The Planning of a monitoring system should be embedded
in an overall concept which considers beside technical and logistic
aspects the principles of the aircraft and engine maintenance
concept. The split of the entire system into onboard equipment
and local or central ground stations should be agreed in an early
design phase. All kinds of changes (engine modifications, variation
in operating conditions and mission profiles, software updates
and processor hardware development) which are likely to become
necessary over a period in service of more than 30 years should
be envisaged. Adequate means for configuration control as well
as the required compatibility and interchangeability should be
provided.
Feedback from service now allows to quantify the
advantages gained with an onboard life usage monitoring
system in terms of spare part savings and flight safety increase.
Furthermore the scatter in life consumption has been analysed
and it can be concluded that the often heard opinion that life
consumption is directly correlated with the mission type is not
supported by the observations. Of course, it is possible to indentify
particular parameter constellations which occur under normal daily
engine operation but cause either excessively high or very low
life consumption. However, most of the scatter must be accepted
to appear randomly.
In a couple of projects 1, 2, 3, 4 - which
include jet and helicopter engines - MTU have gained experience
in the design and operation of onboard engine life usage
monitoring systems. Based on the difficulties overcome during
the development and introduction phase of these projects as well
as based on some feedback from first years in service a lot of
information has been gathered which allows for the deduction of
some general rules.
Engine Monitoring - and particularly engine life
usage monitoring - is not a stand alone task. It must fit into
an overall maintenance concept which includes aircraft maintenance
as well as engine maintenance. Clearly, the overall requirements
must satisfy the needs of the customer. But the customer himself
needs the support of aircraft manufacturer, engine manufacturer
and electronic system specialists to get detailed insight into
the design target of the monitoring system and into all the functions
of the entire system. Figure 1
illustrates the mutual dependencies.
Generally, an aero engine monitoring system must
be understood as a distributed system, what means that some parts
of the system are located within the aircraft and engines and
others on ground. For most of the components it is a priori fixed
where they must be located. But for other components the decision
depends on a number of criteria. This is particularly true for
the algorithmic functions, where it must be considered when and
for what purpose which information (i.e. raw data, preprocessed
data, results) is needed. It is essential that all these parts
cooperate smoothly. This requires regular exchange of data between
them, where between most of them the data transfer is not a oneway
route, meaning that bidirectional communication is absolutely
necessary (Figure 2).
It follows that the complete functionality of the respective
interfaces should be clearly defined in an early stage of the
project.
This requires an intensive dialogue between all the
parties involved. One of the main problems in the development
phase is a communication problem, since more parties (i.e. companies,
institutions, authorities) and more people (specialists of all
the interdisciplinary subjects necessary) are involved than originally
thought. There are examples where the lack of communication has
led to uncoordinated design and development of engine related
onboard parts, aircraft related hardware and software and
ground equipment. This in turn compromises the functionality of
the whole system and finally yields in a very low level of system
acceptance at the customer's side.
The introduction of an engine life usage monitoring
system provides the user with new and advanced capabilities what
in turn requires adequate adaptation of the overall engine maintenance
concept. This forces him to think in new terms. Generally, the
engine operators were used to count engine life consumption in
(engine flying) hours or (engine flight) cycles, where an engine
flying hour just meant an hour of engine flight or an engine flight
cycle simply one flight. Life usage monitoring systems (which
have been employed since life consumption is not proportional
to the number of flying hours or flights) measure the consumed
life in damage related physical or technical units. For example,
if life consumption due to fatigue damage (based on a crack initiation
lifing concept) is monitored, the unit will be 'Reference Cycles'
(Figure 3).
It should be noted that this type of cycles is completely different
from the above flight cycle. The reference cycle is a stressing
cycle describing the local stress range at a considered critical
area under engine design conditions. Since an engine is normally
not (in fact: never) operated exactly to the assumed design conditions,
the real life consumption deviates from 'one cycle per flight'.
On the other hand, the released lives of critical
parts or components (what includes parts or components which are
worth being life monitored) are determined in reference cycles.
To overcome the problem with the different units, one could simply
measure the consumed life relatively as a percentage of the released
life, but this will provoke additional difficulties when the released
life of a component will change during the engine life time.
