have been a source of strength throughout my life. ii :1 .h ACKNOWLEDGMENTS The author wishes to express her deepest appreciation and gratitude to Dr. Robert Summitt, her major advisor, for his constant encouragement and friendship, and especially his valuable assistance in the preparation of this thesis. A special thanks to Dr. D. J. Montgomery for his guidance and encouragement in our meetings. The author's gratitude also goes to her husband, Ali-Akbar, without whose encouragement and tolerance this volume would not have been completed. Finally the author wishes to thank the United State Air Force under contract Number F33615-79-C-5122 for providing the fund for this research. TABLE OF CONTENTS LIST OF TABLES. . ............. . .......... LIST OF FIGURES ......................... I. INTRODUCTION ........................ II. BACKGROUND ......................... A. Description of 3-52 Aircraft .............. B. Related Research .................... 1. Atmospheric Corrosion Studies ............ a. Investigation of Atmospheric Exposure Factors that Determine Time-of-Wetness of Outdoor Structures 0 O O O O O O O O O O O O O O O O O O O b. Predicting a Metallic Corrosion Cost of Operating and Maintaining Building and Utility Systems . . 2 o opera ti Gila] Fattors o o o o o ‘0 o o o o o o o o o o o a. A Statistical Analysis of Maintenance Costs of Large Jet Aircraft . . . . . . . . . . . . . . . b. A Method For Adjusting Maintenance Forecast to Account for Planned Aircraft Sortie Length . . . c. Corrosion Prediction in F-4 Aircraft Assigned to the Pacific Air Force. . . . . . . . . . . . . . C. Aircraft alloys ..... . ............... 1. Corrosion Modes. .................. 2. Magnesium Alloys .................. 3. Aluminum Alloys ................... D. PACER LIME ....................... 1. Test Method and Materials. ............. 2. Corrosion Rates Compared with Environment. ..... 3. The Corrosion Severity Classification System . . . . a. Environmental Variability ............ iv vii ix 10 10- 11 12 L 12 15 16 18 18 23 28 34 35 35 36 36 b. Environmental Severity Algorithm for Aircraft Corrosion . . . . . . . . . . . . ........ 39 E. Time-of-Netness and Environmental Parameters ...... 42 III. USAF MAINTENANCE DATA COLLECTION SYSTEM .......... 47 A. Inspections . . . . . ................. 47 B. MDCS Data and USAF Processing ............. 50 C. Permanent Data. .................... 56 IV. 8-52 CORROSION MAINTENANCE DATA ............... 60 A. Data Base and Overview ................. 60 B. Comparison of 3-52 and C-141A Maintenance Data ..... 65 C. Organizational-Level Maintenance . . . . . . . . . . . . 68 1. Maintenance Effort Compared with Possession. . . . . 68 2. Organizational-Level Maintenance by How Malfunction Code . . . . . ........... . . 69 ' 3. Cumulative Maintenance . . . . . . . . . . . . . . . 74 4. Statistical Distribution of Organizational-Level Maintenance. . . . . . . . . . . . . . . . . . . . . 78 5. 1Comparision with Environment ..... . ...... 86 V. MAJOR AIRCRAFT INSPECTION PROGRAMS .............. 87 A. Reliability Centered Maintenance. . . . . . . . . . . . . 88 1. Application of the MSG-2 Technique in a Military Aircraft ....................... 90 B. Analytical Condition Inspection ............. 91 C. Corrosion Statistics ................... 103 D. Analytical Condition Inspection Sample Size ....... 111 VI. ESTIMATING MAINTENANCE INTERVALS .............. 117 A. The Model: Cost-Unit Time Per Operation ........ 118 B. Corrosion Reliability from Neibull Statistics: Organizational-Level Maintenance ............ 119 V VII. SUMMARY. . . BIBLIOGRAPHY. vi IO. 11. 12. 13. 14. 15. 16. 17. 18. LIST OF TABLES Age and Flying Hour Comparison C-141A and BSZX ......... Distribution of 8-52 Airplanes in Various Airbases . ...... Authorized Inspection and Maintenance Programs, B-SZD ..... MDCS Record Format B-SZX . . . . . . . . . . . . . . . . . . . . B-SZD Organizational- and Depot-Level Corrosion Maintenance by How Malfunction Code, 10 1978 - 3Q 1979 ...... . . . . . B-SZG Organizational- and Depot-Level Corrosion Maintenance by How Malfunction Code, 10 1979 - 3Q 1979. ............ B-SZH Organizational- and Depot-Level Corrosion Maintenance by How Malfunction Code, 10 1979 - 30 1979 ..... . . . . . . . Comparison of 8-520 Organizational- and Depot-Level Corrosion Maintenance Costs by How Malfunction Code, 10 1979 - 3Q 1979 . . Comparison of 8-526 Organizational- and Depot-Level Corrosion Maintenance Costs by How Malfunction Code, 10 1978 - 3Q 1979 . . Comparison of B-SZH Organizational-and Depot-Level Corrosion Maintenance Costs by How Malfunction Code, 10 1978 - 30 1979 . . Distribution of Maintenance Man-hours Among Corrosion-Related How Malfunction Codes for C-141A and 8-520, 6, H Airplanes . . . Distribution of Failures Among Corrosion-Related How Malfunction Codes for C-141A and B-SZD, G, H Airplanes . . . . . Corrosion-Related Maintenance Costs Compared with Ownership B-SZD Airplanes, FY 1978-79. . . . . . . . . . . . . . . . . . . Corrosion-Related Maintenance Costs Compared with Ownership B-SZG Airplanes, FY 1978-79. ........ . ......... Corrosion-Related Maintenance Costs Compared with Ownership B-52H Airplanes, FY 1978-79. . . . . . . . . . . ........ Corrosion-Related Labor Costs, B-SZD Airplanes, 1978-79, By Airbase and Work Unit Code ..... . . . . . . ........ Corrosion-Related Labor Costs, B-SZG Airplanes, 1978-79, By Airbase and Work Unit Code . . . . . . ............. Corrosion-Related Labor Costs, B-52H Airplanes, 1978-79, By Airbase and Work Unit Code . . . ......... . . . . . . . vii 5 8 48 58 62 62 63 63 64 64 66 7O 7O 19. 20. 21. 22. 23. 24. 25. 26. 27. 8-520 Organizational-Level Corrosion Maintenance, Mean Monthly Man-hours per Airplane, 10 1978 - 3Q 1979 . . . . . . . . . . . B-SZG Organizational-Level Corrosion Maintenance, Mean Monthly Man-hours per Airplane, 10 1978 - 3Q 1979 . . . . . . . . . . . B-SZH Organizational-Level Corrosion Maintenance, Mean Monthly Man-hours per Airplane, IQ 1978 - 30 1979 ......... . . B-SZG ACI Airplanes FY 1978 .................. B-SZH ACI Airplanes FY 1978 .................. 8-526 ACI Airplanes FY 1979 .................. B-SZH ACI Airplanes FY 1979 ........ . ......... Cumulative P1 Values for Fleet Leader Aircraft for First Case . Cumulative P2 Values for Fleet Leader Aircraft for Second Case. viii 82 83 83 93 94 95 96 114 115 IO. 11. 12. 13. 14. 15. 16. ' 17. 18. 19. LIST OF FIGURES 8-520, 6, and H Airbases . . . . . .............. , . 6 Corrosion Rates of AZ31B Magnesium Alloy Compared with Environmental Ratings ............. . ....... 38 Aircraft Corrosion Damage Algorithm ...... . . ..... . 41 Panel Temperature vs Time T.=250, T.=27S, and Tm=300 ...... 46 AFTO 349 Obverse ........................ 62 AFTO 349 Reverse . . . . . . . . . ....... . ....... 63 Cumulative Corrosion Maintenance on B-52D Airplane Serial Number 55000087, 1975-1979 ................... 76 Cumulative Organizational-Level Corrosion Maintenance on 8-520 Airplane Serial Number 57005616, 1975-1979 . . . . . . . . . . . 87 Cumulative Corrosion Maintenance on B-SZH Airplane Serial Number 60000004, 1975-1979 . . . . . . . . . . . . . . . . . . . . . . 78 B-SZD Organizational-Level Corrosion Maintenance to Individual Airplanes, 1078-3079 . . . . . . . .......... . . . . . 80 B-SZH Organizational-Level Corrosion Maintenance to Individual Airplanes, 1078-3079 . ......... . . . . . ...... . 81 8-520 Organizational-Level Corrosion Maintenance Compared with Environmental Ratings, 1078-3079 . . . . . . . . . . . . . . . . 84 8-520 Organizational-Level Corrosion Maintenance Compared with Environmental Ratings, 1078-3079 . . . . . . . . . . . . . . . . 85 B-52H Organizational-Level Corrosion Maintenance Compared with Environmental Ratings, 1078-3079 . . . . . . . . . . . . . . . . 86 Organizational-Level B-52H Corrosion Maintenance Before and After PDM in FY 1978 ...................... 97 Organizational-Level B-52H Corrosion Maintenance Before and After PDM FY 1979 ....................... 98 Organizational-Level 8-528 Corrosion Maintenance Before and After PDM FY 1978 ....................... 99 Organizational-Level 8-526 Corrosion Maintenance Before and After PDM FY 1979 ....................... 100 Organizational-Level 8-520 Corrosion Maintenance Before and After PDM for ACI Airplanes FY 1979 .............. 101 ix 20. 21. 22. 23. 24. Organizational-Level 8-526 Corrosion Maintenance Before and After PDM for ACI Airplanes FY 1978 .............. Corrosion Damage Detection v§_Time . . Sequential Corrosion Events and the Probability of Damage Detection ........................... Damage Tblerance Rating v§_Cumulative Probability of Damage Detection ........................... 102 107 108 109 116 I. INTRODUCTION As aircraft operational life is extended, corrosion becomes a major factor' in ownership cost. It is estimated that USAF spends more than one billion dollars each year for detection and prevention of aircraft corrosion problems (1). Recent analysis of corrosion indicates that airplane structures exhibit corrosion rates which are related to time and environment rather than flying hours. Unfortunately there is no established relation between corrosion damage and service environment. The maintenance intervals currently in use in USAF aircraft are based upon wearout predictions where corrosion is not a factor. The existing maintenance schedules are either isochronal (fixed flying hours) or periodic (fixed calendar interval), when part or all of an airplane receives "Inspection, Repair as Necessary" (IRAN). Currently, USAF aircraft are under a complex mixture of isochronal- and periodic-scheduled corrosion maintenance, which, in a few systems, is based on “Reliability-Centered Maintenance. (RCM). Such programs are called Reliability Centered Maintenance because they are centered on achieving the inherent safety and reliability capabilities of equipment at a minimum cost. The RCM could be ideal if the complexities of corrosion processes were better understood to include it as a part of RCM logic. The corrosion tracking and prediction of airplanes first was initiated in 1975 under a contract with Michigan State University (2). In that study, the corrosion maintenance of C-141A aircraft was analyzed. The C-141A aircraft is a long-range, high-speed, high-altitude monoplane designed for use as a heavy logistic transport. It was manufactured in the middle of the 1960's. Most of the C-141A force are used for airlift missions. The present study extends corrosion tracking and prediction to the B-52 fleet, another large, but older aircraft system. The 8-52 airplanes are long-range bombers and have been 2 in service for more than 25 years and they are expected to remain in service for at least ten years longer. In-service operational data, repair maintenance histories, material information, weather, airbase factors, and other relevant information are the basis of this study. The primary source of data was the Maintenance Data Collection System (MDCS) (55). The 8-52 study is divided into four phases. In phase I aircraft series and airbases were selected for more detailed study. Although several 8-52 airplane series were manufactured, only the D, G, and H series remained in service. (As the study ended, the remaining D-series were scheduled for retirement.) The 79 still-flying D's had been rebuilt recently (3) hence their corrosion problems might be atypical; but South Pacific area deployment quickly had cancelled this advantage. Although initial attention was focused on the 6- and/or H-series, AFLC, SAC, and Boeing personnel urged inclusion of the D-series because of these unique characteristics--newly rebuilt and Pacific service--as well as subjective opinions that it was in worse condition corrosion-wise, hence a source of useful information, particularly with respect to corrosion protection systems which had been added to the aircraft at the time they were rebuilt. Consequently, closely-tracked airplanes and gairbases represented all three series. Phase II is the comparison of the 8-52 corrosion maintenance data with that of the C-141A fleet. The 8-52 airplane differs from the C-141A in terms of utilization, age, and airbase assignment. Although C-141A aircraft spent most of their life in the U.S., the 8-52 experienced some stay in southeast Asia. If our study can find similar corrosion damage for both of these aircraft fleets, it will be possible to conclude that all large aircraft have similar corrosion damage. If so, then we can establish an overall pattern of corrosion related-maintenance. It is likely that smaller airplanes could be treated as well. Phase III is the assessment of variations in maintenance repair rates from one airbase to another. Also addressed were questions of materials variation with a view to impacting Programmed Depot Maintenance package decisions on the basis of material type compared with similar materials installed in inaccessible locations. Finally, statistical methods were applied to the selection of representative airplanes for intensive inspection (Analytical Condition Inspection). Phase IV is the prediction of proper scheduled maintenance intervals for 8-52 airplanes based on corrosion problems rather than flying hours. Although it is more likely that corrosion will be a sufficient concern to. modify corrosion rates, there is some opportunity, however, to modify the airplane selection process. . This thesis addressed especially the statistical questions relating to the selection of Analytical Condition Inspection airplanes and the determination of maintenance intervals based on corrosion factors. In addition, some questions, related to the physics of moisture deposition related to atmospheric conditions, were addressed. For clarification purposes, some related material from the Final Report of the Project is included, together with earlier work on the PACER LIME project, to which the author contributed. II. BACKGROUND A. DESCRIPTION OF 8-52 AIRCRAFT The 8-52 bomber has been a primary weapons delivery system of the USAF Strategic Air Command for more than 25 years. There were eight series of 8-52's: A through H. Only the D, G, H series are still in service, but the D-units are scheduled for retirement within two years. The chronologically-older 0 series, however, actually were the newest 8-52 airplanes, having been rebuilt and modified extensively under Engineering Change Proposal 1581 (ECP 1581, code-named PACER PLANK), completed in 1977 (3). Data related to inventories, flying hour utilization, and chronological age of these three 8-52 series are shown in Table 1 together with similar data for the C-141A (2). In the time period of this study the 8-52's were located at AnderSen (Guam), Carswell (TX), Dyess (TX), March (CA), airbases; the B-SZG's were located at Barksdale (LA), Blytheville (AK), Castle (CA), Fairchild (HA), Griffiss (NY), Loring (ME), Mather (CA), Robins (GA), Seymour-Johnson (NC) and Hurtsmith (MI); and the B-SZH‘s were located at Castle (CA), Ellsworth (SD), Grand Forks (ND), K. I. Sawyer (MI), and Minot (ND) Figure. 1. The first B-520 was tested in September 1956. There were originally 157 of these aircraft, 69 of which were built by Boeing-Wichita and the rest at Seattle. There were 186 8-52G's which were all built by Boeing-Wichita. The first 8-52G was tested in October 1958. The H variant of the B-52's were produced by Boeing-Wichita. There exist 104 of these aircraft. Major maintenance and most modifications of B-SZG's have been performed by the Oklahoma Air Logistic Center at Tinker AFB and of B-SZD and H's by San Antonio Air Logistic Center at Kelly AFB. Generally, one wing (nominally 16 airplanes, but from 14 to 19 in practice) is assigned to l mo 00—' )- 95 No. of - 84 )- Aircraft _ F r 50—. 45 46 _ F " _}_4 3 "' l6 l6 ll l4TlelO9 B .9201 211.21 C-l4lA B-SZ’G B-SZM 1975 data 1983 data 1983 data AGE IN YEARS High 24,000 15,514 13,493 12,332 Flying Hours Average 15,343 l3,726 8,819 7,402 Low 5,000 l2,512 3,269 2,482 C-l4lA 8-520 B-52G 8-52H T975 l983 l983 l983 data data data data Table 1. Age and Flying Hour Comparison C-141A and B-52X. .mmmanufi< a mum .o .numum .H ousmfim m4m=._.>.