You are on page 1of 8

Prospective Applications for SHM on

Commercial Aircraft
H. KAPOOR, C. BOLLER, K. WORDEN, G. MANSON and A. LUGO

ABSTRACT

There is current agreement between aerospace industry and researchers that the
drive for implementing SHM onboard aircraft should come from the need to
increase availability. Scheduled inspection intervals across an aircraft’s operational
life are identified as potential areas to realise such benefits. However maintenance
intervals are complex and before identifying where SHM can be applied, a series of
challenges arise. Firstly, maintenance tasks to be undertaken in an interval are not
explicitly defined; secondly, an optimum or near-optimum map of the interval
process must be established. The map should outline the sequence of jobs to be
carried out but also encompass key uncertainties that are likely to exist. The initial
challenge is addressed by mining through maintenance procedures and documents
to generate realistic inspection phases and the second challenge is met by using
optimisation methods and uncertainty principles. The paper intends to emphasise
the important issues and methods that should be used to meet the challenges
highlighted. A realistic commercial airliner example has been used, taking the
methods outlined to establish prospective SHM areas that increase aircraft
availability.

INTRODUCTION

Recently there has been a huge shift towards commercial airliners using their
high-priced assets “better” in attempt to reduce costs and remain competitive. A
significant driver that airliners are aiming to reduce is operational costs. A factor
that significantly influences operating costs is aircraft availability [1].
Aircraft availability can be defined as the time when an aircraft is available
for service for the airline operator. The time in service is usually compared to
when the aircraft is unavailable and measured as a percentage. It is considered a
key performance indicator amongst the commercial airline industry.
There are two major technical reasons why aircraft become unavailable.
Unforeseen damage or malfunction of components and systems yield some form
of unscheduled maintenance. Additionally, constant scheduled inspection
intervals also cause an aircraft to become unavailable. These occur at intervals
over an aircraft’s life and are mandated by the authorities for aircraft to remain
airworthy. Though significant costs are incurred when unscheduled actions are
necessary, scheduled inspections form a significant proportion when the aircraft
is out of service.
________________
C Boller Prof. Dr.-Ing., Fraunhofer IZFP, Campus E3.1, 66123 Saarbrucken, Germany,
c.boller@izfp.fraunhofer.de

223
Any new technologies which can save time in scheduled and unscheduled
maintenance would be greatly welcomed by airline operators. References [2, 3]
have stated that significant benefits could be attained through reduction of
inspection time by automation and elimination of access jobs. However applied
locations and the benefits such solutions would bring, are still are fundamental
questions that need to be addressed. An approach should be centred on
identifying time critical items within an inspection interval. Time critical items
are those that have an influence on the total inspection interval time i.e. items
that when altered individually change the entire inspection interval duration.

The paper introduces the reader to the main challenges involved when
attempting to find time critical locations in scheduled inspection intervals. The
nature of maintenance today allows a commercial airliner the flexibility to
initiate schedule inspection intervals suitable to their own operations. The initial
challenge is specifying tailored inspection intervals and how best to perform
inspections with the amount of available resources. The next challenge posed is
how common uncertainties effect an interval’s duration and the time-critical
items throughout the maintenance phase.

To address the challenges, maintenance items are collated, optimised, and


analysed from maintainer’s documents. Supplementary to the approach
prescribed a real mid-range aircraft’s maintenance interval is taken and used to
highlight a time critical component.

FORMATION & OPTIMISATION OF SCHEDULED MAINTENANCE


INTERVALS

Forming a scheduled maintenance interval for analysis would involve producing


a mock up that accurately reflects the real case. The initial effort of forming an
interval would thus be collecting all available inspection elements at a specified
interval as a planner would do at a maintenance facility.

The majority of scheduled inspection items arise from the MPD (Maintenance
Planning Document) and SBs (Service Bulletins). The MPD is a generic aircraft
type document (i.e. Airbus A320) that contains a collection of items to be
inspected for systems and structures. Inspection items come from a variety of
sources such as the MRBR (Maintenance Review Board Report) and VR
(Vendor Recommendations) etc. SBs are maintenance intervals that are initiated
whilst aircraft are in service, both mandatory and optional SBs exist. For the
purposes of this paper SBs are neglected from the analysis.

