Royal Society Publishing

Fatigue in aerostructures—where structural health monitoring can contribute to a complex subject

Christian Boller, Matthias Buderath

Abstract

An overview of the aircraft design and maintenance process is given with specific emphasis on the fatigue design as well as the phenomenon of the ageing aircraft observed over the life cycle. The different measures taken to guarantee structural integrity along the maintenance process are addressed. The impact of structural health monitoring as a means of possibly revolutionizing the current aircraft structural monitoring and design process is emphasized and comparison is made to jet engines and helicopters, where health monitoring has already found the respective breakthrough.

1. Introduction

Fatigue of aerostructures has been an issue since the early twentieth century. Once accumulation of damage resulting from cyclic loads could be proven to be valid, aircraft were specifically designed such that they would withstand the loads for a defined life without visible cracks. Progress achieved in fracture mechanics has been then taken advantage of in a way such that damage (e.g. cracks) can be allowed to be present in the structure, as long as its propagation can be controlled. This has led to lighter weight design, which is always a major design driver for aircraft but has also required more and scheduled inspection to be done over the aircraft's operational life. The balance between gain through lighter weight versus loss resulting from enhanced inspection effort has still been positive with regard to direct operating cost (DOC). This is roughly speaking the way aerostructures are handled nowadays with respect to their integrity. There is a well-established design and maintenance procedure for all this, which has resulted in codes of practice, procedures and handbooks having been established and improved over decades (Anon. 1995, 2005; MIL-Handbook 5; http://www.esdu.com).

Much was learned from dramatic aircraft accidents that happened in the past such as with the Comet in 1954 (fracture mechanics and damage tolerance), the Aloha Airlines Boeing 737-200 (ageing aircraft) and two older generation Boeing 747 from Japan Airlines in 1985 (insufficient repair) and China Airlines 2002 (delayed maintenance). All of these accidents were related to metallic structures. Major concern regarding composite aerostructures was suddenly raised after the American Airlines flight 586 with an Airbus A300-600 that crashed near New York in November 2001 and where a picture of the broken composite fitting of the fin was shown all around the world. However, the much more detailed investigation showed that this picture was only the result of a variety of unlucky implications resulting from controls and handling of the aircraft, training and possibly partially even personal flying attitudes of the pilot who was going more to the limits than initially assumed in the design. Finally, the helicopter world learned very quickly from different serious accidents in the 1970s that vibration monitoring systems need to be integrated into the gears and rotating shafts. This has resulted in their health and usage monitoring systems (HUMS) which is an integral part of nearly any helicopter sold nowadays.

Despite all these accidents, the major issue for any aircraft operator must still be to decrease its DOC without compromising any safety or reliability issues. The options which exist with regard to the balance mentioned above are either to allow for more controllable damage in the aerostructure and thus hopefully achieve further weight reductions, or to improve and automate inspection with the aim of reducing inspection cost or even try to go for both.

There are different ways on how to achieve this but one of them is definitely to get more information regarding the aircraft's structural behaviour that can then either

  1. go into the analytical procedures that allow calculation of damage accumulation and thus consumed life in a much more appropriate way, or

  2. give much more frequent and thus relevant information on the current damaging stage of the aerostructure compared to the way inspection is done nowadays, and/or

  3. allow monitoring of areas whose damaging behaviour was not possible to be monitored before due to economic terms.

The dynamism in sensor development these days which among others can be observed in terms of miniaturization, performance and price, combined with the remarkable progress achieved in sensor signal processing through mushrooming computation power and advanced algorithms has brought in a new wave of structural technology development that can be termed structural health monitoring (SHM). New and further sensors will allow monitoring of operational loads at various locations on the aircraft in much more detail which will further allow calculation of consumed operational life much more according to the real usage and can train pilots with regard to the implications their way of flying has with regard to the usage of the aircraft. Further to this, there are now more and more sensors emerging that allow monitoring of damage on structures in situ and where information can be retrieved at virtually any time, possibly even on a wireless basis. This definitely can help to avoid a large amount of dismantling and re-assembly, which is normally required to access damage critical areas for simply obtaining the information ‘no damage found’.

Throughout the following, the current process of aerostructures design and maintenance will be described and ways will be discussed on how SHM principles can be ‘organically’ and thus beneficially implemented into a relatively complex maintenance process that has been established over decades.

2. Principles of fatigue design

Fatigue is the effect resulting from a component being repeatedly loaded. It results in striations starting in grains of a metallic material which then nucleate a crack that at a certain size can be detected by means of non-destructive testing. The number of loading cycles at a defined load is the characteristic for a fatigue life. The following gives a brief description on the various elements being involved in fatigue design.

(a) Materials and component data

The basis of all fatigue design is materials and component data are usually described as a fatigue life curve with stress and sometimes also strain plotted over the number of cycles in an at least semi-logarithmic scale. The geometry of the component is characterized by the stress concentration factor Kt, such that data determined with a certain material on a specific component can be transferred to a component of different shape but equal material and Kt at least during the pre-design phase. Materials and component data can be found throughout the literature, in handbooks (http://www.esdu.com; MIL-Handbook 5; Boller & Seeger 1987; Eulitz et al. 1999) or may be obtained from software (http://www.lmsintl.com).