—lm mmus z. m seer 44u3m¢< ~ 11 a... m. m 1 r z. a a 4m :hnzmhual mxzcuno uz—zoq aa—xuz—cu :hzozmaau beau: each airbase, but three airbases (Carswell, Barksdale, and Ellsworth) had two wings each. In general, 8-52 airplanes are traded infrequently from one airbase to another. There were major transfers of airplanes in the time period 1975 to 1980, but these were related to the closing of certain airbases and re-alignments of commands. Occasionally, transfers were effected when an airplane entered PDM, but moved to another airbase on output, from modification fly-in programs, and to keep units at required strengths. Table 2 shows the distribution of B-52 airplanes among airbases for FY 78-79. There is some geographic homogeneity among the D-Model airbases, only Andersen is not in the U.S., the rest are in the Southwest U.S. Carswell and Dyess both are in Texas, about 150 to 200 miles apart, and March AFB, near Los Angeles, is in a near-desert environment similar to that of Texas. Similarly, of the H-series airbases, three are located in the Northern Plains states, within perhaps 300 miles of one another. K. I. Sawyer AFB in the Upper Peninsula of Michigan, shares a similar climate, but it is not in the northern Great Plains, rather in a forested area near Lake Superior. Castle AFB is in northern California and perhaps somewhat different from the others. Nevertheless, three airbases have more-or-less similar environments. The G-series airplanes are based in a wide range of environments. But among the ten airbases, there are apparently similar subsets. Four (Barksdale, Blytheville, Robins and Seymour-Johnson) are in the southeast United States, two in the Mississipi valley and two in the Appalachian Piedmont. Four (Fairchild, Griffiss, Loring and Nurtsmith), although widely separated, share a northern environment and similar weather. Castle and Mather, both in central California near one another, might have similar environments. TABLE 2 - DISTRIBUTION OF B-52 AIRPLANES IN DIFFERENT AIRBASES. D-AIRPLANES FY 1978 FY 1979 Andersen 15 14 Carswell 37 34 Dyess 19 15 March 24 14 G-AIRPLANES FY 1978 FY 1979 Barksdale 3O 29 Blytheville 14 15 Castle 14 15 Fairchild , 16 14 Griffiss ' 15 - 15 Loring 14 14 Mather 14 14 Robins 14 14 Seymour-Johnson 14 14 Hurtsmith 15 14 H-AIRPLANES FY 1978 FY 1979 Castle 8 8 Ellsworth 30 28 Grand Forks 16 16 K. I. Sawyer 19 19 Minot 17 16 8. RELATED RESEARCH In previous years there have been some studies on maintenance and relation of corrosion to the maintenance cost of airplanes. These studies considered different types of causative factors for maintenance cost or corrosion behavior of airplanes. Since each of these studies could be a guideline fer our research, we will first classify causative factors and then review each study. Aerospace vehicle maintenance is necessitated by many causes; however, for the purpose of our study, these may be classified as: 1. Environmental factors. 2. Operational factors. 3. Housekeeping factors. 4. Miscellaneous. The environmental factors are considered to be active primarily while the aircraft is on the ground. Since military aircraft in general are flown less than commercial aircraft, the role of environmental factors would be greater in causing failures in military systems than they would be in commercial aircraft. Environmental factors consist of weather and atmospheric pollution. (a). Heather parameters include temperature, precipitation, solar radiation, wind direction, wind speed, relative humidity, dew point, cloud cover, and fog. (b). Pollutants include particulates, both solid and liquid in particle sizes from 0.1 to 100 um. Dust, grit, fly ash, and visible smoke particulates larger than 20 um settle to the ground somewhat quickly. Smaller particles remain suspended much longer and may be dispersed over extremely wide areas. Thus large particulates might cause corrosion problems close to the source 10 (sea salt spray is a special case), whereas small particulates can be important at great distances from their source (8). The operational factors are those conditions to which the system is exposed under field conditions. The parameters used to characterize these factors include the number of sorties per month, the airborne vibrations, the operating hours per month, the average sortie length, and other factors. Housekeeping factors refer to the quality and effectiveness of housekeeping and preventive maintenance. Although this should be relatively easy to control, it is in fact a serious problem in USAF. The quality of housekeeping varies from one airbase to another and thus conceivably might be considered an environmental variable. Since it is not easily measured, and it varies from time to time, it can not be considered in'a rating system. Miscellaneous factors have to do with such items as Time Compliance Technical Orders, accidents, and design problems. Though this list of causative factors is not exhaustive, it gives a reasonable indication of the range of causes for aircraft maintenance. 1. ATMOSPHERIC CORROSION STUDIES a. INVESTIGATION OF ATMOSPHERIC EXPOSURE FACTORS THAT DETERMINE TIME-OF- NETNESS OF OUTDOOR STRUCTURES Grossman (5) studied the atmospheric factors that determine Time-of-Netness. Hetness of outdoor structures is caused partly by condensation of water from the atmosphere as dew. Rain also is a source of wetness. The heat exchange process by radiation from the exposed surface to the sky is examined . Temperature differences between an insulated black surface, facing skyward, and ambient air conditions were observed to be as 8 °C. Effects of wind velocity, orientation, and surface characteristics were given. Time-of-wetness measurement for test panels exposed at Miami, 11 Florida were reported, including the “black box" exposure method used for coated panels. A regression analysis of corrosion rate 33 different variables was made. Grossman emphasised that Time-of-Hetness, not time of exposure, is a variable. Other variables were temperature during the wetness period and sulfation activity of the air during the wetness period. And at the end he drew some conclusions about exposure testing for the degradation forces of wetness. 1. Time-of-Hetness is an important parameter for defining the interaction of a given material with a specific environment. 2. If Time-of-Hetness is measured, the accelerating action of increased temperature can be established and correlated. 3. If sulfation activity is measured, the accelerating action of higher 502 content in the air can be established and correlated. 4. Insulated test panel exposure methods can be used to accelerate the effect of wetness at any given test site. 5. The positioning of test panel with respect to the cold sky is an important test parameter. b. PREDICTING THE METALLIC CORROSION COSTS OF OPERATING AND MAINTAINING BUILDING AND UTILITY SYSTEM. The problems of corrosion have long been recognized by military services but serious analyses of facility maintenance corrosion costs have never been undertaken. These facilities are mostly buildings and utilities which have direct relation to the public sector. Hahin (6) tried to determine 1) how much the United States Army spends for corrosion related maintenance, 2) are those costs predictable from some data or installation description, 3) how accurate are the cost predictions when compared with real spending? 12 The facility maintenance organizations of the Army and the Air Force were investigated to determine what percentage of work was corrosion related. To develop a total cost model, it was necessary to examine the correlation between component parts with some independent variables. The independent variables were selected as: 1) Mean annual temperature, 2) mean relative humidity, 3) annual rainfall, 4) mean dew point, 5) soil resistivity, 6) air pollution data, and 7) water quality data. Hahin developed two cost prediction models. The first model, termed a “work center model“, predicts the cost from the knowledge of labor, material, and equipment costs. The second method of cost predicting required only knowledge of system unit dimensions, climate, geographic, and environmental variables, and was termed a “system dimension model“. Hahin found that a "work center” approach is more accurate than a “system dimension model“, because in the “system approach”, several trades may be maintaining one system. The overall corrosion cost is simply obtained by adding all the individual corrosion related costs: Tetal corrosion cost I (Operating + Maintenance Cost) + (Energy Losses) + (Out-of-House contracts). At last, Hahin classified the corrosion forms as: 1) Atmospheric, 2) high temperature oxidation, boilers, and system condensate, 3) refrigeration and other cooling media, 4) potable water corrosion, and 5) underground corrosion, and then developed the cost distribution for each class of corrosion. Hahin determined that the total cost of corrosion is between 7.72 to 252 of total expenses, depending on local conditions. 2. OPERATIONAL FACTORS a. A STATISTICAL ANALYSIS OF MAINTENANCE COST OF LARGE AIRCRAFT In 1968, Gilster (7) published "A Statistical Analysis of Maintanance Cost on Large Jet Aircraft“. The Boeing B-52A-H aircraft were used as a sample for his analysis. The sample ran from 1965 to 1966. In order to generalize 13 his results to a wide range of aircraft, Gilster selected only those systems which are common to all jet aircraft. These analyses included the airframe, landing gear, flight controls, power plant, pneudraulics, and fuel systems. The data of Gilster's study was provided through the AFM 66-1 data collection program. The statistical method of costing was the approach through this study. The cost analysis model used in this study were defined as a Policy Response model. Multiple regression analysis was the basic technique used for this type of costing. Gilster claimed that multiple regression analysis had several advantages. First, the regession line normally provides a better approximation of the true cost function than the conventional method of average-cost accounting, especially if there are fixed costs and curvatures in the true cost function. Second, the regression technique provides a more realistic allocation of cost between influential factors. The age of the aircraft, the mission length, the time of low altitude flight, and the technical difference between aircraft systems were the factors that were found to be more dominant on manhour costs. Age has a high significant effect on aircraft failure rates. Calendar age provided more powerful explanatory variables for this effect than did the accumulated flying hour time on the aircraft. Gilster derived several conclusions from an evaluation of individual flight data collection at Nestover Airbase, MA. The conclusions were: (a) Failure rates were non-constant through a mission's length. The sensitivity of failure to the length of the mission varied among the systems. (b) Approximately 20 percent of the total maintenance actions taken on the aircraft resulted from inadequate diagnosis and repair of previous equipment failure. (c) An evaluation of individual flight data indicated that a recursive relationship existed between the criterion variables of this study. 14 Malfunctions affected manhours, but were not in turn significantly affected by the number of manhours expended on system repair. (d) The individual flight data of the Nestover sample was not characterized by adverse statistical properties. Although failures appeared to be generated by a Poisson process, a recommended deflation of the data decreased the reliability of the estimated model parameters. The changes in failure rates, skill levels, and manhour availability experienced by the Strategic Air Command during the sample period were reflected dramatically by the maintenance criterion variables. The decline in the skill composition of the maintenance force led to an increase in the aircraft failure rates. Gilster employed the opportunity cost concept to assign a dollar value to manhours. The writer felt that the wage offered mechanics by the commercial airlines provided a more realistic value of the cost of a direct hour of labor on the aircraft than the figures derived from the Department of Defence Planning Guides. “The airline wage is the true cost of retaining and utilizing mechanics in the Air Force.“ Finally, Gilster claimed that there were indicators that failure rates and manhour efficiency were affected by the climates experienced by aircraft and personnel. Several systems showed higher failure rates at northern bases during the winter. He defined the climatic condition in terms of two factors, viz., temperature and precipitation. In the first reference the geographical areas of the world were classified into 16 climatic groups. The southeastern 8-52 airbases are characterized by humid subtropic and tropic climates. The northeastern bases fall into the humid continental cool summer climate classification. The northwestern bases were dominated by a mid-latitude semiarid climate and the southeastern bases fall into either the dry summer subtropic or the tropical semiarid classifications. In the second reference, 15 the discomfort index was utilized. The discomfort index was based on the temperature and amount of moisture in the air. The final reference utilized was Air Force regulation 91-8 which established criteria for the type of cooling equipment authorized for facilities at Air Force bases. Four general climatic zones were defined; A, B, C, and 0. Zone A is hot and dry; Zone 8 is warm and humid; and zones C and 0 experience milder summer conditions. In general, the northern bases were included in zones C and D, the southern bases in zones 8, and the southeastern in zone A. After classifying airbases in terms of these criteria the author added-seasonal effects to the geographical structure. This means that temperature and humidity could not sufficiently specify the climatic condition of each airbase and there should be other factors to explain these changes. For example, it was not clear to the author why fewer malfunctions were recorded during winter in the northeast than in the northwest of the United States. In our point of view the author should consider the environmental condition rather than the climatic condition of each airbase. Lots of factors should be considered to explain the degree and type of corrosivity at each airbase. On the effect of climate, Gilster states: “Aircraft components are normally subjected to rigid temperature and stress tests by the manufacturers before they are approved for installation on the aircraft, therefore, we might expect aircraft systems to be relatively insensitive to climatic conditions normally experienced in the United States.