The MPD of an aircraft contains the necessary information for planning items to
inspection intervals. A description of every item to be inspected is accompanied
with a threshold (the first time the inspection is to be scheduled), an interval
frequency, an estimated duration and special applicability to particular models.

Thresholds and frequencies are initially used by planners to form inspection


intervals. Items to be inspected are assigned intervals either in terms of FH
(Flight hours), FC (Flight cycles) or C (calendar years). Inspections related to
structural fatigue are designated in FC.

224
An aircraft would be scheduled downtime based on operational requirements.
The planner knowing the current aircraft’s scheduled slot and its next available
slot uses the MPD to collect pending items. Items that have to be completed
before the next time the aircraft is scheduled are included in the current interval.
For example, a short-range aircraft may be scheduled a maintenance slot at
approximately 20000 FC, 25000FH, or 6 Years; the next slot would be at 25000
FC, 32500 FH, 8 years. All maintenance scheduled items that lie in between this
inspection interval would be grouped at the 20000 FC, 25000FH, 6 Years
interval. Figure 2 clarifies how maintenance intervals are formed.

Figure 2 how maintenance items are grouped from the MPD into intervals

Various types of methods for packing intervals exist. Conventional block checks
are focussed on grouping all maintenance items in large rigid blocks (A, C, D).
Equalised maintenance, spread inspections over more frequent intervals. Both
philosophies have their advantages and are suited to different operators [4].
Currently many airliners are adopting “letter leading checks”, a segmented
approach that takes advantages of both philosophies. The example interval
tested within the paper used a letter leading check interval.

Once maintenance items are packaged to their intervals, common planning


practices involve manually allocating pending items to available manpower.
Organising tasks to labourers is undertaken manually and finding a minimum
duration time solution becomes highly unlikely with the large number of
permutations available. As this planning process is undertaken manually the
sequence of items allocated to labourers has scope to be time-optimised.
Additionally, the benefit of implementing SHM solutions can only be compared
to a fully time-optimised interval without any technology. Scheduling
maintenance items to related maintainers to minimise interval duration time is a
combinatorial optimisation problem, which has been extensively explored in
other areas of research, particularly manufacturing. A standard optimisation
problem that fits very closely to the maintenance challenge posed is the JSP (Job
Shop Problem) [5, 6]. In summary, the JSP is concerned with allocating
manufacturing operations (drilling, machining etc) to machines such that
specific objectives (minimising waste, minimising makespan etc) are optimised.

There are many similarities between the JSP and the maintenance problem.
Manufacturing operations are comparable to maintenance items, machines are
equivalent to labourers and the primary optimisation objective is to minimise the
interval duration. A variety of methods have been used to solve the JSP in
manufacturing. Conventional local nearest neighbour algorithms have been
adopted but are a computationally expensive choice that does not converge well
to optima for larger sized problems. Simulated Annealing and Tabu Search

225
methods have also been used to solve the JSP; these are more complicated to
implement but perform better. GAs (Genetic Algorithms) have been extensively
used within JSPs. Research has suggested that on many occasions, GAs have
outperformed other approaches on standard benchmark tests. For this reason a
bespoke GA was devised for the example interval. Figure 3 shows similarities
that exist between maintenance and the JSP. Figure 4 demonstrates a typical
output produced, showing the sequence and allocation of tasks to labourers.

Figure 4 showing the similarities between the JSP and interval definition in maintenance

Interval Duration

J12

J22

J11 J13

J21 J23

Figure 5 output produced of a typical maintenance interval with jobs to labourers

226
SIMULATING FOR UNCERTAINTIES AND DETERMINING SHM
POTENTIALS

The output generated above could be used to analyse time-critical locations.


However times provided within the MPD are deterministic estimates. The
estimates are prone to changes due to uncertainties inherent in the maintenance
process. From the experience of maintainers, the main source of uncertainty lies
when problematic areas cause changes in inspection times. It is crucial to
classify such uncertainties, as changes could alter time-critical items and the
interval duration.

Analysing changes in inspection times can be achieved through modelling and


simulation. Attaching a representative range of values to highly likely uncertain
regions within the optimised output and simulating how changes affect time-
critical items would give greater insight than currently exists. Simulating such
changes would be more applicable to discrete event simulation [7]. Arena has
been outlined as a suitable tool as it has been used to simulate models and
improve other processes for aircraft availability [8, 9]. Arena was used to model
and investigate common uncertainties found on the GA optimised schedule
example. Further details of its application have been included in the next
chapter. Figure 6 elucidates how changes in individual duration times can cause
time critical components to alter.