Fatigue life of different but geometrically equal components, made of the same material and loaded under the same loading condition still results in significantly different numbers of fatigue cycles until fracture. A factor of two in fatigue life scatter and possibly even more can be considered as absolutely normal. This scatter mainly results from slight variations in the material properties or due to the fact that material on a microscopic scale cannot be considered as fully homogeneous. To avoid any unwanted failures during the operational life of a component or material, scatter bands have to be determined from the representative number of experimental results obtained which are based on the safety factors imposed on the design. It is these latter fatigue life curves (usually related to a very low rate of failure) which the fatigue design of a component is then based upon.

In excess of fatigue life curves, there are also crack propagation data which characterize the fatigue behaviour of materials and components. Such data are again obtained from fatigue tests performed under constant amplitude loading on specific typed specimens such as pre-cracked flat plates or compact-tension specimens. Crack length is recorded at different numbers of fatigue cycles for tests performed at different stress levels. The introduction of a stress intensity factorEmbedded Imagewhere Y is a correction factor that accounts for geometry of the specimen used; Δσ the stress range; and a the crack length, allows all crack propagation data to fall in one relatively narrow band when plotting the crack propagation rate da/dN versus ΔK. This proves that the crack propagation behaviour can mainly be drawn independent of the stress level applied to the specimen. Stress intensity factors and their related correction factors can again be found in handbooks (Rooke &Cartwright 1976; Murakami 1989) and crack propagation data can be obtained from the literature, handbooks (http://www.esdu.com) and software.

(b) Load spectra

Fatigue life of materials and components such as described in the preceeding paragraphs is usually determined under constant amplitude loading. Components operated in service are, however, usually subject to variable loading. The load spectrum a component will have to anticipate therefore needs to be characterized. This can be done in a way that the time domain strain signal recorded is analysed according to well-defined loading cycle counting procedures on a similar type component. The most recent and possibly now also most widely accepted one is the rainflow cycle counting procedure where each loading cycle can be defined as a closed hysteresis loop along the stress–strain path and the hysteresis can be characterized in terms of stress amplitude and mean stress, respectively (figure 1). Different classes of stress amplitude and mean stress are then defined, which allow determination of the respective distribution of hysteresis loops recorded and thus comparison of different load sequences.

Figure 1

Rainflow hysteresis cycle counting method based on stresses and strains.

Load spectra are usually defined during the design phase of a component. In many cases, they are based on the experience gathered in the past, possibly topped by some additional loading conditions assumed for the new component. Load cases have also been standardized in a variety of cases and specifically in aerospace such as with TWIST (civil aircraft), FALSTAFF (military fighters) and HELIX–FELIX (helicopters).

Loads and load sequences on components or better systems such as an aircraft are usually measured at a very few discrete locations. For the variety of remaining locations being prone to fatigue damage on the aircraft system, load sequences have to be determined along transfer functions, which may be ideally done nowadays when digital models of the aircraft are available.

(c) Fatigue life evaluation

Fatigue life evaluation, especially when it is done numerically requires a load spectrum, a description of the stress concentration for notches of the component—usually described by a stress concentration factor Kt—and a fatigue life curve (S–N curve). The S–N curve may be determined experimentally on notched components with the respective Kt value or determined from fatigue data obtained on un-notched specimens using a notch strain relationship (Seeger et al. 1990, 1991). Fatigue damage is determined for each cycle using the Palmgren-Miner damage accumulation rule, which can be written as:Embedded Imagewhere D, n and N denote the accumulated damage, the number of load cycles considered for the load sequence and the number of cycles the component would endure under the specific load for constant amplitude load sequence and a specific damage criterion (usually either initiation of a crack at defined size or complete fracture). Once D achieves unity, the component is considered to have its expected fatigue life achieved.

(d) Safe life versus damage tolerant design

Fatigue and damage tolerance are dominant factors throughout the airframe design process as well as the following in-service life. These factors can only be adequately satisfied when loading conditions, resulting stress distributions and material properties are known. A question of design philosophy has to be linked to this, which mainly depends on a component's inspectability during service, its repairability or replaceability in case of damage or its criticality towards loss of the aircraft in case of the component's failure. Meeting these criteria or not, very much depends on either applying safe-life or damage-tolerant principles (figure 2).

Figure 2

Airframe fatigue design principles.

Safe life means that the structure is designed such that it is able to withstand a defined fatigue life, normally termed in flight hours, for a given load spectrum without any inspection being required. Once the fatigue life has been achieved and the load spectrum has not been exceeded, the component has to be replaced irrespective of the fact that no damage has occurred so far. If loads have exceeded the loading spectrum, the component has to be reassessed. This can be limited to a numerical evaluation only, resulting in a reduced allowable fatigue life or may even require specific inspection using non-destructive testing procedures.

Damage tolerant design allows for a damage to grow. This may be either achieved such that a crack may grow at any time up to a certain length where it will then be stopped by a crack stopper or the component will have fractured and the loads transferred by that component will be transferred by some other component (multiple load path). Damage tolerance can however also be based on assuming a crack to be available at a badly inspectable location and to determine how much the crack is allowed to grow until it finally reaches a critical stage.

3. The aircraft fatigue design process

(a) The process by itself

Design of complex engineering structures such as aircraft is an iterative process. It starts with the customer expressing the requirements, followed by a design concept. This design concept then needs to be analysed which again may result in modifications to be made with regard to the customer requirements. This process is also often termed as the ‘design wheel’.

Once the conceptual design is available more emphasis is laid on aspects such as aerodynamics, weight or propulsion within the ‘initial layout’. It is only in a next step called ‘revised layout’ where structural design together with landing gear (in the case of aircraft), cost and others becomes relevant. Solid-state structural aspects therefore come in at a relatively late stage in design. The major task with structures is to guarantee that structural integrity and performance is guaranteed all throughout the life of the system designed. This requires a good knowledge of past experience as well as powerful tools to estimate future performance.