Experience shows that it is necessary to change the
released lives when
An improved lifing concept for a part with life limitation
due to fatigue could mean the introduction of a damage tolerant
concept. In contrast to the crack initiation lifing concept, the
damage tolerant concept allows for the presence of a crack in
the considered part. Now the size of this crack (normally measured
in mm or mm) could be used as a quantitative measure for life
consumption. Nevertheless, the actual size of the crack is not
known until detected by inspections. To be on the safe side, the
prediction of the possible crack size has to consider worst case
conditions, meaning that the predicted crack size is significantly
larger than the actual one.
The fracture mechanics crack growth prediction methods
are capable of providing correlations between the number of applied
loading cycles and the resulting crack size. The crack size itself
is of no practical use for the customer (unless observed by inspections).
But based on the above correlation between the accumulated number
of cycles and the corresponding crack size, the monitoring system
can internally handle the respective fracture mechanics algorithms
and convert the damage into numbers of reference cycles for external
communication. Thus, for the application of both the lifing concepts
the customer needs only one unit for the measure of life consumption.
An additional advantage of this procedure is that
internally both lifing concepts can be combined without changes
at the interface and without any appearence outside the algorithmic
part of the monitoring system.
Typically, the creep damage - technically represented
by changes in strain - should be indicated in units of creep life
consumption during a predefined test run at high temperatures.
Other damage mechanisms can be treated in a similar way, allowing for a unique appearance of all life monitoring results at the customer's interface.
Fatigue and creep life consumption - the most important
life limiting damage mechanisms in monitored aero engine parts
- generally result from stress - temperature sequences. Affected
are only a few locations of a part, the so called critical areas.
Monitoring the life consumption of these critical areas simply
means to determine the complete history (i.e. from the virgin
part to the definite life expiration) of stress (or some other
relevant stressing parameter) and temperature development at this
area, and to assess the resulting damage.
Details of the applied procedures were already presented
at several occasions 1, 2, 5, 6. Here only a short
summary is given.
The core tasks for life usage monitoring are
(Figure 4)
These core tasks for life usage monitoring are on
principle the same as for engine design and for the analysis of
mechanical integrity. For the monitoring purpose, of course, special
algorithms have been developed which are fast enough to allow
for onboard real time calculation. Furthermore, the monitoring
algorithms encompass not the complete engine structure but only
the critical components and the monitored areas. To achieve the
requested processing speed for monitoring, some reduction in the
accuracy (compared to the engine design analysis) must be accepted.
The achieveable accuracy was already reported about 5
.
These core tasks alone do not suffice for engine
monitoring. Additional tasks as input data acquisition and conditioning,
the detection of a start criterion (i.e. when the engine is started
and the calculations shall begin) and an end criterion (i.e. when
the engine is shut down and the calculations must finish) are
necessary (Figure 5).
In order to cope with the 'imperfection of the real
world' some checks of data are required. The input signals are
checked for plausibility. In case of implausible signals (e.g.
due to sensor failure) substitute values are estimated. Smoothing
of the signals can be achieved by appropriate filtering.
The monitoring results - i.e. the calculated life
consumption per flight - also needs to be checked. In case of
implausibility again the results are restored by substitute values.
Not necessary to mention that the substitutes are determined such
that always conservative results (i.e. overestimation of life
consumption) are provided.
The life usage monitoring results are evaluated for
logistic purposes. This means that life consumption per part,
per engine or per fleet is continuously observed, trends estimated
and the times for life expiration forecasted. Based on these forecasts
maintenance actions are planned and spare parts purchased.
The main parts involved in an engine life usage monitoring
system are
During the life-time of an aero engine - this is
usually more than 30 years - all these parts necessarily undergo
modifications. The engine itself is modified to improve for shortages
or as reaction to more comprehensive knowledge of its behaviour.
New variants of this engine type are introduced to fulfil customer's
requirements for changed mission profiles, higher thrust, better
performance, lower fuel consumption, extended life, etc.