Inspection A Inspection A
L4 (Critical) L4
Inspection B Inspection B
L3 L3
(Critical)
Removal B Uncertainty Effect Removal B
L2 L2 (Critical)
Removal A Removal A
L1 (Critical) L1

t t

Figure 6 uncertainties in jobs causing change in critical components

OPTIMISATION AND SIMULATION OF REAL MAINTEANCE


INTERVAL

To identify time-critical locations, a mid-range aircraft’s MPD was used as an


example. The study was only confined to heavy depot maintenance over the
wing section due to the large number of tasks within the MPD. The wing was
chosen as an interesting area by experienced maintenance staff at Aero Mexico
for the chosen aircraft. The maintenance interval analysed would be initiated at
approximately 20000FC and executed at least once more before the aircraft’s
service goal would be achieved. Structurally related items related to fatigue
cracks were only considered in the study. Due to the symmetry of the wings,
modelling and analysis could be confined to one wing section. This reduced the
computational complexity allowing a more efficient search for the optimal
solution.

227
To package the scheduled inspection intervals, items belonging to the same
leading letter checks were consolidated. Specifically, left wing items belonging
to the 8C leading letter check (heavy depot maintenance) were grouped to form
the analysed interval.

Organising the information for optimisation required further work. An MPD


item describes inspection of a specific location along with the necessary panels
for access. It was found that items in the same zone would require the same
aircraft panels to be removed. The optimum way of carrying out maintenance
would lie if removal of all panels in a zone would be undertaken once, followed
by inspection of all specified items in the zone and reinstallation of the panels
removed. Grouping items to their zones would reduce the search space when
attempting to find an optimal solution.

Each zone could be broken into 3 jobs (access, inspection and removal). Jobs
would be the specific actions which the maintainer must complete in the correct
sequence to complete inspection at the zone.

Another parameter defined was the number of available labourers and


profession for the interval. For the example interval analysed, 4 labourers were
assumed to inspect half the wing, 2 were designated as inspectors to look for
cracks and 2 as mechanics who would be responsible for attaining access. The
assumed numbers of labourers were based on real estimates from Aero Mexico.

To minimise interval duration time, a GA was formulated in Matlab to search


for the optimal allocation and sequence of jobs (removal, inspection,
installation). A permutation representation was used to structure the problem
and convert solutions to schedules. After fine tuning or crossover and mutation
probabilities, the GA converged to a minimal duration time of 99.1 hours. As a
form of validation the optimal solution generated was compared a greedy search
algorithm. The GA outperformed the greedy approach by 25% which converged
to 120.6 hours.

The GA optimised tasks based on estimated times from the MPD; however as
discussed, time-critical items are subject to change when uncertainties are
introduced. The preliminary map determined by the GA was thus transferred
and modelled in Arena to accommodate for uncertainties. The model was built
in a flow chart like fashion, which considered the aircraft as an entity. The entity
would move through a series of process modules representing jobs (removal,
access, inspection at zones). The responsible labourer of a job was specified at
its process block and to ensure the correct sequence was kept, it was assigned a
priority number. Duration time of the interval and individual process blocks
were parameters monitored. To verify the Arena scheme was modelling
correctly, estimated times from the MPD were inputted into the model and
compared to the GA schedule. The interval duration from the model run with the
estimated times was 99.1 hours consistent with the GA findings.

To investigate the effect of major uncertainties, 3 problematic inspection zones


that showed considerable variations in time were pointed out by Aero Mexico.
The range of variation for the highlighted zones could also be approximated.
The Arena model generated was simulated a number of times to recognise
changes in the critical zones.

228
During each model run in Arena, the highlighted uncertain zone times would be
incremented by a proportional amount (alpha) and the effect of to start and end
times would be investigated. As times were gradually increased a change in
some of time-critical items were noticed. All time-critical structural zones over
the entire range of uncertainties investigated would be assessed for their
applicability to SHM technologies.