A more detailed view into the fatigue design process is given in figure 3. It shows that fatigue analysis has to be done at least in two steps. Along the first step, a fatigue analysis is done which is based on the preliminary loading spectra, being mainly based on the available g-spectrum and read across results from past experience. This at least allows definition of first allowable stress levels for the different components being considered which again allows roughly the appropriate fatigue design shape to be found. It is at this stage where the so-called design allowables are generated. Experimental data required usually come from data generated in the past on either the materials and components considered or if not directly available, at least determined on similar materials. In a limited number of cases, material or component data may be experimentally generated at that stage.

Figure 3

Fatigue analysis process for a fighter airplane.

Along a second step, fatigue analysis is done in further detail. This is when stress analysis of the whole structure and thus detailed design has been done, and thus load transfer functions as well as geometric limitations are known. Material and component data need to be generated experimentally in case sufficient experience has not been gathered up to that point of design.

Once an aircraft has been entirely designed, it is in principal ready to be built. However, to fully understand how the complete aircraft structure performs under close to real loading conditions, this structure has to be assessed on the ground along a major airframe (full scale) fatigue test (MAFT; figure 4). The airframe structure is mounted on a test rig and is loaded repeatedly by different loading cylinders according to the loading combinations and sequences for the duration of the aircraft's qualified life times the security factor, which in the case of the example shown in figure 3 would be 6000 flight hours×3=18 000 flight hours. The aircraft structure is inspected regularly and any cracks or damage monitored is fed back for redesign of the aircraft structure or amendments to the maintenance procedures.

Figure 4

Major airframe fatigue test (MAFT) setup of a fighter airplane.

A chronology of the aircraft structure fatigue verification process is given in figure 5. It shows that loads and fatigue assessment are the result of an iterative process and thus appear several times. It further shows that MAFT is not even finished when the first flight is done and may even not be finished when the first aircraft enters into service with the customer.

Figure 5

Chronology of the aircraft fatigue verification process.

Since the operational life of military aircraft has increased significantly over the past years with 50 years due to become standard and up to 100 years to be discussed for specific cases (e.g. the B-52 bombers) a fatigue assessment becomes appropriate once the aircraft type is close to reaching its mid-operational life. This is mainly required because operational loads resulting from aircraft modifications, change in payloads, flight envelops or flight environment may have significantly changed over that first period of the operational life and may have implications on the aircraft's remaining operational life. In some of the cases, an aircraft structure taken out of service is therefore reassessed in a so-called mid-life update MAFT.

4. The aircraft structure support process

(a) Inspection sequences

Aircraft, civil as well as military, are inspected according to well-prescribed procedures. It starts with a pre- and post-flight visual inspection of the aircraft and ends up in a full dismantling or at least disassembly of major components such that fatigue, corrosion or wear can be determined at any of the damage critical locations. In the latter case, visual inspection is highly supported by visualization aids such as borescopes, dye penetrants and non-destructive testing techniques (NDT) with the latter including techniques such as ultrasonics and eddy current. Inspection intervals are usually defined according to flight hours used and partially also according to days of operation. This is incidentally amended by the pilots' judgements, in case they may have faced a situation damaging the aircraft.

With damage tolerant aircraft, the inspection interval is also defined according to the allowable crack propagation life. Based on an initial crack size that may still not be detectable by conventional means, crack propagation life is calculated up to the point where the crack will become critical. Divided by a safety factor, this then defines the true inspection interval which can be either matched with the overall maintenance sequence or requires a redesign of the component such that the inspection interval becomes longer and the overall maintenance sequence can be matched.

(b) Estimation of crack length in damage tolerant structures

Allowable crack lengths are defined regarding the circumstances of how they can be reliably detected. Examples include complex lapped joints where the crack may have to grow a substantial period of time and thus have to achieve a length of possibly tens of millimetres until it can be reliably detected on the surface. A significantly long crack has therefore to be assumed as an initial crack for determining the allowable crack propagation life. If the initial crack length could be determined at shorter length the allowable crack propagation life would become longer, which would result in longer inspection intervals or, if this is not desirable, in higher allowable stresses which would result in lower weight.

The same phenomenon exists with complex inaccessible structural components. An example for this is frames and stringers behind galleys and toilets. Since the effort is too high to dismantle the galleys and toilets for inspecting the frames' and stringers' integrity, the frames and stringers have to be considered to be fully broken for the damage tolerance analysis and the detectable initial crack is a small crack on the fuselage panel's surface that once being there, propagates at relatively high speed and thus only allows for a relatively short inspection interval. If these stringers or frames were inspectable, a much longer crack propagation life, and thus inspection interval, could be allowed (Schmidt & Schmidt-Brandecker 2001).

(c) Changes in operational conditions

A review of the usage experience of specifically fighter aircraft reveals that a number of features are changing within the lifetime of the aircraft where the following can have specifically an influence on loads:

  1. mission changed or added,

  2. flight envelope expanded,

  3. role equipment changed or added,

  4. increased weight,

  5. configuration changes,

  6. increased engine thrust, and

  7. possibly others.

These changes become obvious when looking at the different variants of a fighter aircraft having been introduced after it has been launched where a sample of such a scenario is shown for the Panavia Tornado in figure 6.