“ Only stress tests and temperature are not enough to show the property of each component. Different metals display widely diverse corrosion behavior in a given environment and also from one environment to another. Some alloys are more resistant in marine locations than industrial and the reverse is true for others. There are unknown factors which affect the behavior of metals 16 in different environments. Ignoring the climatic effects results, Gilster's research gives a good guideline for more research in the economics for maintenance in general, but not specifically corrosion. b. A METHOD OF ADJUSTING MAINTENANCE FORECAST TO ACCOUNT FOR PLANNED AIRCRAFT SORTIE LENGTH Howell (8) attempted to develop a technique by which the Logistics Composite Model (LCOM) simulation engineer can readjust the failure rate of aircraft subsystems with the amount of sortie length of a developmental aircraft. LCOM is a general model which can be adapted to a particular aircraft or a weapon system. The aim of this study was to determine whether sortie length is a significant factor in making comparability analysis and, if so, how the simulation design engineer can account for an expected sortie length. Data were collected for military cargo, bomber, and civilian transports because the most likely application would be modeling developmental cargo aircraft. The data of this research were based on maintenance data of the Boeing 747, a civilian airliner, 8-52D aircraft, and some MDCS information for USAF C-141A, and C-130E cargo aircraft. The mean number of maintenance actions per sortie was plotted against average sortie length, and regression analysis smoothed the data to give mean maintenance actions per flight hour. The regression gave a high standard error and the correlation coefficient was very low, which means that other factors are involved in this regression. For example no effort was made to include the difference in mission profile, maintenance concepts (method and policies), weather (environment), aircraft age, aircrew techniques, utilization rate, and many others. Furthermore the airplanes selected are not at all similar in terms of their mission, utilization, age, and even 17 structure. C-141A and C-13OE are designed for use as heavy logistic transports; 8-52's are long range bomber aircraft; and the civilian airplanes were designed for transportation. Therefore, it is easy to see why they show different amounts of sortie length and numbers of sorties or even varying amounts of maintenance actions per sortie. Military Aircraft in general are flown less than commercial airplanes, the military aircraft are often stationed on the ground and some of them do not even fly for some months. Looking at Howell's tables (8), we find that civilian airplanes had large average flight hours, but needed fewer maintenance manhours than 8-52's. As mentioned above, these two type of airplanes are totally different in terms of construction, and utilization. It makes more sense to compare military aircraft with each other than to compare them with civilian airplanes. Even in the case of military airplanes, we may be able to compare some aspects of maintenance actions but not all of them. As will be shown later, the 8-52 airplanes experienced approximately the same amount of corrosion as C-l41A airplanes. Therefore this result gives us a hint that we may be able to compare other military aircraft in terms of corrosion maintenance manhours and develop a model for corrosion maintenance for all the aircraft. c. CORROSION PREDICTION IN F-4 AIRCRAFT ASSIGNED TO THE PACIFIC AIR FORCE Harrington and Teomy (9) studied corrosion predictability in F-4 aircraft assigned to the Pacific Air Force in 1974-76. The population of this sample was defined as all present F-4 aircraft in USAF inventory. The F-4 is a twin engine jet fighter which is deployed world wide. Mission-design- series of the F-4 in the inventory include the F-4C, F-4D, F-4E, and RF-4C. The sample was limited because of command interest and the availability of data. The original sample of 170 aircraft was reduced to 80 F-4C, F-4D, F-4E, and RF-4C aircraft. The data of their research were obtained from PACER 18 LIME interim corrosion factor values (4), base-level corrosion maintenance programs, Programmed Depot Maintenance (POM), and contract corrosion maintenance manhours. The airbases were Cadena AB in Japan, Clark AB in PI, and Osan A8 in Korea. They adapted Summitt's (2) logic in selecting corrosion-related maintenance actions and used the same How Malfunction codes. They used a multiple regression model to compare base-level corrosion maintenance manhours with available data such as airbases, time since last PDM, and manhours in last contract corrosion maintenance. They had to delete the corrosion severity factor bacause of a high correlation with airbase. A high correlation was found between maintenance manhours and each airbase. The exclusion of variables like corrosion severity factors and other corrosion-related data has lowered the explanatory power of their model. Their regression model resulted in a high standard error and low coefficient of multiple determination (R2). An R2 of .50 is described as a minimum level useful for management engineering and they got .61 which is still low. Nevertheless, the conclusion can be 'drawn that multiple regression analysis is a prOper tool to use in constructing prediction models in the presence of enough data at the base level. 19 C. AIRCRAFT ALLOYS 1. CORROSION MODES Historically aircraft in the USAF inventory had a service life generally less than 15 years. The World War II B-29's, for example, were in service for 10 years, and post war B-47‘s for 13 years. The 8-52's, however, built in the 1950's and 1960's remain in use today and are expected to see service until the year 2000. Therefore, corrosion prevention and control, a limiting factor in service life, is a primary concern. Air Force Technical Order 1-1-2 states (10): "Corrosion or deterioration of metal starts the instant the fabrication or manufacturing process is completed and continues until the material is exhausted or salvaged. The speed of deterioration or corrosion will depend on many factors but primarily on the type of chemisty of the material used; environment to which it is exposed; fabrication and/ or assembly method used; heat treatment and degree or method of protective measures including such things as shot peening, etc., taken to retard the corrosion process. The design or project engineer will be concerned with all the design factors, i.e., mission reliability, maintainability, cost and corrosion which can have a detrimental effect on the equipment. Maintenance personnel in accomplishing structural repair will find that about 50 percent or more of actions or work will be related to corrosion or deterioration in some way. Corrosion is a natural phenomenon which destroys metal by chemical or electrochemical action and converts it into a metallic compound such as an oxide, hydroxide, or sulfate. . In the 8-52 airframe structure, a variety of alloys were used, e.g., aluminum, magnesium, titanium and steel. The corrosion susceptibility of titanium and steel are well known and there is no need for further discussion of these alloys. Therefore we concentrate our study on magnesium and aluminum alloys. Because of differences in the corrodibility of these two metals, it is convenient to divide the operational atmospheric environments into different types. The broad atmospheric categories are 20 industrial, marine, urban, rural and tropical. A metal that resists corrosion in one environment may lack resistance in another, therefore, the relative corrosion performance of the aircraft changes with location. Corrosion in the atmosphere can be classified as: wet atmospheric corrosion; moist atmospheric corrosion; dry atmospheric corrosion. It is not always possible to differentiate these three forms of atmospheric corrosion because gradual transition from one form to another is possible. 'Het atmospheric corrosion" is caused by moisture on the material surface from condensation at 1002 humidity or direct wetting. "Moist atmospheric corrosion.I occurs at relative humidity less than 100% and proceeds under a thin film of electrolyte formed on the surface by capillary action or chemical condensation. "Dry atmospheric corrosion" occurs on the metal surface in the absence of moisture (11). Factors which affect atmospheric corrosion include weather, pollutants, time of wetness, and the nature of corrosion products (12). ' Heather conditions affecting corrosion rate or type include relative humidity, dewpoint, temperature and rain fall. As Vernon (13) found, a, perceptible increase in rate of corrosion occurs above a "critical humidity". Below this point the corrosion rate is slow. The value of the critical humidity is dependent on the material. Rain is considered a harmful source of moisture because it provides moisture and washes away soluble corrosion products. On the other hand, it may be beneficial since it also washes away pollutant deposits. Atmospheric pollutants are substances which accelerate corrosion or induce condensation. An important pollutant is sulfur dioxide. Atmospheres polluted with sulfur dioxide have been found to be among the most corrosive. 21 Another class of pollutant is particulates. Particulate sizes vary from .01 to 10 um. Oust particulates for example, influence corrosion rates markedly, especially if the dust consists of water soluble particles. A duration of the moisture on the surface or, as Grossman calls it “Time-of-Hetness", is an important parameter for finding the interaction of a given material with a specific environment. In section II. E., we will define the condition of dew appearance on the outdoor surfaces. Corrosion products can act as a barrier and as protection between the atmosphere and metal. There are cases, however, where corrosion products are not protective, e.g., when soluble salts are present. The formation of corrosion products proceeds directly in an anodic region of the metal surface. The protective characteristic of corrosion products depends on the time of year in which the metal is exposed to the atmosphere for the first time (8). In winter, the greater surface accumulation of combustion products (mainly sulfur compounds) produces a less protective initial corrosion product which affects. the subsequent corrosion rate. Aircraft are subjected to different types of atmosphere and environmental conditions and exhibited several modes of corrosion: Pitting Corrosion: Pitting corrosion starts with the breakdown of passivity at favored nuclei on the metal surface (14). The breakdown of nuclei has not been definitely clarified. The work by Foroulis and Thubrikar (15) indicates that pit nucleation is caused by the absorption of chloride ions under the influence of the electric field, followed by the formation of soluble aluminum hydroxychlorides. This process will have a high probability of repeating itself and penetrating at the same site at a constant electrode potential. The electric field will tend to be strong at the point where the oxide film is thinned by the dissolution process. 22 Metal used in a passive condition, such as aluminum alloys, Fe-Cr-Ni alloys, are sensitive to pittng corrosion, especially in environments containing chloride ions. The propagation of a pit is caused by a large difference in the composition of the environments, in and outside the pit (16). Crevice corrosion: Crevice corrosion is produced at the region where non-metallic materials are in contact with passive metals. This kind of corrosion is a form of local attack which can be observed within crevices or other shielded areas of metal surface exposed to a corrosive environment. The process is triggered by a local difference in composition of the environment. Galvanic Corrosion: Different metals may be present in the same construction. If these metals are in direct contact with each other, and the construction is exposed to a corrosive environment, the corrosion rates may be different from those of single metals in the same environment. The metal with the lowest corrosion potential, when exposed separately, wdll show an increased corrosion rate when it is connected to a metal with a more positive corrosion potential. The latter material will have a decreased corrosion rate. The anodic reaction dominates on the part of the combination with increased corrosion rate (anodic area); the cathodic reaction dominates on the part of the combination with the decreased corrosion rate (cathodic area) (16). In a system which shows galvanic corrosion, the ratio of anodic to cathodic area is of importance. If the combination has to be used with a large difference in the corrosion potentials of the separate parts, the cathodic area should be small compared with the anodic area. Intergranular Corrosion: In a homogeneous polycrystalline metal, crystals of different orientation are separated by grain boundaries. In these transition regions the atomic packing is more imperfect than in the matrix. In a corrosive environment these regions can be preferentially attacked in 23 comparison with the adjacent crystal faces (16). The difference in reactivity is usually so small that macroscopically a uniform attack is observed. There can be, however, large differences in reactivity, if along the grain boundary, zones are formed with a composition different from the matrix. Several engineering alloys exhibit this metallurgical state under certain conditions of composition and heat treatment. Corrosion Fatigue: Metals can fail by fatigue in an inert environment, if subjected to cyclic stresses lower than the static yield strength. In most nonferrous alloys, the allowable cyclic stress always decreases as the required number of load cycles decreases. In ferrous alloys, an endurance limit is reached below which metal apparently can by cycled indefinitely without failing. Corrosion fatigue is the combined action of a corrosive environment and cyclic stress. It leads to an appreciable reduction in fatigue life (16). The fatigue process starts with a crack initiation period which is followed by a visible propagation of the crack. There is no general agreement how both periods are divided over the total fatigue life. One of the problems is that the length of the initiation period depends on the sensitivity of the crack observation method. In an inert environment, the crack initiation starts with the formation of cyclic slip bands at the surface, gradually developing into intrusion and extrusions, at which a microcrack along a slip plane or a grain boundary is formed. The initiation process is accelerated by corrosion reactions. In general, four principal effects are considered: pitting corrosion, preferential solution of anodic aareas induced by deformation, protective film destruction, and surface energy reduction by absorption of species present in the environment. The initiation process is only accelerated if the general corrosion rate of metal exceeds a cri tical value (17) . 24 2. MAGNESIUM ALLOYS Several magnesium alloys have been used as wing and cowl skins of B-52's. The use of magnesium, however, is limited because of the inability of the protective film as a barrier in neutral and acidic conditions. This, together with its high electromotive potential, severely limited the application of these alloys (18). Studies of the corrosion of magnesium-rich alloys, with and without protective coatings, were begun in 1929 with the object of improving corrosion resistance. Corrosion was found to be of two distinct types: a) Pitting, with no evident relation to the microstructure of the metal, analogous to the rusting of ferrous metals, and b) intercrystalline, where only slight evidence may be seen on the. surface, but which penetrates into the interior, largely between the grain boundaries (18). . Protective Film: Electron diffraction studies indicate that magnesium alloys, when exposed to humid air, exhibit excellent stability, with oxidation rates 0 less than 100 A per year, provided no condensation occurs (19). When the air is saturated with water vapor so that condensation occurs, vacuum-deposited 0 films of magnesium about 400 A thick are destroyed in less than a day. In this case, in addition to amorphous material, the film contains crystalline magnesium hydroxide (brucite). The good resistance of magnesium to water near room temperature decreases as the temperature increases. Alloy AZ318 has a corrosion rate of 17 mpy in . water at 100 oC compared with 1200 mpy at 150 °C. The stoichiometry and equivalence of all oxygen positions in brucite indicate that it is a true hydroxide and not a hydrous oxide. It has a layer 25 lattice and undergoes the easy basal cleavage expected from such a structure (20). This interlayer weakness might have been responsible for the apparent cracking and curling of the film that was observed in electronmicrographs of the surface of pure magnesium that has been exposed to distilled water (21). Reaction of hydroxide film with acid gases in the atmosphere has an important influence on the weathering of magnesium. The atmosphere contains a substantially constant 300 ppm of C02, and contains less than one part per billion of S02, unleSs polluted by nearby sources of combustion products, when the $02 concentration may reach a few ppm. Bengough and Whitby (22) and the X-ray diffraction analysis of Casey and Maupin (23). showed the dominant film anion after weathering is COZ‘Z. The latter found 3M9 CO3.Mg(OH)2.3H20 (hydromagnesite), 5MgCO3.Mg(OH)2.9H20,MgCO3.H20 (nesquehonite). and M93.5H20 (lansfordite) in corrosion product scrapings from electrolytic pig stacked outdoors at an urban-industrial site and at a rural site for 18 months. MgSO4.6H20 and MgSO3.6H20 were detected by diffraction only on scrapings taken from within the stack at the urban-industrial site. A similar examination of corrosion products from A2318 alloy sheet exposed outdoors by Bathwell (24) for four years revealed only the diffraction pattern of hydromagnesite and Mg5Al2(OH)15CO3.4H20 (Hydrocalcite). Magnesium is in the high electromotive metal series, therefore, it is subjected to galvanic corrosion when coupled with another metal. Hhen exposure to marine mists or de-icing salts is likely, an important first step in controlling galvanic corrosion is to provide appropriate drain holes in the design of the structure so that chloride solution does not accumulate at the couple. If the dissimilar metal juncture is well sealed, so that the solution does not enter between the faying surfaces, the current