From the example aircraft studies, a zone that continuously remained a critical
job despite changes from uncertainties was inspection of the wing inspar area
inboard section. The area holds fuel and runs from the wing root just past where
the nacelle is attached. A significant inspection within this area is the rear spar
at the lower surface. The lower surface is in tension in flight and can be subject
to fatigue around fastener locations. It is at the fastener locations of the ribs
trunion fittings and nacelle fittings where inspection is stated. Access requires
defueling and ventilation of the tank and removal of the baffle door that
regulates the fuel to the centre of the wing. To inspect the area, an inspector
must enter a manhole and undertake inspection within the tank manually with
conventional ultrasonic equipment. Figure 8 shows the zonal area highlighted on
the left wing portion along with a schematic of a typical fastener found on the
lower surface of the rear spar as looked on from above. The abbreviations FWD
AFT, and OUTBD specify the forward, aft and outboard directions respectively.

Zone 530
Inspar Area

Rear Spar

Figure 8 shows a critical zone despite changes from uncertainties [10]

229
CONCLUSIONS

The paper addressed the significance of aircraft availability for commercial


airline operators and discusses a novel approach that involves taking available
information from maintenance documents to find time-critical inspections items;
such items are important for two reasons. Delayed critical tasks are likely to
postpone an entire interval; also a reduction in the total interval duration can be
achieved if time critical tasks can be reduced. Therefore application of SHM
solutions to critical items has the possibility of reducing the interval duration
and preventing delays.

Optimisation tools from the JSP in manufacturing research have allowed


minimum interval durations to be realised. Determining critical items can be
undertaken at this stage, but a more realistic conclusion could only be drawn
after considering major uncertainties. The example provided a deeper insight
into the combined optimisation and simulation approach. Only structural items
from a heavy depot interval were used. To find critical jobs, maintenance items
belonging to systems and avionics should also be considered within an interval.

A GA was devised specific to the maintenance problem to find a schedule with


minimum interval duration. Remodelling changes in maintenance job times that
were prone to uncertainties enabled a greater understanding of the dynamics
within the process. Changes of critical jobs were identifiable by using discrete
event simulation software. The combined optimisation and simulation method
highlighted interesting areas. A time critical area unaffected by uncertainties
was the rear spar in the inboard tank over the lower surface. Here, manual
ultrasonic inspections are needed to inspect for cracks, the confined area makes
the inspection difficult and intensive. Identifying such inspection areas and
understanding the impact of SHM from an operational perspective would give a
clearer insight into SHM’s role in aircraft for the airline operator.

REFERENCES
1. Jensen, D. (2007) Special Report: TATEM Europe's Future View on Maintenance Volume,
2. H. Speckmann, J.P.D., Structural Health Monitoring for Airliner, from research to user
requirements, a European View, in CANEUS, Conference on Micro-Nano Technologies. 2004:
Monterey, CA, USA.
3. Boller, C., H. Kapoor, and W. Goh, Procedures for the Assessment of Structural Health
Monitoring Potentials, in 3rd European Workshop on Structural Health Monitoring 2006:
Granada, Spain.
4. Razniewski, M., Maintenance Packaging and Bridging in Boeing Fleet Maintenance Seminar.
2006, Boeing: Seattle, Washington
5. R. Cheng, M.G., Y.Tsujimura, A Tutorial Survey of Job Shop Scheduling Problems Using GA I.
Representation Computers and Industrial Engineering, 1996. 30(4).
6. R. Cheng, M.G., Y.Tsujimura, A Tutorial Survey of Job Shop Scheduling Problems Using GA II.
Hybrid Genetic Search Strategies Computers and Industrial Engineering 1996. 36(2).
7. Ragsdale and T. Cliff, Spreadsheet Modeling and Decision Analysis, 3rd ed. 2002, Ohio: South-
western College Publishing.
8. S.E.Vigus, A Simulation-Based Analysis of the Impact of In-Sourcing A Major Process Element
on the Coast Guard HH-60J Depot Maintenance Process, in Systems and Engineering
Management 2003, Air Force Institute of Technology Wright-Patterson Air Force Base, Ohio p.
67.
9. Shyong and Shenn-Rong, Advanced Planning and Scheduling in the United States Air Force
Depot Level Maintenance. 2002, Air Force Institute of Technology: Ohio.
10. Boeing, Part 4 Ultrasonic, Lower chord of the rear spar, in 767 NTM (Non-destructive test
manual). 2009

230

You might also like