Figure 6

Different configuration standards within the Panavia Tornado fleet.

This enhanced generation of variants has further justified the implementation of load monitoring systems, of which the first generation has been introduced with Panavia Tornados.

Civil aircraft are usually anticipating less variation in their design over the lifetime especially with regard to their operational conditions. The major operational change is the conversion of civil airliners such as the Airbus A300 and A310 or Boeing 727, 747, 757 or 767 from passenger to cargo configurations, or recently also the consideration to convert used Boeing 767 into air refuelling tankers.

(d) Loads monitoring and resulting implications

As mentioned above, operational loads monitoring systems are becoming increasingly popular with fighter aircraft (Krauß 1988; Hunt & Hebden 2001) and have partially also been considered and developed for civil aircraft (Ladda & Meyer 1991). Major tasks and objectives for loads monitoring in fighter aircraft include

  1. in-service measurements,

  2. fatigue life evaluation, and

  3. maintenance planning.

Data recorded include either the time domain strain signal monitored by a strain gauge at a defined location or the time domain signal from a variety of sensors already built in for monitoring flight parameters which are then used to determine the load sequence by use of the aircraft's load transfer functions. Data recorded on the aircraft are downloaded by a ground-loading unit and are processed accordingly. This is mainly done in the way that a rainflow cycle counting procedure is applied to identify the different cycles, which are then accumulated and stored in the respective rainflow matrix that finally allows one to characterize the loading spectrum the aircraft has gone through.

Recording and processing of all these data can accumulate a significant amount of work and it is thus that strategies have been developed which include the following three types of aircraft tracking:

  1. individual aircraft tracking (IAT),

  2. selected aircraft tracking (SAT), and

  3. temporary aircraft tracking (TAT).

IAT is performed on 100% of the aircraft and is based on a limited set of data called the pilot parameter set (PPS). SAT is only performed on 10% of the aircraft and includes recording of the full parameter set (FPS). TAT is performed on less than 10% of the aircraft (sometimes even just 1%) and includes FPS extended by strain measurements at highly specified locations on the aircraft. The principal idea behind this concept is to validate the fatigue consumption calculation with the limited amount of flight data available through the calculation performed by FPS.

The loads data monitored are of multiple values. In a first step, they can be used for comparing the actual load spectrum flown with the design or qualification spectrum such as used along MAFT (figure 7). The data shown in figure 7 indicate that the squadrons have so far neither exceeded their MAFT qualification spectrum nor are they flown as severe as the MAFT spectrum has been configured to be. Furthermore, the squadrons' IAT spectra tend to be slightly more conservative when compared to the more detailed in-service SAT data. With respect to prognostics, the monitored spectra can be used to extrapolate the spectrum up to the incident when the initially defined fatigue life expressed in flight hours has been achieved. Further to that it allows determination in a prognostic sense of how much residual life may still exist in a case where the prescribed number of flight hours has already been achieved but the spectrum the aircraft experienced was less severe when compared to the design spectrum.

Figure 7

Comparison of MAFT load spectrum compared with spectra recorded in different squadrons.

(e) Mid-life updates

Mid-life updates are measures more popular with regard to military aircraft. It stems from the fact that military aircraft are flown over a much longer period when compared to commercial aircraft and that the operational environment as well as technology itself has significantly changed over time. It has, therefore, become increasingly popular with some aircraft to revisit the whole aircraft system regarding the aircraft equipment and design. With Panavia Tornado, this has been done mainly as a result of using this aircraft in many more roles and flight durations than anticipated initially; flying it in and even exporting it to regions of the world possibly never considered initially, and upgrading it with a variety of new avionics, electronics and reconnaissance equipment.

With respect to the aircraft's structure, this has resulted in reperforming a MAFT with a significantly used aircraft taken out of service to mainly determine the residual life as well as the locations the aircraft structure might be prone to fatigue damage in the longer term. Further to this, maintenance plans have been established according to which major components of the safe-life designed aircraft are replaced by new ones. Although these components do not show any cracking, as a result of the well-known scatter in fatigue life they may last much longer than the operational lives achieved, the components have to be removed because they have been designed to be crack free.

(f) Ageing aircraft

The ageing aircraft issue discussed more from an academic point of view for a decade so far became fully apparent through the Aloha Airlines Boeing 737-200 accident in 1988. Up to that time damage (e.g. cracks) was considered to appear at single stress raisers such as rivet holes and crack propagation was assumed to start from a single crack which finally also defined the inspection period (figure 8). What was not considered was the fact that in an aged and thus significantly loaded structure with a variety of notches such as a rivet line, the likelihood of more than one crack generating from each of the different holes is significantly increased when compared to the pristine material. This multi-cracked configuration, which has also been denominated as multi-site damage, leads to a much shorter crack propagation life, which is schematically shown in figure 8 and has to be considered with regard to the inspection intervals whenever operation of an ageing aircraft is considered.

Figure 8

Multi-site damage and consequences.

The conclusion from years of assessing aged aircraft was that aircraft can be inspected according to the traditional procedures up to a defined age but then have to be inspected more carefully. This more detailed and mainly more frequent inspection does not have to take place for the whole aircraft but for specific components and areas of the aircraft which are specifically prone to damage. Figure 9 shows, as an example, some of the areas which have been identified in that regard for the Airbus A300.

Figure 9

Fatigue damage susceptible areas in ageing Airbus A300 aircraft.