in the electrolyte 26 portion of the circuit is forced to flow along a thin and therefore highly resistive film. This limits the galvanic activity almost entirely to zones about 1/8 in. wide on either side of the line of juncture between the two metals. Even then the megnesium alloy within this zone can suffer severely if the cathode metal does not polarize readily. Magnesium alloys do not show any tendency toward intergranular corrosion except in chromic acid containing traces of chloride or sulfates. Primary magnesium ingots can undergo a peculiar form of layer corrosion which Schiebold and Siebel (25) found to be clearly transgranular. In this type of corrosion, the large grains of the ingot are corroded preferentially along their basal planes. If the angle between the basal plane of a grain and the ingot surface is less than 40° (Schiebold and Siebel reported 29°), the corrosion continues until the entire grain is converted to thin layers of metal pushed apart by corrosion products. If the angle is greater than 40°, pits develop that are elongated the basal plane, but there is no layer formation. Thus, straining of the angle in a direction of poor support by stresses produced by the corrosion products appears to be essential to the stress. Chloride solutions are necessary for lamellar corrosion to occur (26). The requisite solution can be formed by the absorption of water from humid air by chloride contaminants on the ingot surface. Potential sources of chloride contamination are all electrolyte and (or) melting flux from the pigging operation, sea mists, and industrial dust and fumes. In magnesium alloys exposed to strong electrolytes, severe pitting can be caused by cathodic impurities and surface contamination such as mill-scale from rolling or carbonaceous lubricant residues from forming, forging or impact intrusion. Magnesium alloys, either painted or bare, can suffer filiform corrosion 27 when exposed to humid air, even at times when wholly immersed in chloride solution. The formation of tiny, meandering trenches during exposure to an electrolyte in bulk has has not been previously reported for magnesium or any of the many other metals that undergo this form of corrosion (26). Crevices and poultices in magnesium structures should be avoided since they are potential reservoirs of electrolytes. Bothwell (26) has observed crevice corrosion at the unprotected faying surface of pure magnesium assemblies and of HK31A alloy assemblies immersed in 3% sodium chloride solution. Crevice corrosion was not observed on A2318 alloy exposed under the same conditions. Differences in corrosion rate of various lots of HK31A alloy during exposure to 3% sodium chloride were small in the absence of crevices, and did not relate to the crevice corrosion susceptibility. However, the more susceptible lots have higher corrosion rates in MgClz solution than the less susceptible lots. It is probably significant that A2318 alloy has substantially better resistance to magnesium chloride solutions than either pure magnesium or HK31A alloy. Thus all the facts indicate that crevice corrosion is another case of localized attack caused by cation (and anion) concentration cells. Loose and Barbian (27) have discussed the stress-corrosion testing of magnesium alloys. They pointed out that residual stresses from fabrication are subject to relaxation through creep. Bothwell found that, of the commercial alloys only wrought ZK60A and alloys containing more than approximately 1.5% aluminum are sufficiently susceptible to stress-cracking to warrant preventive measures. Castings are generally much less susceptible to stress cracking than wrought products of similar composition. Thus, A2318 sheet and extrusion, ZK60A and forgings are moderately susceptible, whereas cast AZ318 alloys seem to be immune. The weathering of magnesium alloys can be explained in terms of two 28 processes (28): 1. Conversion of a protective surface film to soluble bicarbonates, and sulfates. 2. Stimulation of local cells by chloride ions. Except in close proximity to the ocean, the first process is dominant. Hhen the magnesium alloy contains aluminum, it is found that aluminum ions concentrate in the film during weathering as the magnesium ions are leached away. Although there is only about 1 aluminum atom to 40 of magnesium in AZ318 alloy, hydrotalcite, which has a ratio of 1 aluminum atom to 4 magnesium atoms, concentrates in the film. This concentration of aluminum in the film is greater on the rainwashed skyward surface of the panels, and is increased by a high level of atmospheric sulfur dioxide. This resistance of the aluminum content of the film toward dissolution is a probable reason for the improved resistance to weathering of magnesium when it is alloyed with aluminum. . 1 Magnesium exhibits inherently different corrosion rates depending on the amount and type of alloying element present and location and humidity. Humidity playes a major part in the corrosion or tarnishing of magnesium and its alloys (28). Magnesium alloy AZ318 shows very good corrosion resistivity even at 98% relative humidity. Dry sulfur dioxide plus humidity produced an increase in corrosion rate to those found in the moist air (28). 29 3. ALUMINUM ALLOYS Aluminum surfaces oxidize rapidly when exposed to air, and acquire a dense, adherent, protective aluminum oxide film which retards further oxidation. The oxide film is stable over a pH range of about 4.5 to 8.5. The pH value alone, however, does not determine the solubility of the film, because the presence of certain anions and cations, as well as OH‘ and H+ ions, exert an influence (29). The film is dissolved by most strong acids and alkalies, and accordingly,.aluminum is termed an atmospheric metal. It consists essentially of amorphous aluminum oxide (Al203) and in various degrees of hydration (Al203oxH20), depending on history, especially the condition of relative humidity and the temperature of formation (30). When aluminum is exposed to a moist atmosphere or immersed in water, the oxide film thickens (the growth rate is much more rapid in water). In both cases the rate of growth increases with temperature. Although dissolution of aluminum in specific chemicals will produce distinctive Corrosion products (e.g., aluminum nitrate in nitric acid), by far the most common corrosion product is hydrated aluminum oxide. The main difference between the oxide produced by corrosion and that in the surface film is that the oxide on the surface forms compactly and has a protective influence, whereas the oxide in corrosion is produced at a distance from the metal-environment interface and is nonprotective. The temperature increase has a strong accelerating effect on corrosion of aluminum. However, in the atmosphere, heat can be beneficial, because it increases the rate of drying and thus reduces the period of wetness. Also temperatures above 40 °C tend to reduce the rate of pitting of aluminum in water. The alloying elements: Magnesium has a beneficial influence and Al-Mg alloys have good corrosion resistance. Magnesium increases the resistance of 30 the aluminum alloys to alkaline solutions such as lime and sodium carbonate. Copper reduces the corrosion resistance of aluminum more than other alloying elements. It leads to a higher rate of general corrosion, a greater incidence of pitting, but when added in small amounts (e.g., 0.152), a slower rate of pitting penetration. The influence of copper is greatly dependent on the amount present, its form, and distribution in the alloy microstructure. In general, aluminum-copper alloys (ZXXX series) have relatively poor corrosion resistance, and require surface protection when used in corrosive environments. Without protection, they suffer extensive corrosion in marine and industrial atmospheres. Wrought Al-Cu alloys are protected by cladding- with commercially pure aluminum, and in this form in the absence of excessive diffusion and when properly heat treated, these alloys become the-most resistant to atmospheric corrosion. They show negligible deterioration even after 20 years in severe marine atmospheres, even in the fbrm of 0.050 in sheets. . Aluminum-zinc alloys (high-strength aircraft alloys 7075, 7079) require protection in severely corrosive environments. They tend to have relatively low resistance to stress corrsion in the short transverse direction. Galvanic corrosion: When aluminum is in contact with a dissimilar metal other than magnesium, zinc, cadmium, or chromium, in the presence of an electrolyte, it tends to corrode more rapidly than it would if it were exposed by itself to the same environment. Galvanic corrosion of aluminum is negligible in inland rural atmospheres, and in mild industrial atmospheres. It is pronounced in marine atmospheres, depending on the amount of wind-blown spray, and is worst on board ships. A marine atmosphere may be defined arbitrarily as that along the sea coast for a distance of about 1 mile inland, depending on the prevailing wind and 31 local topography. The severity of galvanic corrosion decreases rapidly with distance from the shore. Crevice corrosion of aluminum in the atmosphere is negligible except in a marine atmosphere, in which it can become appreciable and is probably proportional to the amount of windblown salt spray. Crevice corrosion of aluminum in marine atmospheres has been noted in a five-year exposure test of plate samples containing crevices from which the following recommendation was based. “In the design of aluminum structures which are subjected to marine atmosphere corrosion, no provision need be made for corrosion in aluminum-to-aluminum crevice for service life of less than five years, except where very thin wall sections are involved (e.g., less than 0.04 in.) In structures to be exposed to a marine atmOSphere for longer than five years, measures should be taken to prevent crevice corrosion. These consist of coating faying surfaces before assembly with an inhibitive paint system and, where possible, filling the crevice with a jointing compound or resilient gasket material to prevent the ingress of moisture," (31). PTTING CORROSION: Pitting is the result of electrochemical action in local cells on the surface of a metal. At the point of initiation, corrosion occurs at local anodes, while the cathode is the immediately surrounding metal surface. There are four possible stages in the life of a pit: initiation, propagation, termination, and re-initiation. Some pits initiate and terminate in a relatively short time. Others first initiate, propagate for longer times, and then terminate, while others continue to propagate (31). The resistance of aluminum to initiation of pitting depends to a marked extent on its purity; the purest metal (zone refined) is the most resistant. The presence of other elements in aluminum except for Mn, Zn, Mg, V, Ga, either as impurities or as deliberate alloying additions - increases its suseptibility to pitting in fresh water. 32 Intergranular Corrosion: Those aluminum alloys in which grain-boundary precipitates do not occur, or those in which the precipitation has the same solution potential as the matrix, are not susceptible to intergranular corrosion. The Al-Zn-Mg-Cu alloy AA-7075 is susceptible to intergranular corrosion under unfavorable conditions. The degree depends on the heat treatment, and is least in the T-73 condition. Cladding (with AA-7072) is an effective preventive measure for AA-7075. Cathodic grain-boundary precipitates which can develop in aluminum alloy are CuAl and Si. The most susceptible alloy of this type are AlZCu alloys. However, in a properly heat- treated condition, AA-20l4 and AA-2024 are susceptible to intergranular corrosion only in severe environments, such as highly industrial atmospheres, severe marine, or immersion in water. Stress Corrosion: Cases of stress-corrosion cracking have been limited_ mainly to the so-called "strong" wrought alloys such as Al-Zn-Mg-Cu (AA-7075) and Al-Cu (2024). High-strengh aluminum alloys have been used fbr years in aircraft construction. The very few failures that have occured have resulted mainly from residual stresses acting continously in the short transverse direction relative to the grain structure. Longitudinal stresses, on the other hand, have rarely caused stress cracking of aluminum alloys. It is important, therefore, to recognize the sources of stress, and to design for their elimination or reduction. For AA-7075, it has been shown that lowering the temperature of aluminum heat treated from 465 °C (normal) to 435 °C increases the resistance to stress corrosion, though at a sacrifice of mechanical properties. Corrosion Fatigue: Aluminum alloys, like many steels, have relatively poor resistance to corrosion fatigue. According to Liddiard (32). failure by fatigue normally occurs in two stages. In the first stage, (1) the crack starts along slip planes, particularly those oriented parallel with the 33 shear direction, and then propagates into the material in the form of a crack at an angle of 450 to the direction of stress. From there it may change direction to the second stage, which proceeds at right angles to the direction of stress. Liddiard reported that, in addition to lowering the fatigue limit of aluminum, corrosion tends to alter the mode of failure and often reduces stage 1 and causes several stage 2 cracks to form, most of which are nonpropagative. For aluminum alloys, in which corrosion fatigue initiates by stage 1 cracking, cathodic protection by impressed current or by sprayed coating (that is anodic) is effective in prolonging the life, although the mode of failure may change to stage 2. Cathodic protection is ineffective, however, if the fatigue failure initiates as stage 2 (30). Atmospheric Corrosion: The corrosivity of the atmosphere to metals varies enormously from one place to another. The most important factors that affect the rate of metallic corrosion in the atmosphere are the degree and type of chemical pollution, the period of wetness of a metal surface (33-35) and the temperature. The most common corrosive pollutants are sulfur dioxide from burned fuels (especially coal) in urban and industrial areas, and wind-borne salt from the sea. Thr relative corrosivity of the atmosphere depends on the metal. Aluminum alloys corrode in the weather by mild roughening of the metal. The maximum depth of penetration is the measure of the extent of corrosion. The Al-Cu and Al-Zn-Mg-Cu high strength alloy groups are an exception, and may suffer intergranular or lamellar corrosion. It is hard to detect the corrosion of aluminum in rural areas. It is usually less than .Ol mpy. AA-2024 has a poor resistance to corrosion in marine atmospheres and requires surface protection in the form of cladding or Inaintained paint coating to avoid corrosion. In a marine atmosphere the surface of aluminum dulls to a grey color in a few years. Most aluminum 34 alloys have good corrosion resistance in industrial atmospheres deSpite acid conditions on the surface which result from sulfur, producing pitting, surface roughening, and a general loss of strength. The Al-Cu alloys have low corrosion resistance to industrial atmospheres. In the tropical atmosphere, the alumunum surface remains the same and the amount of corrosion is negligible. 35 E. PACER LIME PACER LIME was first initiated in 1965 by the needs of the Strategic Air Command and the Air Force Logistics Command to develop a corrosion severity classification for each operational airbase. It was later implemented in 1971. PACER LIME was a two-phase effort: (1) Development of an equation or algorithm for computing 3 251251 a numerical corrosion factor which combines weather and other environmental factors; (2) experimental measurement of corrosion severity at selected locations through atmospheric corrosion tests. The experimental data would be used to “calibrate" the computed corrosion factors. An initial corrosion factor equation, combining certain weather geographical factors, was developed in 1971. Interim numerical classifications were published for SAC airbases in 1972, and for 95 USAF and 27 ANG airbases in 1973. A complete list of corrosion factor ratings, combining certain weather and geographical factors, was distributed in 1974 under the title” PACER LIME Interim Corrosion Severity Classification.