An aspect to be mentioned in that regard is the management of the aircraft airworthiness within the ageing aircraft process. Major activities within this include:

  1. direct inspections with regard to fatigue cracks,

  2. direct inspections and sampling programmes to determine the condition of the aircraft structure as well as its equipment, and

  3. prepare corrosion prevention plans (CPP).

The importance of these activities increases significantly with the age of the aircraft, which mainly results in an increase of information handling and processing. The directions issued with regard to detecting fatigue cracks depend greatly on the monitoring techniques being available, while the sampling programmes turn out to become a knowledge base for any experience gathered with the aircraft fleet and hopefully shared among the different aircraft operators as well as the aircraft assembler, maintenance organizations and airworthiness authorities. It is only on the basis of such a knowledge base that effective CPPs as well as repair methods can be configured, certified and finally applied.

(g) Engineering support systems

As a consequence of the various monitoring systems implemented into aircraft nowadays or being a part of the aircraft's integrity certification process, concepts and initiatives have emerged that are intended to handle the data flow on ground as well as the follow-on logistic processes. Derived from the strong SHM initiative around the Eurofighter Typhoon aircraft, analysis and synergy has to be established based on information being generated from the following tools, systems and actions:

  1. aircraft system health (ASH),

  2. structural health monitoring (SHM),

  3. engine health monitoring (EHM),

  4. secondary power system health monitoring (SPS),

  5. logistic software package (EFLog),

  6. non-destructive inspections (NDI),

  7. experience capturing systems (ExCS), and

  8. aircraft integrated systems (AIS).

Information regarding all this is currently downloaded by different means and protocols and it is a major need to get this better standardized not just for one type of aircraft, such as is again proposed for the Eurofighter Typhoon in figure 10, but also across the different aircraft types including fixed and rotary wing aircraft from the civil as well as the military sector.

Figure 10

Engineering support system information process routes for Eurofighter Typhoon.

This major issue is currently trying to be tackled by an EU-funded Integrated Project entitled TATEM along the 6th Framework Programme that started in March 2004 (http://europa.eu.int/comm/research/fp6/projects.cfm). It is only with more standardized processes of the type proposed here that information generated can be better used in a way that follow-on logistic actions such as the provision of spare parts and maintenance personnel can be done effectively and advanced technology such as wireless communication, web-based logistics and troubleshooting and last but not least the integration of further advanced sensing into aircraft structures and systems can be brought further ahead. The latter will be discussed in more detail below.

5. Advanced aircraft structural health monitoring

The way aircraft are monitored today has not very much changed over the past decades (Ladda & Meyer 1991). Most of the monitoring is done by visual inspection supported by other NDT such as ultrasonics and eddy current. Monitoring is done at prescribed intervals which are defined by the weakest links in the aircraft system. As a consequence a huge amount of information is generated, which needs to be processed directly by the maintenance personnel involved. In close to 100% of the cases, the information is always the same: no damage found. Further to this, the major effort in achieving this information is related to dismantling and reassembling the aircraft structure to get access to the component considered. Is this huge amount of effort required to get such little information? Are there no easier means to obtain this information in a more efficient way? In many cases, we even replace the component with no sign of damage only because we do not consider any means to receive more continuous information from these components. In Valeika (2003), this is compared to a situation where all kidneys would have to be replaced for all 55-year-old people because in 0.1% of the cases damage to kidneys has been observed. Why do not we apply such principles for human beings also on aircraft? The answer is quite simple. Human beings have orders of magnitude more sensors than aircraft have. Their sensors are orders of magnitude more primitive than the few sensors we use today in aircraft. Sensors in human beings are highly redundant while the ones in aircraft are mainly not. Biologic sensors can regenerate themselves by self-healing which the ones in engineering cannot. These are just some of the differences between biology and engineering, so the logic questions with respect to a revolution in maintenance technology arising from this are as follows.

  1. Can we integrate sensors into the structure that will give us more efficient information than we have today?

  2. Are there low-cost simple sensors around which can be integrated into structures in high quantities and would provide the sufficient redundancy?

  3. Will it be possible to process the high quantity of information generated?

  4. Will this revolution in maintenance through structure-integrated monitoring be beneficial for the operator without compromising safety?

The answer to all of these four questions is: yes. An explanation on how this can be achieved is given subsequently in conjunction with what can be denominated as advanced aircraft SHM.

SHM is considered today to be the integration of sensors into structural components that allow continuous monitoring of the structure combined with automated advanced signal processing. To keep consistency with established designs in engineering, SHM is based on the engineering design principles applied nowadays and tries to automate and extend the monitoring process to the benefit of the engineering system considered. It uses sensors such as optical fibres, piezoelectric elements, micro-electro-mechanical systems (MEMS) or possibly even nanostructures to just name some of the ones being mentioned most. These sensors allow monitoring of strains, acoustics, electrical fields, temperature, pressure, humidity, chemicals and possibly more. Information is retrieved either by wires but more recently even wireless. Sensor signals are processed using advanced data acquisition cards and multiplexers combined with FFT-analysers, wavelets, genetic algorithms and artificial neural networks to again just mention a few. Further information on what is ongoing can be found in a textbook (Staszewski et al. 2003) as well as in the proceedings of conferences (Balageas 2002; Chang 2003).

(a) What could be monitored?

Since design principles in engineering are very much established and monitoring is just a consequence from all this, the central question with regard to monitoring results in: what are the design parameters which we have to assume in design and which we are thus most lacking, with regard to more light weight and cost-effective design?

All structural designs are based on loads (static as well as cyclic) which we have to assume prior to configuring the structure. These loads do not have to be limited to mechanical loads only. They can also include other environmental loads such as temperature, humidity, chemical corrosives, etc. We then impose a safety factor which has to cover all uncertainties we have with these loads. If we could, however, have more information on when which load occurs (i.e. the real load sequence), we could reduce the original safety factor since safety would be covered by the additional information provided. This would allow us lighter weight design without compromising safety. The means being required here is therefore loads monitoring.

The other phenomenon that needs to be monitored and which is a consequence from our design is damage. Damage needs to be monitored because

  1. operational loads as well as material properties are subject to scatter, which as a consequence can influence the incident of damage initiation as well as the period of damage propagation significantly,

  2. loading of the structure can go beyond design allowables (overloading) either by accident or intentionally with respect to enhancements or life extensions, and

  3. damage is allowed to occur in a controlled way (fail-safe and damage tolerance).

Today, this is solved by large and obviously also costly inspection initiatives, where the cost could be reduced through automation without compromising safety. The means therefore considered here are damage monitoring.

Recent development in sensing and sensor signal processing technology gives rise to what has been and could be done for the enhancement of loads and damage monitoring and is thus summarized in the following.

(b) Loads monitoring

The way loads are currently monitored in aircraft is either by implementing the strain gauges at well-selected locations or by using the flight parameters monitored on the aircraft. In both the cases, either strain gauge or flight parameter based, the information recorded and downloaded on the ground is fed into a digital model, which is mainly the loads model of the aircraft structure on a finite element (FE) basis. This load information can then be used to virtually calculate the damage accumulated at any location of the aircraft structure. The problem with the current loads monitoring systems is, however, that loads are monitored at fewer locations than this should be with respect to the aircraft's complexity. The Eurofighter Typhoon's structural loads are monitored with just 16 strain gauges being implemented over the whole structure of the aircraft. The major driver for that limited number of sensors is restrictions in data storage and processing. However, these decisions were taken in the past with the respective technology of that time. In terms of technology available nowadays, this looks to be far too little sensing and any further improvement loads monitoring may be able to achieve has to be seen in the context of the following.

  1. Clear identification of the areas (e.g. notches, joints, lugs, fittings, etc.) of the structure being prone to damage.

  2. Monitoring of the load sequence in or very close to the areas being prone to damage, not only on a single- but also on a multi-axis basis such that the load sequence provided is similar in quality to a load sequence having just been measured in the notch to be monitored.

The former can be achieved by a clear structural analysis. Current software analysis nowadays allows the mapping of a structure with respect to stress and strain concentrations, as well as damage having been accumulated. The result is some colourful pictures of which an example is shown in figure 11. These pictures allow one to select and decide upon the locations worth monitoring with regard to their load sequence that is then fed into a FE analysis and a follow-on fatigue life calculation.

Figure 11

Stress distribution of a fighter centre fuselage frame.

One of the limitations in using electrical strain gauges for strain monitoring is their relatively high amount of wiring. Two wires are required for each sensor. A much more elegant method in monitoring exists with fibre optic sensors and here specifically with fibre Bragg grating (FBG) sensors. These sensors have the advantage of being light weight, having all passive configurations, low power utilization, immunity to electromagnetic interference, high sensitivity and bandwidth, compatibility with optical data transmission and processing, long lifetimes and low cost (as long as one uses silicon fibres). Their disadvantages mainly appear when being integrated into a material such as a composite where repairability of the sensor is mainly excluded. The overwhelming advantage of FBG sensors is however that they can all be lined up as hundreds and even thousands of sensors along a single optical fibre and each can still be identified due to their different grating pattern and thus be multiplexed. Applicability of this method has been proven in a variety of cases where test cases on real aircraft structures have been described (Trutzel 2000; Betz et al. 2001). Similar results of success are reported with the integration of FBG sensors into a composite hull air cushion catamaran (Johnson et al. 1999). Work is still ongoing with regard to fully integrating such a sensing system into a diagnostic and prognostic system.

FBG sensors have a further advantage that they are able to also monitor temperature as well as pressure. With temperature and pressure profiles being able to be recorded, this allows one to get a broader picture of a structure's loading environment.

Another type of sensor of interest for loads monitoring is MEMS. These are small micro-machined sensors on a silicon basis that allow conversion of thermal, mechanical, magnetic or electrostatic energy into electrical signals and thus allow monitoring of a variety of parameters such as strain, temperature, vibrations, humidity, pressure and even a variety of different type of liquids. So far, MEMS sensors have been specifically developed for corrosion monitoring (Wilson et al. 2001) and are currently evaluated in a longer-term in-service test on a Delta Airlines Boeing 767 (Trego 2003).

(c) Damage monitoring

Damage is monitored by non-destructive means. Conventionally, this requires dismantling of most of the structural system since most of the areas prone to damage are very much hidden and thus difficult to access. It is therefore more dismantling and reassembly which cause the relatively large effort for inspection when compared with the monitoring effort for finding the damage itself with an NDT technique. Sensors fully integrated or adapted to the structure to be monitored that remotely send out the monitoring signal upon request can therefore help minimize the current dis- and reassembly effort required to the situation where damage is truly detected and repair is unavoidable.

To find out where sensors may be useful for integration and where sensors can be avoided, the stress distribution map such as shown in figure 11 and specifically a damage distribution ‘map’ of the structure is required as the first step. Further to this, any recordings from scheduled maintenance may be useful, that allow underlining of the analytical results or extension of the information pool, which is specifically relevant when considering corrosion damage.