“ These interim values were to be compared with corrosion maintenance experience and the results of the PACER LIME atmospheric testing program. The need for environmental guidelines was so great, that PACER LIME ratings were used to develop maintenance interval guidelines, e.g., for washing and corrosion inspection, despite the poor correlation between PACER LIME results and field-level experience. The experimental phase of PACER LIME would provide a calibration reference point for the corrosion factor equation by measuring corrosion rates at several airbases. To complete the program, it was assigned under a contract to Michigan State University in 1978 because of inadequate in-house USAF resources. The objectives of the contract were to complete the program by analyzing results 36 of the corrosion exposure test program, the Base Corrosion Severity Classification, and to develop an improved classification system. 1. TEST METHOD AND MATERIALS Test stands were constructed to hold about 125 test panels by means of porcelain insulators at 30° to the horizontal facing prevailing winds. Test sites installation was accomplished at eight sites in March 1972, two more in September 1973, and the last one in late 1975. Six alloys were selected and tested in three different configurations. The riveted assembly of three aluminum alloys was intended to provide galvanic corrosion. Thus the corrosion damage to these panels should represent fairly the behavior of aircraft in a given environment. The corrosion of steel and magnesium was thought to be predictable and fast enough to give results in a reasonable time. Since aluminum usually does not exhibit extensive general corrosion, it was anticipated that sporadic, difficult to interpret results might be obtained. The corrosion behavior of titanium was considered predictable, but it also was known to be exceedingly slow. 2. CORROSION RATE COMPARED WITH ENVIRONMENT Sufficient data were available for environmental comparisons for A2318, 2024, and 7075 alloys only. In the case of 4340 and the aluminum assembly, there were only four data values each, hence they were insufficient. For 7079 there also were only four values, but there were several corrosion rates available from other studies to warrant comparison with the environment. A semi-quantitive comparison with environment is made by plotting experimental corrosion rate vs the Corrosion Damage Algorithm rating for the respective test site. This is shown for the alloys AZ318 in the Figure 3. Shown with the PACER LIME corrosion rates are the values obtained for the same alloys in other studies. 37 For the various sites, the Corrosion Damage Algorithm (CDA) provides a two-letter scale, viz., 88, AB, etc. The first letter refers to the rating derived from a set of less-tolerant threshold values for environmental parameters, and the second from a more-tolerant set. Thus a second-letter A indicates a more severe environment, with respect to one or more environmental factors, than does a first letter A. Environmental ratings range from the mildest C through 8 and A to the most severe AA. For the purpose of data plotting, these letters have been assigned a numerical scale of 1 to 4 for C to AA, and the two letter values are summed. Thus an A-B environment yields the sum 5, and AA-AA, yields 8. Data for the magnesium alloy, Figure 2, show very good correlations with the CDA rating, with only one or two discrepancies. The ASTM data for State College PA and Newark NJ both are high for their respective environmental ratings. The PACER LIME data, however, are quite consistent. In the case of 2024 Alclad, nearly all the results are consistent with the exception of the ASTM McCook IL and Richmond VA values. The data for 7075 T6, are similar to those of 2024 T3.- The PACER LIME 7079 T6 data consist of only fbur points, which are plotted in together with ASTM results. They are all consistent with the CDA environmental ratings. 3. THE CORROSION SEVERITY CLASSIFICATION SYSTEM a. ENVIRONMENTAL VARIABILITY The variability of environmental corrosion severity has been well established by atmospheric testing programs (36-41). Relative severity is commonly indicated by designating an environment as rural, urban, industrial, marine, or an appropriate combination of these terms. Moreover, many studies (42-43) have shown that certain environmental factors, e.g., moisture, salt, pollutants, are responsible for the more rapid corrosion observed in 38 2 6 L. Magnesium A2318 4 .. 2 __ {DO ON OR? -4 ,, 10 a — '9 I- i. a - . To - 0 so Ca ‘ KB .3 6 L_ ‘.. 3 - 011-? PR 2 4 _ 5 o R A Andrews MD '3 .. o “‘1’ 8 Barksdale LA 3 o '1' o A D Davis Monthan AZ ‘6 3 R K8 Kure Beach NC 0 2 _ ‘ ’1' 3 N Newark NJ . 0 Panama Canal 3 3:3":1. el d RF Rain Forest PR’ Point Reyes CA 10-5 _ R ROanS GA _ T Tinker OK' 8 ._ SC State College . _. w W-P Wr1ght-Patterson 6 II- Corrosion Severity Index (Combined) Figure 2. Corrosion Rates of AZ31B Magnesium alloy compared with Environmental Rating. 39 environments containing them. Consequently an environmental rating scale which takes into account those factors could provide a more useful indication of relative severity which could be used in management of aerospace systems. It would be difficult to devise a general rating system which would predict the corrosion damage to every metal. Different metals display widely diverse behavior in a given environment and also from one environment to another. Some alloys are more resistant in marine locations than industrial and the reverse is true for others. The several factors which influence corrosion are present in a unique combination for a given site, and precise information relating the corrodibility of a specific alloy to every environmental factor is not available. b. Environmental Severity Algorithms for Aircraft Corrosion We proposed an alternative set of algorithms, based on locally-measured environmental factors and which rely in part on maintenance experience as contained in AFM66-1 records. Three algorithms rate environmental severity with respect to corrosion damage to alloys and corrosion protection systems from the ambient level or concentration of several parameters compared with a set of Horking Environmental Corrosion Severity Standards (WECS). The WECS were established from the extreme and median values observed in the continental United States. If a parameter exceeds the WECS level, it is rated corrosive, whereas a lower value is rated noncorrosive. The WECS environmental parameters listed in decreasing order of significance for corrosion are (a) proximity to airborne salt source, e.g., seacoast, (b) moisture as humidity or rainfall, and (c) atmospheric pollutants, including particulates, $02, photochemical oxidants as 03, and nitrogen oxides, NOx. In evaluating environmental risk to corrosion protection systems (e.g., paint), 40 the intensity of solar radiation also is considered. There are questions about the relative weighting of these factors in the algorithms. High humidity and high rainfall are considered equivalent, but this is a minor point, since these factors are generally parallel. Atmospheric pollutants are weighted equally, but evidence that oxidants or "particulates“ are as corrosive as $02 is less than compelling. The corrosive effects of salt, S02, and moisture, especially in concert, are well established. The algorithm could be revised so as to deemphasize some factors by raising the threshold values. Summitt (4) drew an algorithm for washing, repainting, and corrosion damage Figure 3. The Corrosion Damage algorithm (CDA) considers first distance to the sea; if very close, it leads either to very severe (AA) rating or a consideration of moisture factors. After moisture factors, pollutant concentrations are compared with WECS either for S02, TSP, or 03. High values for any one of the three pollutants together with a high moisture factor leads to the A rating, but if all are low, together with high moisture factor, the severe (8) rating results. Low moisture factors with a high pollutant value result in the moderate (8) rating whereas if all are low, rating (C) results. Using these algorithms, airbase classifications have been obtained which were in good agreement with USAF maintenance experience, from AFM 66-1 maintenance records and in good agreement with experimental testing results. >43 >61 )36 543 543 $61 $61; 536 536 c a EIG-‘ECTED CORROSION DAMAGE AA VERY SEVERE A SEVERE B MODERATE C {TLID Figure 3. Aircraft Corrosion Damage Algorithm 42 E. TIME-OF-WETNESS AND ENVIRONMENTAL PARAMETERS Grossman (5) investigated the relationship of Time-of-Wetness and corrosion rates and found a direct correlation between them. Time-of-wetness, in principle, should be calculable from meteorological conditions, which lead to dewfall. When certain meteorological conditions coincide shortly after sunset, dew collects on outdoor surfaces and usually disappears an hour or so after sunrise. An investigation follows concerning the conditions necessary for dew to appear as well as the amount of time needed for evaporation, thus leading to the time-of-wetness for given ambient conditions. At night, outdoor surfaces radiate heat to the sky and become colder than the ambient air. An object cooler than the surrounding air, however, will absorb heat from it because of convective flow of air over the object. The radiative cooling effect is opposed by a convective warming effect; and when this effect is absent, the object temperature drops slowly and approaches the equilibrium temperature. To show this transformation, consider a thin horizontal plate of area A and thickness L at temperature T. Assume that the lower surface is isolated, and that the upper surface is suddenly exposed to the sky at effective temperature Tsky at time t = O. The plate will lose heat by radiation to the (cold) sky, and will gain heat by convection from the (warm) air at temperature Tair° Because of the insulation on the lower surface, both radiation loss to the ground and conduction loss to the supports will be negligible. By the First Law of Thermodynamics, the rate of net gain of heat to the plate, dq/dt, will equal the rate at which work is done by the plate, dW/dt, plus the rate of internal energy gain of the plate, dE/dt: dq/dt = dW/dt + dE/dt . There are three contributions to dq/dt, namely, from conduction, radiation, 43 and convection. Under the conditions assumed here, they are respectively: d /dt = 0 ; qconduction 4 4 dq /dt = e a A (T - T ). radiation sky where e is the emissivity of the plate and a is the Stefan-Boltzman constant; and dq , /dt = hA (T - T ). convect1on air where h is the convection heat-transfer coefficient for still air at the ambient conditions. Since the plate does not move and its volume change is negligible, the work done~is zero, and dw/dt = O. The rate of change in internal energy of the plate is just the rate of change of sensible heat: dE/dt a CpLA dT/dt, where C is the specific heat of the plate material, and 9 its density. Upon collecting appropriate terms and eliminating A, the differential equation for the approach of the plate temperature to equilibrium is dT/dt = (1/ch) [-ea (14- r4 1 + n (1 , -1)1. sky a1r Although the exact solution of this equation is staightforward, it is tedious. Since the drop in temperature of the plate is small compared with the air temperature, it is convenient to introduce the auxiliary variable ZE(T.-T)/T a a1r ir 44 _ This quantity is small compared with unity; its differential dZ is -dT/Tair. The T4 term is then expanded by the bionomial theorem, and only the first-degree term in Z is retained. If further we take Tsk ” O K, the differential equation becomes d2 3 1 [(4 + h ) _ 1 ], "73" dt pCLT ea air air which is easily solved to give -t/r T - T(t) = f T , (1 - e ). air a1r with the definitions "'3 "I 1 / (4 + h/eaf3 ) air and "I III pCL/(4caT3 + h). air It is apparent that T(t) decreases from T(O) = Ta rto T(~) = (1 - f)Tair' whence f is to be interpreted as the ultimate fraztional decrease in initial temperature. The fraction f is seen to decrease with increasing h and decreasing c. The temperature drop proceeds with time constant t, which increases with the thermal inertia of the plate (density 9 times specific heat C times thickness L) and decreases with the emissivity e and the convective coefficient h. A numerical estimate is given for a representative case, a 1-mm thick aluminum plate at air temperature 30 0C: L = .001 m p = 2707 kg/m2 C = 896 J/kg PC o a h a e a T 3 Then f = and T 3 Thus in a the plate 45 5.62 x 10'8 w/mZ-k4 10 W/mz- °c 0.1 303 K. o 0.147 = 1/68.0, and ft . = 4.5 C ; a1r 226 sec. time interval of 3 times 226 seconds, or a little over 11 minutes, temperature drop will have attained all but 1/e3 - 52 of its 0 ultimate value, and will be about 26 C. The dewpoint for 30 °C air with 792 relative humidity is 26 °C, and hence we may expect dew to begin forming at about 11 minute in 79% RH air. Moister air will produce dew sooner, drier air never. Moreover, if the cloud cover is considerable, the effective temperature of the sky will be higher, and the temperature drop will not be so low; hence dew will be less likely to form. ‘1- )- I‘m- ” T-3OO // . I m I- " r-275 / 1'11 \\ ¢- ‘2‘- {8‘1- T=250 Illi- 4oo ”0 Mac) Figure 4_ Panel Temperature y_s_ Time. III. THE USAF MAINTENANCE DATA COLLECTION SYSTEM (MDCS) USAF extensively documents maintenance actions which correct failures or modify airplanes. Routine service, e.g., washing, cleaning, and touch-up painting, was documented in less detail until recently. "MDCS is the primary source for AF reliability and maintainability data; basic understanding of its objectives, uses and limitations is essential to R 8 M [Reliability and Maintainability] users. The system was designed primarily as a base level production credit and management information system. Objectives are to provide [field-level] maintenance managers with information about production accomplished by the assigned maintenance personnel; to identify the equipment on which work was accomplished, why it was required, and the action required to do the job. The MDC system identifies maintenance requirements and problem areas so that appropriate management action can be taken to effectively support [sic], and meet the established operational requirements. In addition, the system is designed to provide data to AFLC for maintenance engineering and logistic management." (44) A. INSPECTIONS A maintenance action is initiated by a Discrepancy Report, which usually --particularly in the case of corrosion --is generated in the course of a scheduled inspection. Authorized maintenance and inspection programs currently in effect for 8-520 airplanes* are listed in Table 3. In addition to the scheduled inspections listed in Table 3, there are several other special inspections. Each discrepancy report carries a "When Discovered" designator to indicate the type of inspection in effect. *H and G series inspection programs are not available, but presumably are the same. 47 48 mDODzHHzoo 2am mHHS ZOHHUZSHZOU «59262 we anew CON MHHmmHzH 883 £282 EEBSH gunman Samoa: 303 52.8.55 82.328 48:»; Aznmv uozNA Roman onHommmzH ammNA OH H Qmo mz<fluomm ZOHHONmmzH\uozuum~=a=o .N muawuh mmmp paw cow paw ll N llamsueg gueoz\mc=o;-=az m. ;.=Oz\mcaog-=az ma mp<¢ muz311aowumuwcmmuo 2.33.550 asap amp :2. .54 :84 .54 gas .54 gas m. u. fig 11 m" w m. mv 1. X. a at I. a \‘O a J cI|\\ta|5 .... s. ... w \L. 00:0: 0:98.— \51 “On-0n— .. . suegxtaogueuz 8 5:22:23... 8 2.5. 3522.22: gm: .osodunaoa .oqmoocam 568552 .w ouswa @3— m3— pan. :2. .3. :25 :5 11..-- c \\ \\ 5 OJ \\\ a U. a H. .m mm .1.ocm w. a. s ...r a m m m. m .. w m m. w W o m w. IJ 89 V s 1 A 4.. mm 1! fl (laces mm «Lao: l1 8N— ucaou IIIIII wwmnm @38on L 8....— 18o— ) 9:8 ..| 8cm 1 82. .l: 2.3 SUDOH'NW 3411v1nwn3 78 .m~¢film~ad .acoooooo Hana“: uuauom onudnud< mNnIn ac m8. @3— 22 _:a cow .aa cua paa coa Va w ||.|| m S l D. .m 35:35.: 3 ~88 \ ‘ \\ st... II...‘ ‘Ks. 5:9{23gé2 we figs—{952-55 @— ;uchmsagucg me unannoufluux flo«00Huoo w>qumH=E=U .m wuawwm 22 22 .2. =2. .2. =2. .2. - o 3 II can m. o a D n H as m 1 m In 3 w H n .1 98 n S mL:b = J 82 acacu uuuuuuu :lem L Dcmp 388cm coo p 88 coon 7 25¢ F 88 SUflOH'NW 3A IlV'lflHflD 79 maintenance for a given airbase is a constant regardless of what factors may drive it. This rate reflects environmental corrosivity and other factors which determine the amount of effort . A second method of determining field maintenance rates is to determine the total effort by aircraft serial number from AFM 66-1 data and compare this with the time period the airplane was assigned to that airbase, i.e., total manhours divided by possessed months. 