Monitoring with structure-integrated sensors can be done on the basis of a variety of different physical parameters. To keep compatibility with state-of-the-art NDT techniques in aeronautics, ultrasonics and eddy current are the most popular techniques.

Ultrasonic and thus acoustic waves can be sent into structures by attaching and/or integrating piezoelectric elements to and/or into a structural component. Acoustic waves are sent out by the piezoelectric element where Lamb waves are possibly one of the most efficient since they operate as guided waves. The reflected and/or transmitted signal can then be again recorded by a piezoelectric element. Systems like this have been made commercially available such as the Smart Layer™ from Acellent Technologies (http://www.acellent.com) where piezoelectric elements are positioned according to structural needs on a Kapton Layer and are electrically wired by copper wiring using PCB techniques for the manufacturing process. A variety of aircraft components have been monitored this way where examples are given in Boller et al. (2001) and Betz et al. (2004).

The acoustic waves emitted do not necessarily have to be sensed by piezoelectric elements. Fibre optic sensors will do it as well and here specifically FBG sensors are catching up (Ihn & Chang 2004). This type of sensor will be specifically considered in areas where electromagnetic interference may be of concern or where other parameters may be useful to be monitored with the same sensor (e.g. strain or temperature). MEMS is another type of sensor that can be used in that context as well. A very fast and efficient way of monitoring is also laser scanning vibrometry (Staszewski et al. 2004). In that case, a laser scans the surface of the component to be monitored with respect to the Lamb wave going through and can thus detect changes in the Lamb wave propagation due to damage. This method is however limited in having the surface of the component accessible, i.e. the effort of dismantling and reassembly may often still prevail.

There are also eddy current sensors and arrays available that can be attached to structures (Washabaugh et al. 2002). These sensors are shaped-field sensors designed as conductive metallic windings which are placed on a carrier such as Kaptone using micro-fabrication techniques. A magnetic field is generated from one of these windings which is then recorded by the other windings. This allows monitoring of crack propagation in metallic structures. The system can also be extended to a high-resolution eddy current imaging system by introducing a single spatial wavelength or periodic, square-wave inductive drive winding with a linear array of inductive sensing elements. Further to this, the wavelength of the magnetic wave can be varied which allows detection of cracks, inclusions and corrosion even in thicker metallic components. The system has been proven to work on cracked aluminium panels, around rivet holes and on a C-130 flight deck chine plate.

There are also the more passive methods being discussed, where acoustic emission is possibly the most widely explored (http://www.ultra-scs.com; Saniger & Dupuis 2002). This method, which ‘listens’ to the noise being generated through a crack by monitoring the respective stress waves being generated allows one to use the piezoelectric elements mentioned before. However, these stress waves are only generated when either high loads are applied to the structure, which is relatively seldom in the case of the randomly loaded structures of an aircraft, or the structure may be significantly aged such that the condition of ageing may turn out to be critical. Further to this, the system must be ‘alert’ at any time during in-service so to not miss any of these seldom events. ‘Black outs’ are therefore not allowed. Finally, suitable filtering techniques have to be applied such that the signal resulting from damage can be clearly identified from any other noisy signals being around. Independent of all this, it may be worth observing the research effort generated in that field.

(d) Systems monitoring

Health monitoring in engineering is not limited to structures only. It also includes full systems and it may be specifically this area where everything regarding SHM started. Monitoring of gears is one of the big initial areas where vibrations are monitored and characterized on the basis of frequency spectra and where changes in these spectra can be related to specific types of damage.

Jet engines have traditionally been monitored for a long time and engine condition monitoring (ECM) is nowadays standard in jet engines. Systems used in that regard include the full authority digital control (FADEC), the remote data concentrator (RDC), the engine's spool speed, engine distress monitoring system (EDMS) and the ingested debris monitoring system (IDMS) (figure 12). With respect to ECM one of the major philosophies is to take data from control, which are then fed into the thermodynamic model of the engine and allows parameters related to engine, compressor and turbine efficiency to be determined. These parameters are observed over time, which allow trends to be recognized. However, trends based on a single or very limited number of sensors often turn out to be not sufficient which is why pattern recognition combined with neural networks has turned out to be of great value. In that case, data from FADEC, RDC and the engine's spool speed have been combined. EDMS and IDMS are sensors that monitor debris in real time. IDMS monitors the quality of air coming into the engine, which is an important information regarding environmental loads the engine has to withstand. It allows determination of which deterioration the engine must have suffered. EDMS monitors the quality of air coming out of the engine with regard to debris and allows determination of whether these debris are resulting from the air the engine is operating in or from any deterioration inside the engine, such as wear of abradable seals and coatings, high levels or carbon being produced, debris-producing faults in the gas path, such as blade rubs, combustor faults, seal break-up or others. It may be interesting to report that this debris monitoring information allows a better prediction of the engine's condition compared to most of the vibration or strain sequence monitoring. Owing to the extreme operational conditions, structure-integrated non-destructive monitoring techniques can be mainly excluded. More detailed information regarding engine monitoring can be found among others in Fisher (1997).

Figure 12

Different sources of information for gas turbine monitoring (Fisher 1997).

For helicopters, HUMS have been developed and introduced for around three decades now and HUMS has become a standard equipment in any newly delivered large helicopter. The HUMS is mainly based on monitoring vibrations generated in components such as gears, shafts and couplings, bearings, rotors, and any further components critical for the flight performance. Data are continuously recorded using accelerometers based on piezoelectrics as described before, processed and compared to a threshold value that describes the allowable damage. Figure 13 shows such an example where early damage could be clearly identified. All data recorded are summarized in a database which allows one to gather experience and to observe specific trends that can further help to improve HUMS.

Figure 13

Monitoring of a helicopter-bull gear pinion shown as fourth statistical moment (kurtosis) over time (Larder 1999).

The need for HUMS in helicopters and getting HUMS implemented into helicopters already straight away has also driven the idea of an integrated system. Figure 14 shows the concept with regard to what equipment is required onboard. As well as accelerometers a main processor unit, a RDC, a cockpit display unit and a data transfer unit are required, where some of the units may already be on the helicopter for other purposes. Data are then downloaded on the ground using either a PCMCIA card, the internet or any other type of networking standards which then allow one to further process the data in the ground based system and to store the data in the database.

Figure 14

HUMS equipment proposed for UH-60 (courtesy of BF Goodrich; Watson 2000).

An issue not to forget is all activity related to regulatory bodies. All components integrated in the aircraft need to get airworthiness approval and specifically if they are due to operate during flight. Regulatory authorities can be of great help for implementing advanced technologies, especially if the benefits and needs are obvious, which is definitely the case when reliability of the system such as a helicopter is improved. The regulation authorities are therefore continuously working on making HUMS mandatory in helicopters. Further details regarding this can be found at http://aar400.tc.faa.gov/Programs/agingaircraft/rotorcraft/index.htm.

6. Quantified benefits of structural health monitoring and how to analyse them

Although SHM is still in the area of research to a large extent, benefits to be expected are most likely. This has already been proven to some degree for jet engines where the integration of sensing has already led to extended twin operations safety for an aircraft to operate on a single engine beyond 180 min or the extension of intervals between overhauls. With respect to helicopters, the UK Civil Aviation Authority analysed 63 airworthiness related issues from North Sea helicopter operators (McColl 1997), of which 70% were successfully detected by HUMS. Six of the issues that had been successfully picked up by HUMS were categorized as potentially catastrophic and hazardous, and estimated that one or two of them may have ended up in accidental failure had the HUMS system not been available.

In Staszewski et al. (2003), an example is given as to how far structure integrated damage monitoring can be beneficial with respect to reducing the inspection cost. The conclusion has been that even though such damage monitoring systems are currently still mainly available as a hand-made solution only, they already significantly pay-off in areas where significant effort in dismantling and reassembly is required before the component to be monitored can be reached. Benefits with respect to allowable operational stress of the order of 20% or more are also predicted with respect to using SHM for extending the idea of damage tolerance design (Schmidt & Schmidt-Brandecker 2001).

In addition to automating the inspection process, there is 20% weight reduction expected for specific components through SHM by expanding the damage tolerance principle (Schmidt & Schmidt-Brandecker 2001). This is currently under evaluation and if this number can be proven to be true, SHM will have a significant impact on any future aircraft design.

Generally, analysis of each aircraft type can be advised with respect to the locations prone to damage, the effort related to dismantling and reassembly, the maintenance process itself and the implications SHM would have in automating the inspection process. This evaluation procedure is currently done with too little effort or even not at all. However, going through that assessment would much more clearly show where the potentials are and where the engineering community developing these SHM systems should truly focus on.

(a) Technological threats and challenges

SHM-related technological development described above is far from being at the limits. There is a variety of further materials and signal processing threats along the technology development road. However, as long as benefits of SHM have not been clearly evaluated and communicated, pursuing these threats may not be much worth doing. Much consideration will, however, be required to get the sensing and actuation devices attached onto or integrated into the structural material in an appropriate, reliable and cost-effective way. Further threats can be seen with respect to providing energy to the respective elements. This will become specifically important if wireless communication is considered. Will batteries be required or can the system charge itself through energy harvesting from the structure? Wireless technology will play another important factor with increasing numbers of sensors and this becomes more and more relevant when size and cost of the sensors decrease. Such technology will further promote development of smart sensing coatings that can be easily integrated into or attached onto the structures to be monitored. With nanotechnologies emerging, the first candidate materials already seem to be on the horizon.

7. Conclusions

Maintenance of aircraft is a complex process that has consolidated over the past decades significantly. This complexity combined with all the safety and reliability issues related to it have made this process difficult to modify. However, this complexity should not prevent one from continuously questioning the different steps performed with respect to advanced technology being provided.

The variety of activities mentioned in this article with regard to designing, monitoring and managing fatigue or deterioration in general of aircraft and specifically their structures show that structural integrity combined with ageing issues is a major concern. Despite this significant activity, the process is unfortunately not complete, mainly due to the fact that the interface between fatigue monitoring systems, the subsequent management of aircraft life limitations and repair limits is still missing. A broader thinking in terms of life-cycle cost has become highly important.

Advanced sensing, materials and data processing technology are possibly the biggest challenges to be seen in revolutionizing the inspection process and in automating the procedure leading to the most common information of ‘no failure found’. In case this automation can be achieved then a further, and possibly much bigger, potential can be seen in the extension of the damage tolerance principle such that significant components of the airframe structure can achieve remarkable weight savings. Further pursuing this way will allow new materials, manufacturing and repair techniques to be easier certified and more efficiently applied. There is therefore big potential still available waiting to be explored.

Footnotes

  • One contribution of 15 to a Theme Issue ‘Structural health monitoring’.

    References

    View Abstract