4. STATISTICAL DISTRIBUTION OF FIELD MAINTENANCE The statistical distribution of corrosion maintenance data was determined for field maintenance rates measured by both slopes of cumulative maintenance vs, time and total maintenance divided by possessed months for each airbase. The data were plotted as monthly manhours per airplane by serial number against the per cent of total population for each airbase, as in Figures 10 and 11 for B-SZD and H systems. Similar data plotted for the G—series are comparable, but are not reproduced because the number of similar airbases yields too complex a graphical presentation. The data are summarized, however, in tabular form together with those of the D-and H-series. Figures 10 and 11 are charts of the data obtained by dividing total manhours by possessed months. Plots of slopes from cumulative maintenance li- time are similar but are not reproduced. The results closely parallel those of the C-141A system in that the data are distributed essentially normally. From these charts one can deduce three types of information: (a) the data are more-or-less normal, (b) a mean value, and (c) the standard deviation. These data are collected in Tables 19-21 for all series. Results are shown both for values obtained from the cumulative maintenance 12- time charts as well as the total effort 22° possessed months values. Again as in the case of the C-141A, there is a striking difference from one airbase to another. In the case of the D-models, monthly manhours MEAN MONTHLY MAN-HOURS PER AIRPLANE 80 llO . o 100- o o 90 - O 80- O I o oo o 70'. o . O o 'I o I 50 .. ‘ ‘ ° -- , o o . A o . A 50 P ' ' Ci A A A o o 40 " ' Andersen o _ o Carswell . 3° F . ' ' March A ‘ Dyess 0 20 1. 1 ‘ 1 1 I, I J .1 l J, .1 1 1. l 2 5 TO 20 30 4O 50 60 7O 80 90 95 98 99 PERCENT OF TOTAL POPULATION Figure 10. Individual Airplanes, 1Q78-3Q79. B-SZD Organizational-Level Corrosion Maintenance to MEAN MONTHLY HAN-HOURS PER AIRPLANE 81.. I I A O 0 O I I ° . o‘3 0 a 0.. A . D 0". 0 Castle - K.I. Sawyer O 0 ° . ‘ Ellsworth . lO -— a _ Grand Forks A Minot 0 o l J 1 1 J 1 1 l l I I L L I 2 TO 20 30 40 50 60 70: 8O 9O 95 98 99 RECENT OF TOTAL POPULATION FigurelJ.. B-SZH Organizational-Level Corrosion Maintenance to Individual Airplanes, 1Q78-3Q79. 82 TABLE 19. B-SZD ORGANIZATIONAL-LEVEL CORROSION MAINTENANCE, MEAN MONTHLY a MANHOURS PER AIRPLANE, 10 1978- 30 1979. AIRBASE CUMULATIVE EFFORT MEAN MONTHLY EFFORT ANDERSEN 73 (22) 107 (53) CARSHELL 66 (16) 62 (22) DYESS _ 48 (21) 4O (19) MARCH 50 (13) 47 (10) a PARENTHETIC VALUES ARE STANDARD DEVIATION. ‘83 TABLE 20. 8-526 ORGANIZATIONAL-LEVEL CORROSION MAINTENANCE, MEAN MONTHLY a MANHOURS PER AIRLANE. 10 1978 - 30 1979. AIRBASE BARKSDALE BLYTHEVILLE CASTLE FAIRCHILD GRIFFISS LORING MATHER ROBINS SEYMOUR-JOHNSON NURTSMITH CUMULATIVE EFFORT 31 54 91 4O 61 58 39 36 43 44 (8) (24) (20) (14) (15) (15) (8) (8) (12) (12) 30 86 81 35 56 55 38 36 39 45 MEAN MONTHLY EFFORT (10) (33) (19) (18) (26)- (24) (9) (6) (19) (17) TABLE 21. B-52H ORGANIZATIONAL-LEVEL CORROSION MAINTENANCE, MEAN MUNTHLY MAQNHOURS PER AIRPLANE, 10 1978-30 1979. AIRBASE CASTLE ELLSWORTH GRAND FORKS K. I. SAHYER CUMULATIVE EFFORT 74 4O 3O 3O a PARENTHETIC VALUES ARE STANDARD (27) (14) (5) (10) DEVIATION. MEAN MONTHLY EFFORT 78 3O 26 24 (25) (14) (6) (11) 84 per airplane range from 40 to more than 100 for the four airbases. In the case of the G-series, they range from 30 to more than 85 manhours per month, and for H -airplanes, from 24 to 78. 5. COMPARISON WITH ENVIRONMENT A preliminary comparison of monthly corrosion maintenance manhours per airplane with environmental corrosion severity indices obtained from PACER LIME is shown in Figures 12-14. Results for D-series airplanes show excellent linear correlation with environmental severity ratings. In the case of G-series airplanes, the agreement is not good, probably because there are unknown factors in the corrosion equation. There are similarities, however: Seymour-Johnson, Robins, and Barksdale are similar both in geography and maintenance, yet Blytheville is anomalous, having a similar environment but a much higher rate of corrosion maintenance. The Castle effort also is high, probably reflecting the training-mission-induced cracking repairs. Griffiss is high in corrosion maintenance, related to delamination problems and possibly corrosion caused by acid rain, also not included in the corrosion severity algorithms. No obvious explanation comes to mind for the high values at Loring, except possibly the acid rain question. The H-series results show good agreement for Ellsworth, Minot, and Grand Forks, which have similar environments. Castle is high, again reflecting training mission damage. K. I. Sawyer may not be indexed properly with its environment, which may be more severe than reflected in the PACER LIME re- sults. MONTHLY MAN-HOURS PER AIRPLANE 80 7O 60 50 '40 "30 20 IO 85 IDA In ." Andersen Guam Carswell TX Dyess TX March CA CORROSION SEVERITY INDEX Figurel2 . B-SZD Organizational-Level Corrosion Maintenance Compared with Environmental Ratings, 1Q78-3Q 79. 86 MONTHLY MAN-HOURS PER AIRPLANE C Castle CA 90 BA Barksdale LA 'BL Blythville AK F Fairchild NA G Griffiss NY .BL L Loring ME L. 80 M Mather CA .C R Robins GA S-J Seymour-Johnson NC N Wurtsmith MI 70 _. 60 - O G I L 50 #- ON I F . R 3° -' ‘ .BA 20 — J I l J J 2 3 4 5 6 CORROSION SEVERITY INDEX Figure 13. B-SZG Organizational-Level Corrosion Maintenance Compared with Environmental Ratings, lQ78-3Q79 MONTHLY MAN-HOURS PER AIRPLANE 8O 7O 60 50 40 3O 20 IO 87 C Castle CA E Ellsworth SD _§F Grand Forks ND KIS K.I. Sawyer MI C C M Miont ND a. “" .KIS .GF' .M I J I I l 2 3 4 5 6 CORROSION SEVERITY INDEX Figure 14. B-SZH Organizational-Level Corrosion Maintenance Compared with Environmental Ratings, lQ78-3Q79. V. MAJOR AIRCRAFT INSPECTION PROGRAMS A. RELIABILITY CENTERED MAINTENANCE Early primitive airplanes with marginal performance capabilities could not utilize system redundancy because of the weight penalty. Thus, maintenance programs necessarily were dedicated to maintaining system reliability because almost every component had a direct relation to safety. Early maintenance philosophies generally were “isochronal". i.e., based on fixed flying-hour intervals. At the end of some predetermined usage period, a given subsystem was inspected and repaired as necessary (IRAN). 0n the other hand, since brand-new, warehoused systems deteriorated even in storage, the concept of ”periodic“ maintenance (i.e., based on fixed calendar interval IRAN) has developed. Both approaches have been used widely as airplane design has advanced. Increasing system redundancy, however, has made possible a basis for airplane maintenance programsjbecause the relation between individual component reliability and safety is less significant. Modern in-service reliability analyses (48) have shown that the real incidence of adverse age/reliability relationships (the basis for presumed effectiveness of preventive maintenance tasks) is much lower than intuition had implied. Therefore, airplane designers developed a new conceptual model for maintenance programs which provides efficient scheduling of maintenance for complex equipment as well as the ongoing management of such a program. These models are referred to as Reliability-Centered Maintenance (RCM) programs because they are focused on maintaining the inherent reliability capabilities of equipment at minimum cost. United Airlines first initiated a generally applicable RCM technique to develop scheduled maintenance programs for new aircraft. An elementary decision ”tree" was developed in 1967 and first formalized in the Air 88 89 Transport Association's publication "Maintenance Steering Group“ (MSG-1) (48). This program became a policy guideline for development of initial maintenance programs for the Boeing 747. Subsequently, it was decided that the experience gained from this technique should be applied to all new aircraft types by deleting certain 747-specific information, resulting in "MSG-2“ (48), which was used to developed the Lockheed L-IDII and McDonnell-Douglas DC-IO initial maintenance schedules. In this program, each aircraft item is defined as a Structural Significant Item (SSI), Functional Significant Item (FSI) or Work Zone Item (HZI) according to their importance to the aircraft's structural integrity, or function, or economic impact. RCM logic defines four basic scheduled maintenance tasks: inspection at apprOpriate intervals, either periodic or isochronal; rework before a specific allowed age; inspection when failure - occurs; or discard before a specific allowed age. These maintenance tasks should reduce the hazards from failure to an acceptable level before the tasks can be considered effective. ' An Air Transport Association (ATA) committee review of the MSG-2 and the experience derived from its use identified several areas which could be improved, e.g., the rigor of the decision logic, the clarity of the distinction between economics and safety, and the adequacy of treatment of hidden functional failures. The revised technique (MSG-3) was built upon the existing framework of MSG-2. The difference between the two appears in presentation of materials and procedural content, with the addition of economic-based decisions. The first task of this technique involves classifying the equipment into three major divisions: power plant, functional system and structure. Each division itself is further subdivided to a point where the failure of a given 90 item is not significant. Once the significant items have been identified, the functions of each must be defined. After these two steps, the analyst uses the decision diagram to identify program tasks. After the types of tasks to support each item have been identified, the interval between successive performance of each task is established. The interval between of most applicable tasks is held conservatively short in the initial program and extended on the basis of analysis of information obtained after the equipment enters service. The process of obtaining and analyzing in-service experience often is referred to as age-exploration. This technique also classifies maintenance tasks to a smaller number of maintenance packages with different time intervals. 1. APPLICATION OF THE MSG-2 TECHNIQUE IN A MILITARY AIRCRAFT There are many differences between military and civilian aircraft, but there are similarities in the objectives of maintenance. These objectives are safety, reliability, preservation, aircraft readiness requirements, and economic considerations. In 1972 the U.S. Navy applied the MSG-2 technique to maintenance planning for S-3 carrier based antisubmarine warfare (ASH) aircraft which was then in development (56). Later in 1972, the Navy directed the Improved Maintenance Program (IMP) based upon MSG-2 for their long-range land-based P-3 ASN aircraft (48). In 1975, USAF applied the MSG-2 technique in 8-52 aircraft. Every item on the aircraft was analyzed and related to one of three basic maintenance processes: hard-time limit (periodic overhaul/replacement), on condition (scheduled inspection/test), or on condition monitor (fix when failed), which are approximately the same as those of the MSG logic described above. The B-52 RCM analyses recognized three basic types of scheduled maintenance: (1) Field maintenance inspections accomplished mostly by SAC military personnel on an isochronal phase (100, 91 200, or 600 flying hours) basis and include critical items that should be looked at often. 2- Programmed Depot Maintenance. These are essentially a periodic major overhaul accomplished on a 48 month basis that include items which are less critical but should be looked at on all aircraft, as well as maintenance actions which require special facilities or skills not available at field level. 3-Analytical Condition Inspection: those items which have not been inspected at field or depot and need inspection on a sampling basis. In the case of corrosion maintenance, the fields have prime responsibility for its inspection. The depot will inspect those corroded areas that were not inspected by the field and the inspection will be according to PDM or ACI statement-of-work. B. ANALYTICAL CONDITION INSPECTION The Analytical Condition Inspection (ACI) Program provides for special inspections over and above normal inspection requirements. In this program, ten airplanes are selected for more careful and thorough inspection in order to find defects which would not be detected or repaired in normal PDM inspection. The results of these programs are used to improve and change various inspections and intervals, especially the PDM package. The airplane sample for ACI is selected on the basis of chronological age, time since last PDM, operational environment, cumulative flying hours, and other factors which suggest that an airplane may have problems not indicated by AFM66-1 data, Material Deficiency Reports, or other sources. Thus the number of aircraft and specific serial numbers to receive ACI are determined by considering aircraft configuration, mission, operational environment, calendar age, and flying hours; Lead the Force (LTF) and Controlled Interval Extension 92 programs (CIE) also are considered as ACI candidates. The number of ACI aircraft is based upon the inventory size and the confidence level desired in the result to indicate the fleet condition. The number of B-52 airplanes selected for ACI for several years has been fixed at ten each for the 79 D, 173 G, and 96 H airplanes, thus, the confidence level will be somewhat higher for D-series and lower for G series fleets. Since enough data are not available to examine the result of the ACI program on all B-52 airplanes, we only concentrate our work on the airplanes which have been in the ACI program during FY 78 and 79. Tables 22-25 show the B-SZG, and H ACI airplanes by serial number and airbase. Figures 15 and 16 show the distribution of quarterly corrosion maintenance manhours of B-SZG PDM airplanes for FY 78 and 79. The star signs (*) indicate those airplanes which underwent the ACI program in that year. Figures 17 and 18 show the quarterly manhours of ACI B-SZG airplanes before and after inspection. These two figures show the variation of manhours before and after ACI. Some of the airplanes had smaller manhours after ACI and some of them vigguggggg, Most of these airplanes were transferred to another airbase and therefore a different environment after the ACI program. Because of these transfers, a different rate and type of corrosion could be seen on those airplanes. In FY 78 the amount of ACI airplane quarterly manhours tended toward the mean value after ACI. Nhereas before ACI, they were covering the smallest and largest amount of manhours. In this year three airplanes had less manhours after ACI and six had more manhours than before inspection. The largest decrease belonged to the airplane which was transferred from Wurtsmith to Fairchild after ACI. Other manhours decrement belonged to the airplanes which were transferred from Griffiss to Mather and to Seymour-Johnson, and Seymour-Johnson to Barksdale after inspection. The two airplanes which were transferred to the same base B-SZG 58-241 59-2566 57-6498 58-170 59-2601 58-165 59-2583 59-2589 57-6492 58-227 Table 22 - ACI PROGRAMS FY-78 SCHEDULE LOCATION SEYMOUR-JOHNSON SEYMOUR-JOHNSON LORING .SEYMOUR-JOHNSON (GRIFFISS GRIFFISS SEYMOUR-JOHNSON MATHER FAIRCHILD ELLSHORTH 93 PDM INTERVAL 36 36 50 38 42 43 41 39 39 45 INPUT pglg 27 OCT 77 10 Nov 77 1 DEC 77 19 DEC 77 7 FEB 78 6 MAR 78 30 MAY 78 7 JUN 78 23 AUG 78 11 SEP 78 OUTPUT Iygg; 31 JAN 78_ 16 FEB 78 7 MAR 78 23 MAR 78 4 MAY 78 31 MAY 78 26 JUN 78 1 SEP 78 21 NOV 78 8 DEC 78 B-52fl 60-022 61-019 61-024 .60-059 61-026 60-011 61-036 61-004 60-008 60-013 MINOT MINOT GRAND GRAND GRAND GRAND GRAND GRAND MINOT MINOT Table 23. LOCATION FORKS FORKS FORKS FORKS FORKS FORKS 94 ACI PROGRAM FY-78 SCHEDULE PDM INTERVAL 37 39 4s 46 46 46 46 47 45 44 INPUT 24.1.5. 17 OCT 77 3 NOV 77 8 DEC 77 6 FED 77 13 MAR 78 17 APR 78 18 MAY 78 7 JUN 78 27 JUL 78 31 AUG 78 OUTPUT DATE 20 JAN 78 7 FEB 78 .14 MAR 78 3 MAY 78 8 JUN 78 13 JUL 78 15 AUG 78 1 SEP 78 25 OCT 78 30 NOV 78 95 TABLE 24. ACI PROGRAM FY 79 SCHEDULE POM INPUT OUTPUT .§;§g§ LOCATION INTERVAL _Jy§g; DATE 59-2581 MATHER 43 5 OCT 78 . 11 JAN 79 57-6509 FAIRCHILD 42 16 OCT 78 19 JAN 79 57-6508 MATHER 42 24 OCT 78 29 JAN 79 59-2590 FAIRCHILD 43 1 Nov 78 6 FEB 79 58-158 FAIRCHILD 42 13 Nov 78 I 15 FEB 79 58-200 -FAIRCHILD 42 20 Nov 78 23 FEB 79 57-6471 FAIRCHILD 42 29 NOV 78 .. 5 MAR 79 57-6503 FAIRCHILD 43 7 DEC 78 ‘ 13 MAR 79 57-6480 FAIRCHILD 42 18 DEC 78 22 MAR 79 58-207 BLYTHEVILLE 42 2 JAN 78 29 MAR 79 L521 61-013 60-001 60-047 61-021- 60-052 61-015 60-012 61-012 60-033 61-018 TABLE 25. LOCATION MINOT WURTSMITH MINOT WURTSMITH GRAND FORKS CASTLE CASTLE CASTLE CASTLE CASTLE ‘ 96 PDM INTERVALS 46 45 47 45 45 45 46 47 47 47 #m 26 11 12 13 13 ACI PROGRAM B-52H FY 79 SCHEDULE INPUT DATE OCT 78 NOV 78 JAN 79 FEB 79 MAR 79 MAY 79 JUN 79 JUL 79 AUG 79 SEP 79 5 6 27 25 12 31 29 2 1 5 JAN FEB MAR MAR JUN JUN AUG OCT NOV DEC OUTPUT DATE 79 79 79 79 79 79 79 79 79 79 Corrosion Manhours Per Calendar Quarter 240 200 180 100 160 ' 120 100 80 97 ""10 2ofiaoso'l'aofl'47oflo oo'lss 98'99l—' Figure 15. Before and After PDM in FY 78. and Line Connects the Means. 10 ZOISL“350607D 80 90 Organizational-Level B-SZG Corrosion Maintenance Starred Values Are ACI Units, Corrosion Manhours Par Calendar Quarter 98 320 300 I f 280 — 260- 240- C 220— 100— *. 100 .— ..... . r ‘* *0 120-' " . 100 5 10 20304050607080 90 95 2 5 10 zoaoooéaomé NJ: Figure616- Organizational-Level‘B-SZG Corrosion Maintenance Before and After PDM in FY 79. Starred Values Are ACI Units, and Line Connects the Means. Corrosion Manhours Per Calendar Quarter 99 320 300 280 200 240 220 200 180 100 140 120 100 00 i / .. ’1 ‘fi ‘ L.”L 1 ,| | | | | | A | | l I ILEL,I | | | | 10 20 304050607080 96 96 10 203040506070 80 Figure 17, Organizational-Level B-SZG Corrosion Maintenance Before and After PDM for ACI Airplanes in FY 78. Corrosion Manhours Per Calendar Quarter TOO 300 200 £00 2‘0 '- 220 — 200 — .180 .— 160-; 140 .— 120 .— 100— 5 10 20 304050007080 90 95 2 5 10 20304050607055 Figure 18. Organizational-Level B-52 Corrosion Maintenance Before and After PDM for ACI Airplanes in FY 79. Corrosion Manhours Per Calendar Quarter 10] ‘40 - . 120— 100- ao-— 60- 40 .- I» I l I I I I I 11 I I I I .1 I l I I I, I I I 510 2030 so 70 90 5 10 2030 60 7.0 90 Figure 19. Organizational-Level B-SZH Corrosion Maintenance Before and After PDM in FY 78. Starred Values Are ACI Units, and Lines Connect Same Airplane. Corrosion Manhours Per Calendar Ouarter 102 300 - 250 I- 200 .— 150 I- 100— ‘ 2 5 10 zoaoaouso'wmé 90|"95 9899 10 20304050007080 90 96 98 Figure 20. Organizational-Level B-SZH Corrosion Maintenance Before andAfter PDM in FY 79. Starred Values Are ACI Units, and Lines Connect Same Airplane. 103 showed different behavior. The airplane from Seymour-Johnson showed less manhours and an airplane from Fairchild showed more manhours after ACI. In FY 79, the manhours distribution of ACI airplanes tended to become wider after inspection. The ACI airplanes were the first ten airplanes transferred to the depot in that year and mostly have lower manhours. In this year only two of the airplanes showed a lower amount of manhours after ACI and the rest of the airplanes showed more repairs. One of these airplanes belonged to Blythevile and returned to the same airbase, and the other belonged to Fairchild, but transferred to Blytheville. B-52H's showed approximately the same pattern of manhours before and after ACI, Figures 19 and 20. In FY 78, one of the airplanes showed a sharp reduction of manhours after ACI after transferring that airplane from Minot to Castle. Five of the airplanes showed higher manhours after ACI and three of them lower manhours. In FY 79, airplanes showed more decrease in their manhours than FY 78. Seven of the airplanes had lower and two of them higher manhours after ACI. He should notice that four of the airplanes were transferred from Castle to other airbases after ACI and one airplane from another airbase to Castle. Therefore airplanes showed a decrease in manhours in the first case and higher manhours in the second case. C. CORROSION STATISTICS A fundamental objective of ACI inspection is the detection of corrosion/ corrosion-cracks in a specific system, but not necessarily the most critically corroded item. Once damage is detected, the corrosion problem would be investigated and, if necessary, an inspection added to all airplanes in the fleet. Corrosion detection in a fleet, Cd, is a function of a cumulative random series of corrosion occurrences and systematic inspections (49). If 104 .more inspections are performed and/or more corrosion is initiated on the specific component, the probability of finding a corroded item increases. The corrosion process itself, once initiated, is a monotonically increasing function of time. It is essential to relate the corrosion detection process to detection capability, which probably can be accomplished by utilizing in-service maintenance data. The random nature of corrosion detection in a fleet is presented by ”typical” detectability characteristics derived from service data (49). This approach also can be used to establish service-demonstrated Damage Talerance Ratings (DTR) which are used to derive minimum required DTR's on a comparative basis (49). In addition to providing a comparative damage tolerance between similar details on various aircraft models, the system provides a simple method of evaluating the effect of varying corrosion rates, inspection methods, improving accessibility, and the use of supplemental inspections. The probability of inspecting an airplane with damage, P1, is a function of the level of inspection, the number of airplanes inspected, the area or component under consideration at the specific level of inspection, the position each inspected airplane holds relative to the operational lifetime, and the cumulative exposure to risk. The probability of detecting damage at a specific threshold, P3, depends on the inspection method used (visual, NDE, etc.) the extend of damage and the mental set of the inspector. The probability of detecting at least one instance of damage at the specified threshold, P2, is the product P1 - P2 or, if corrosion damage at the threshold level has occured on one component of one airplane in the fleet, the probability of detecting it during a typical inspection is P5. For the corrosion propagation characteristics shown in Figure 21, L0 is 105 the threshold of damage detection for a given inspection level and method of inspection; LC is the corrosion damage at which residual strength corresponds to minimum design strength (critical); and T is the calendar time which corrosion will propagate from L0 to Le? Corrosion damage at the time of inspection in necessarily random. A relation between the probability of detection and the extent of corrosion damage can be established from service experience for a large number of components. This data then can be used to integrate the corrosion propagation curves to derive equivalent values for probability of detection P2 during time interval T (see Figures 21-22) (49). If T is the average interval between inspections, the frequency of inspection during damage detection interval T is TIT. Therefore, for a given inspection level and inspection method, the probability that a single corroded component (the first) in the fleet in interval T is pr=1-[1-ps]T/T It is known empirically that when corrosion damage is detected in a specific Component of an inspected airplane (say at A61), additional inspections on other airplanes and/or similar components at another location usually will reveal comparable damage, provided the ACI sample was representative. As we have pointed out frequently, there is not a one-to-one correlation: A specified level of confidence must be known in order to assure that the critical damage level has been determined xlg_the selective inspection. Let dT be the mean time interval between corrosion-damage detection in the fleet for the same component/subsystem (which is a random variable.) If the first damage is detected at T1, the second case should be detected at * This does not take account of L0, the level of corrosion damage when repair can be effected istead of remove/replace,i.e., the optimum damage level for repair. 106 Time L Critical For Fail-Safe Load l Corrosion | L l I \ LD 7 fl . , Damage Detection t T _ Period - Time Threshold of ?//P—. Damage Detection TF- qb —— - LO. 1 '. l - 6 l ' P3 .. EqUTvalent .3 l ' l I L - . 1”“ I6 M“ T 4 Corrosion T 4.1 = PROBABILITY 0F DETECTING CORROSION DAMAGE AT A SPECIFIC THRESHOLD FOR A SPECIFIC INSPECTION LEVEL AND METHOD. Figure 21. Corrosion Damage Detection is Time. 1 107. % . First Corrosion Event T I °.4@ .— °°5§fl7 l-Ie T 4.. 5%., ' In snpection. °--. | Number @-.,.. | /////////////////////// Equivalent Damage Detection Period Second Corrosion Event _-Lr AT I_{ ' (“'o. Time Third Corrosion Event ZAT Time #D- Time _ Figure22 . Sequential Corrosion Events and the Probability of Damage Detection. lO8 T1+AT, and the third at T1+2AT, etc. Thus in the period T, the interval available for detecting first damage, the second, and the third will be T-AT, T-2AT, etc. From this the frequency of fleet inspections during the corrosion period is [T/lehere Y = [1-(n-IIOT/T] n a corrosion occurence number, + = values >zero. Concequently, for a given inspection method and level, the probability of detecting one threshold or critically-damaged corroded component in the time interval T is Pd . 1 - [1- Ps](T/T). or Pd = l'Ll-PrJY . Considering all levels of inspection in a fleet, the cumulative probability of A damage detection is PD-l-[H(1-P1)]. where i is the applicagle inspection level. Although values of PD may be close (0.999 vs. 0.998), the probabilities of ggt_detecting a damaged component, given by (l-Pd), 0.001 vs. 0.002, are widely different, hence the latter probability provides a better basis for comparison. To provide a direct qualitative measure of design and/or maintenance planning actions, an eguivalent number of 50/50 opportunities of detection may be used to define a Damage Tblerance Rating as follows. DTR=log(1-PD)/log 0.5 PD=1-(1/20TR). 109 15 I. 1 l l I 0.6 0.96875 0.908' 0.9999695 Figure 23. Damage Tblerence Rating :3 Cumulative Probability of Damage The required level of DTR will be established by evaanting “acceptable“ Service corrosion using the same system. Therefore, a DTR is a comparative measure of damage detection opportunities in the fleet. The required rating system standards are P1, P2, L1, and AT. The data can be used with corrosion propagation data (kinetics) to derive DTR's. If a derived DTR is found to be unacceptable, compared with previous acceptable service experience, the required level can be obtained by --improving detectability characteristics (accessibility, visibility, more sophisticated inspection methods), --increased surveillance (inspection at a lower, more frequent level, special or supplemental inspections; inclusion of an item in ACI). --extending corrosion propagation intervals 113_material/design/stress level Changes. llO The DTR data can be used with other maintenance requirements to provide a flexible structural inspection program for the Maintenance Requirements Review Board process. The DTR system is a comparative fleet system; therefore the variability of structural maintenance in the fleet should be monitored and controlled by concerted OMS/manufacturer action. A most essential prerequisite to such action is that the manufacturer must be supplied with complete and accurate maintenance data. It is doubtful that this is done currently. Inspection interval escalation reduces DTR. When the DTR level reaches. the minimum level required, maintenance aCtion is necessary. Alternative actions which can be considered are: --Discontinue escalation, --modify the basic maintenance plan for the detail, supplement the basic maintenance plan with 'lead-the-force" inspections (49). 111 D. ACI SAMPLE SIZE The lifetime to corrosion failure of a new structural detail in a fleet of aircraft can be represented by a two-parameter Neibull distribution with characteristic life 8 and o as shape parameter. The characteristic life is the time period at the end of which the greatest concentration of corrosion will occur in a given population, a is a scale parameter which shifts the distribution along the horizontal axis and not the shape of the distribution. A set of objects, exposed to the same corrosive environment, the order of first occurrence of corrosion damage is random. In a more severe environment, first occurrence will appear at shorter time intervals than in milder environments, i.e., the distribution shifts to left or right depending on the severity of the environment. Consequently, supplemental inspections of fleet leader aircraft (i.e., those exposed to severe environments for longest time periods) should give the greatest benefit of damage detection. Requirements for supplemental leader inspections generally will be governed by the additional required level of damage-detectability. This depends directly on the probability of inspecting an aircraft with existing damage, (P1), and the probability that an inspection will detect the damage (P2). Values of P2 depends on the extent of corrosion as well as the inspection method used, and will correspond to the values used during scheduled maintenance with the same inspection method. For the supplemental ACI inspection, P1 is a function of the number of aircraft and the time period of exposure to a corrosive environment, all other factors being equal. The probability of including at least one damaged aircraft in an ACI sample of size n is given by the equation (49) n P" = I-LH R(Ti)]. i=1 where 112 R(Ti) = reliability of an arbitrary detail, e -(11/8)a n = number of sample airplanes i A a airplane number 11 a number of hours completed by the i-th airplane B = characteristic life a = Heibull shape parameter. If it may be assumed that damage has occurred, and the entire fleet is inspected, the value of P" is one. Consequently, the normalized probability that an ACI sample of n airplanes includes at least one or more airplanes with corrosion damage to a specific component is given by: n n7 [1] P1 = Pn/Pn = {1-[n R(r1)]}/{1-[n R(11)]}, T 1-1 1-1 where n is the total number of aircraft in the fleet. If the P1 for each aircraft and P" for the fleet are known, the desired level of P1 is achieved by selecting a sufficient number of aircraft for inspection. Since the highest value of r/B gives the maximum P1 values, inspecting the fleet leader aircraft gives the minimum number of aircraft to be inspected. For some models it may be more convenient to spread the supplemental inspections to some percentage of a larger fleet leader sample. Corresponding values of P1 can be derived by considering the average of a series of random selection of aircraft from the fleet leader sample size. When specific aircraft are selected in the sample, the value of P1 can be determined in a tabulated form for each aircraft. In order to apply this method to the B-52 aircraft, first one must determine the fleet size. The number of aircraft in each airbase varies slightly, hence, it is difficult to observe the effect of airbase environment on each airplane. Therefore, airplanes were selected which were 113 stationed continuously at one airbase for seven or eight calendar quarters, yielding seventy one airplanes. The quarterly corrosion maintenance manhours of each airplane, normalized by mean quarterly manhours of that base, were plotted on Neibull probability paper Fig 24. The data fit a straight line and showed no obvious evidence that they are not represented by a Neibull distribution. Graphical estimates of u and o and all of the related parameters B=1lo and asexp(-u) in the Neibull survivor function S(t) = exp [-(0r)5], can be obtained. The slope of the straight line was approximately 4.5, giving a a 4.5‘1 = .22 . In addition the value of x = .18 correspOnds to the value of y-O and gave the estimate u . .18 . TheSe can be converted into estimates for the Heibull parameters 8 = Ila: 4.5 and as exp(-u) = 0.83. In order to compute the mnnimum number of fleet leader aircraft required to achieve P1 = 0.95 confidence in knowing the Neibull parameter a and B, one calculates n n P" = I-{H R(Tj)} or P" =1-{n exp-(11/B)“} . i=1 i=1 The value of P" became approximately I, therefore, there is no need-to normalize the PI'S. Using Equation [1] the cumulative value of P1 is shown for the fleet leader aircraft in the Table 26. The result shows that eight B-SZG ACI aircraft are needed to achieve a P1 value of 0.95. Secondly, it is possible to determine the minimum number of fleet leader aircraft which must be inspected to achieve the required level of P1 if no airbase is to inspect more than one airplane. The same procedure is used as in the first case, omitting the airplanes which exceed an airbase quota of one airplane. The cumulative value of P1 is shown in Table 27, from which it is seen that eight airplanes must be inspected. These airplanes are from Wurtsmith, Fairchild, Mather, Seymour-Johnson, Barksdale, Griffiss, and Loring. IHALUHLI‘ \I. I \JLICI 99 95 90 80 7O 60 50 4O 30 20 10 C II .I I O 3 I ‘f I I 8 I. I O I O I I O I I I l l l L 11 IL I l J, 1 1.1 11 l l l l 2 3 4567891 2 3 4567891 2 3 4 MAN-HOURS QUARTER Figure 24. Neibull Probability Plot of Corrosion Man-Hours Per Calander Quarter. ThbTe 26. Number 115 Cumulative P1 Va1ues for Fieet Leader Aircraft for Case One. Airbase WURTSMITH FAIRCHILD FAIRCHILD MATHER SEYMOUR-JOHNSON BARKSDALE GRIFFISS GRIFFISS Amount of Quarteriy Manhours 1.85 1.82 1.63 1.54 1.52 1.42 1.40 1.39 reiiability R(N) .6199 .6266 .6545 .6632 .6691 .6845‘ .6862 .6868. P1 .38 .61 .74 .83 .88 .92 .94 .96 Table 27. Cumulative P1 Values for Fleet Leader Aircraft in Case THO. Number 01 Airbase WURTSMITH FAIRCHILD MATHER SEYMOUR-JOHNSON BARKSDALE GRIFFISS LORING BLYTHEVILE 116 Amount of Quarterly Manhours 1.85 1.80 1.54 1.50 1.42 1.4 1.37 1.33 km .6199 .6266 .6632 .6691 .58 .684 .589 .695 VI. ESTIMATING MAINTENANCE INTERVALS Aircraft structures may be corroded at one or more locations, and these corroded areas will deteriorate gradually with time. The relations between deterioration of these areas and environment are not known, nor is there a satisfactory measure of corrosion damage. The only possibility is to define a threshold value for corroded parts or items beyond which performance is considered inadequate. The threshold value X, depending on the nature of the corrosion, might be defined as length of corrosion, depth of corrosion, area of corrosion, or the amount of weight lost or gained. A service lifetime is thus fixed at the time when the parameter crosses the threshold. In this case, it is not the failure law that is given initially, but rather a random process describing the deterioration of the parameter with time. The failure law then is the probability distribution of the times at which that random process first crosses the failure threshold. It will be shown now that, under some reasonable conditions, the reliability can be reduced in this case to a random problem (51). Let x be a corrosion parameter, so that, the deterioration here means an increase in x. We denote the failure threshold by X. The deterioration process then is characterized by a family of distribution functions whose simplest members are P(X, t), the probability that x < X at the time t; and P(x,t;xo, t6), the joint probability that x < X at t and that x < X0 at an earlier time t0< t. Since the item deteriorates with time and will not improve before repair, it is valid to say (50) ‘ P(X.t;X0, to)= P(XO, t& (X g X0: to< t) . In words: If it is known that x < X at some time t, then it is certain that x was below any Xo> X at any earlier time to. This arrangement assures that x can cross any threshold X only once, going from "good“ to "failure" condition. Accordingly, P(X,t) itself is a survival probability up to t: 117 118 F (t) = P(X, t). Since we are considering only the aircraft surface, we may have to deal with a complication. The corrosion condition of a part or an item may not be excellent in the beginning and the part may not be repaired after passing its threshold; in other words, the possibility F (O) < 1, and F (~)>O. In the first instance, the parameter x would have acvanishing probgbility even at the time t a O of being unacceptable; and in the second, it need not have become unacceptable in all cases even at t= a. If this is to be avoided. we must add another assumption, I P(X , O) = 1 , P(X , ~) 8 0. Thus, the failure law fellows directly from the simplest distribution function associated with the deterioration process. A. THE MODEL: COST-UNIT TIME PER OPERATION The failure characteristic of an item as a function of time t, t . 6.....a, is described by its probability density function f(t), the cumulative probability function until an item reaches the threshold X at time t is } P(t) 3 ff(t) d1 , t=0 and the reliability of an item at time t will be R(t) = I-F(t). Selection of time for scheduled removal of an item may create two kinds 0f situations: (1) The item will be removed from the aircraft before it reaches the threshold, carrying a cost for scheduled replacement of C1; and (2) the item will be removed from the aircraft after failure but before reaching the time t, carrying a cost of replacement because of failure of the part passing a threshold C2. The total cost per operation will be R(t) C1+ F(t) ca [2] . } R(T) dT t= 0 119 The minimal cost per unit time of operation will be achieved at the optimal time for scheduled removal, t*, and is obtained by equating the first derivative of Equation [2] to zero. The resulting condition for t* is that 'k [3] Z(t*)‘J)'R(t)d'r - F(t) =i . D where Z(r) is the instantaneous failure rate or hazard function, defined as P(t) Z(t) =.___. 9 R(t) . c2 and D 8 E - 1 . 1 B. CORROSION RELIABILITY FROM NEIBULL STATISTICS: ORGANIZATIONAL- LEVEL MAINTENANCE Aircraft corrosion rates are not related to flying hours, but depend only on "environment”. we have no exact time-based data for each component failure, hence our data could be considered insufficient to establish directly the life distribution with a high degree of confidence. The only available data are the corrosion-related man-hours of each airplane. These man-hours show the effort and amount of corrosion repair of each airplane which is related to the corrosion failure of the components. One may use the Neibull formula as a reliability prediction model for an airplane component. The condition on which the Neibull distribution depends is that the component failure, FE, occurred according to a Poisson distribution, i.e., the probability of exactly i-th component failures is (51) Ll P = u e /i! a where u= t / B is the mean of the distribution of component failures per unit time. 120 The Poisson distribution is valid provided: 1. The equipment operates continuously; 2. a system is not turned off when a component fails; 3. the probability of failure of each component within the system is only a small fraction of the total system probability of failure. Since “corrosion [almost] never sleeps”, these conditions are satisfied. Therefore, the reliability of components should be predicted by the Neibull formula. The probability density function, the cumulative probability function, the reliability function, and the hazard function; respectively, as found in Van Aluen (1966) are (52): B B 1 B [4] f(t)= K t exp (- X t ). [5] ' F(t)= 1- exp (- i t8). 1 B [6] R(t)= exp (- A t ). ' . -1 [7] Z(t)- E t8 . Substituting Eq. [5], [6], [7] into Eq. [3], we obtain for the case of the cost per unit time of operation model, the optimal scheduled removal date, 1*, is obtained when the following equation is fulfilled: * B 3-1 t 1 3 1 B 1 - t* 6 exp(- - T ) dt-l + exp(- - t* ) - - = O . A A A D Analysing the result of the above equation reveals that when 8 approaches one, i.e., when the Neibull distribution is reduced to special, exponential case no scheduled replacement of the item should be performed, since the hazard function has a constant value, Z =.£. The result obtained is that 121 1* a o . However, even if B > 1, but C2 = C1, there is no advantage in performing scheduled replacement of the item since D = 62-1 a v and therefore again t* -+ . . Thus the condition for scheduled replacément is when 8 > 1 and C2 > C1. From the data of B-SZG aircraft we calculated the scheduled (Cl) and unscheduled (CZ) cost for B-SZG airplanes. These costs became approximately the same and B = .67 which is much smaller than one. Therefore it is not relevant to find the optimal scheduled repairment time t* for corrosion repairs. Ne suggest no scheduled corrosion maintenance interval fOr 8-52 airplanes or other aircraft fleets, but thorough corrosion inspection in all maintenance levels. In other words, the corrosion maintenance interval is not as important as detection of corrosion, which can be done at field and depot levels for all levels of inspection. VII. SUMMARY From our results reported at the beginning of our study we conclude that environmental factors and time rather than usage factors, play the major role in the corrosion of alloys and airplane airframe structure. We have verified the accuracy and validity of the PACER LIME Corrosion Severity Indices (CSI) by correlation with corrosion maintenance manhours in each airbase. The CSI would be more accurate if appropriate environmental factors were known and quantified. The field-level corrosion maintenance on 8-52 airplanes was compared from one airbase to another, and it was found that variations can be related to environmental differences in weather and atmospheric pollutants. The extent and nature of such maintenance is similar to that of another large airplane system previously analyzed, the C-14IA, despite significant differences in age, mission, and utilization. It is plausible that all large military airplanes will be similar with respect to the effect of environment on maintenance. Considering the environmental variation from one airbase to another, the statistical principles were applied to the problem of sampling a large airplane fleet for special in-depth discovery inspection. It is shown that a smaller sample of airplanes is sufficient if the selection of airplanes is based on random selection; and large savings would then become possible. 122 123 BIBLIOGRAPHY 1. Summitt R., and F. T. Fink, ”Extending the Lifetime of Operational Systems Through Corrosion Tracking and Prediction,“ MSU ENGR-Bl-OOG, Interim Report, Michigan State University, East Lansing, July 1981. 2. Summitt, R., ”Corrosion Tracking and Prediction for C-141 Aircraft Maintenance Scheduling,“ AFML-TR-78-29, Air Force Materials Laboratory, Hright-Patterson AFB, OH, 1978. 3. Aviation Week and Space Technology, May 10, 1976, pp. 140-144. 4. Summitt, R., and F. T. Fink., “PACER LIME: An Environmental Corrosion Classification System,” AFNAL-TR-BO-4102 Part I, Air Force Nright Aeronatical Laboratories, Wright-Patterson AFB, OH, 1980. 5. Grossman, P. R., “Investigation of Atmospheric ExpoSure Factors that Determine Time-of-Hetness of Outdoor Structures,“ Atmospheric Factors Affecting the Corrosion of Engineering Metals, ASTM STP 646, S. K. Coburn, Ed., American Society for Testing and Materials, pp. 5-16 (1978). 6. Hahin, C.,'Predicting the Metallic Corrosion Cost of Operating and Maintaining Buildings and Utility Systems,“ U.S. Army Construction Engineering Research Laboratory, Champaign, IL, 1978. 7. Gilster, L. 6., ”A Statistical Analysis of Maintenance Cost of Large Aircraft,“ Ph. D. Thesis Harvard University, Cambridge, MA, February, 1968. 8. Howell, L. D., ”A Method for Adjusting Maintenance Forecasts to Account for Planned Aircraft Sortie Lengths," M. S. Thesis, AF Institute of Technology, Nright-Patterson AFB, OH. 9. Harrington, J. 3., and J. Teomy, “Corrosion Prediction in F-4 Aircraft Assigned to the Pacific Air Force,“ M. S. Thesis, AF Institute of Technology, Hright Patterson AFB, OH, 1977. 10. U. S. Department of Air Force, ”Corrosion Prevention and Control“, TO 1-1-2, 1 January 1973. U.S. Government Printing Office, Nashington 1973. 11. Shreir, L. L.,"Corrosion," Vol.1, John Wiley & Sons, Inc., New York, 1963. 12. Sereda, P. J., ” Heather Factors Affecting the Corrosion of Metals,“ “Corrosion in Natural Environments, ASTM STP 558,” American Society for Testing and Materials, pp. 7-22, 1974. 13. Vernon, N.H., Trans. Faraday S29..gz, 265-277 (1931). 14. Uhlig, H. H., “Corrosion and Corrosion Control,“ John Wiley & Sons, Inc New York, 1966. 124 15. Foroulis, Z. A., and M. J. Thubrikar, "The Kinetics of the Breakdown of Preanodized Aluminum by Chloride Ions," J. Electrochem. Soc. 122, 1296 (1975). _ 16. Shultze, H. A. “Corrosion Theory and Practice,“ Laboratory of Metallurgy, Delft University of Technology. The Netherlands. 17. Uhlig, H. H.“The Role of a Critical Minimum Corrosion Rate on Fatigue Damage," in NACE-Z, pp 270-278 (1948). 18. Mutchler, H., “Corrosion of Metals Used in Aircraft,“ Journal of Research of National Bureau of Standards, Research Paper RP1316, (1940). 19. The Dow Chemical Company, unpublished research. 20. Dana, E.S., and H. E. Ford, “A Texbook of Mineralogy," 4th edition, Wiley, New York, p 507. (1932) 21. The Dow Chemical Company, ”Fundamental Study of Natural and Synthetic Films on Magnesium and Its Alloys,“ HADC. Report 54-222, August 1954. 22. Bengough, G. D., and L. Nhitby, Trans. Inst. Chem. Eng. 11, 176-89 (1933). 23. Casey, J. J., and A. R. Maupin, ”Outdoor Storage of Magnesium,” a talk presented to Am. Inst. Mining, Met. and Pet. Eng., New Orleans, February 1957. . ' 24. Bathwell, M. R., ' General Corrosion Behavior," Metal Products Department, The Dow Chemical Company. 25. Schiebold, E., and G. Siebel, Z. Physik.§g, 459 (1931). 26. Bathwell, M. R., ”The Corrosion of Light Metals,“ Electrochem. Soc., NY (1967). 27. Loose, H. S., and H. A. Barbian, “Symposium on Stress-Corrosion Cracking of Metals,“ ASTM and AIMME, page 273-292, (1945). 28. Flores, J. L., ”Microbalance Technique in Atmospheric Corrosion," M. S. Thesis, Michigan State University 1981. 29. Mc Kee, A. B., and R. H. Brown, Corrosion 3, 595 (1945). 30. Godard, H. P. “The Corrosion of Light Metals." Electrochem Soc NY., (1967). 31. Godard, H. P., unpublished work (1962). 32. Liddiard, E. A .G., Corrosion Technology, May 1963, Page 111. 33. Sereda, P. J., ASTM Bull. 228, 53 (1958). 34. Sereda, P. J., Ind. Eng. Chem. 52, 157 (1960). 125 35. Sereda, P. J., ASTM Bull. 238, 61 (1959). 36. Ailor, N. H. Jr., "ASTM Atmospheric Corrosion Testing: 1906 to 1976,“ "Atmospheric Factors Affecting the Corrosion of Engineering Metals, ASTM STP 646,” S. K. Coburn, Ed., American Society for Testing and Materials, 129-151, (1978). 37. ASTM, "Corrosion Test Sites as Measured by Steel and Zinc,” ”Metal Corrosion in the Atmosphere, ASTM STP 435," American Society for Testing and Materials, 360-391, 1968. 38. Ailor, N. H., Jr., "Performance of Aluminum Alloys at Other Test Sites," "Metal Corrosion in the Atmosphere, ASTM STP 435,“ American Society for Testing and Materials, 285-307, (1968). 39. Ailor, H. H., "Seven-year Exposure at Point Reyes, California," “Corrosion in Natural Environments, ASTM STP 558,“ American Society for Testing and Materials, 75-81, 1974. 40. Baboian, R., ”Final Report on the ASTM Study: Atmospheric Galvanic Corrosion of Magnesium Coupled to Other Metals,“ ”Atmospheric Factors Affecting the Corrosion of Engineering Metals, ASTM STP 646, S. K. Coburn, Ed., American Society for Testing and Materials, 17-29, 1978. 41. Scheffer, T. C., “A Climate Index for Estimating Potential for Decay in Hood Structures Above Ground,” Forest Products J. 21, 25-31 (1971). 42. Tomashov, N. 0., “Theory of Corrosion and Protection of Metals,“ translated and edited by Tytell, E. H., Geld, I., and Preiser, H. S., McMillan, New York, 1966. - 43. Rozenfeld, I. L., “Atmospheric Corrosion of Metals,“ B. H. Tytell, transl., National Association for Corrosion Engineers, Houston, TX, 1972. 44. "The Maintenance Data Collection System," TO-OO-ZO-Z, Secretary of the Air Force, Washington, D.C., 1975. 45. "Data Automation Elements and Codes Maintenance, "AF Manual 300-4, Vol. XI, HQ USAF, Washington, D. C., 1975. 46. "USAF Series 8-52 Aircraft Work Unit Code Manual,“ T. 0. 18-528-06, Secretary of the Air Force, Washington, D. C. 1979. 47. Summitt, R., and F. T. Fink, "Extending the Lifetime of Operational Systems through Corrosion Tracking and Prediction,“ AFMAL/MLLN, Air Force Systems Command, Wright-Patterson AFB, OH, 1983. 48. Barnard, R. A., and T. D. Matteson, “Military Aircraft Maintenance -A New Concept, Reliability and Maintainability Symposium, 1975. 49. "Airline/Manufacturer Maintenance Program Planning Document MSG-3," Maintenance Steering Group-3 Task Force, Air Transport Association of America, New York, 1980. 126 50. Drenick, R. F., “Mathematical Aspects of the Reliability Problem,“ Soc. Indust. Appl. Math. 8. No. 1, March, 238 (1960). 51. Horn, R. L., and G. S. Shoup, ”Determination and Use of Failure Patterns,“ The Boeing Company Military Aircraft Systems Division Hichita Branch (1975). 52. Landy 5. P. and M. Aharoni, ' Maintenance Policy of Aircraft According to Multiple Criteria, Int. J. System Sci..§, 1093-1101 (1975). lllll uulllllllm 174 7391 ulllll 3 1293 03 INTI/(Ill