Most of these modifications are likely to influence
the thermal and mechanical behaviour of the engine, the geometry
of engine components and subsequently the temperature, stress
and life reactions. Therefore, it is clear that for most of the
major modifications an adaptation of the life usage monitoring
algorithms becomes necessary. It occurs that after some years
in service several variants of an engine type are operated in
one fleet in parallel. The monitoring system must be able to cope
with all the different standards. Experience shows that - to allow
for unrestricted interchangeability - it is useful to implement
fleetwide unique software capable of calculating the life consumption
of all engine configurations. It is self-evident that the monitoring
system - and hence the implemented software - must be informed
about the configuration of the installed engines. Generally, this
means that for each major engine modification new algorithms must
be developed and a new software version released. Now it may happen
that different software versions together with different engine
configurations are operated in parallel. To ensure working of
the whole monitoring system, the compatibility of software and
engine must be checked and in case of mismatch appropriate measures
taken. It could occur that between two flights either one of the
engines or the software in the monitoring system was changed.
Therefore, it is recommended to perform the compatibility checks
before the beginning of every engine run.
Changes of the electronic hardware of the monitoring
system are also very likely - particularly, if one reflects on
the time scales of ongoing advance in microprocessor and memory
technology. Extension of the hardware capabilities may also result
from the above requirements. Thus, different hardware constellations
should be considered and, consequently, the monitoring system
hardware configuration should be checked for compatibility as
well.
Based upon these considerations, it can be concluded
that the basic concept for a life usage monitoring system should
contain the necessary means for configuration checks and configuration
control. Software and hardware design should allow for easy adaptation
to the foreseeable modifications.
Almost all of the above summarized monitoring tasks
could be performed either on board or on ground. As expected,
each of these solutions has got advantages and disadvantages.
Performing all the tasks on ground means that a continuous stream
of input data must be transferred to a ground station. Such a
procedure may include a lot of implications. The other extreme
would be to perform all the tasks - including the logistic planning
- onboard. This also does not seem to be the best solution.
Trying a trade off between the efforts for all the
single tasks - including considerations about required processing
capacity, data transfer capabilities and requested time for result
availability - MTU came to the conclusion that the best share
of tasks would be to satisfy the following principles.
The principles outlined above seem to provide the
best arrangement for the whole system. The advantages are that
A disadvantage might be that the entire complexity
of the whole system must be born in mind from the very beginning
of the whole project. It does not seem to make much sense to develop
one part of the system (e.g. the onboard hardware and software)
independently of the other parts.
The system OLMOS (Onboard Life Usage
Monitoring System) for the RB 199 engine in the TORNADO aircraft
was introduced in 1987. Now, after a couple of years in service,
some feedback and operational experience is available. First experience
has been already reported about 7 . The benefits quantified
in terms of reduction in spare part costs and increase in flight
safety have also been issued 7 . It can be summarized
that individual life usage monitoring allows nearly double the
time a critical engine part can be kept in operational service
(compared to life consumption counting based on engine flight
hours, Figure 6).
Additionally, individual monitoring reduces the risk
of unintended use of a part beyond its released life to zero
(unintended excessive life consumption can not be prevented with
flying time based life consumption counting).
First of all we can have a look at the statistical
distribution of life consumption per flying hour
(Figure 7). A dimensionless
depiction has been chosen to allow for the inclusion of all monitored
areas of all investigated engines. Life consumption is shown relatively
to the mean value. It is to be noted that the maximum of the distribution
curve appears slightly below the average value. The curve shows
a wide scatter (about one order of magnitude). Since the data
describe the typical operational life consumption of a whole fleet,
it becomes quite clear that a measure of life consumption just
in flying hours is not adequate.
The question may be raised whether or not the variance
is somewhat systematic in nature or just random.
It was observed for several aircraft that some monitored
critical areas of both the engines in one aircraft undergo significantly
different life consumption although - obviously - both engines
are flown to the same mission pattern. Detailed investigations
revealed that the only remarkable difference in the engine operation
was that generally the right hand engine was started first (about
10 minutes prior than the left hand engine) and also shut down
first (about 4 minutes before the other one). It was assumed that
the resulting differences in warming up and cooling down times
(with the engines in idle) could be the reason.
In a systematic analysis,
calculations with identical
mission profiles but varied periods between engine start and take
off as well as landing and engine shut down were carried out which
could confirm this assumption. As an example, the variation of
life consumption with varying warming up and cooling down times
is shown in Figure 8 and
Figure 9 for
two typical critical areas. Generally
is observed that shorter times between engine start and take off
or between landing and engine shut down - although desired in
order to reduce fuel consumption - have a detrimental effect on
life usage. It is evident that more than a factor of two in life
consumption can be caused.
Based upon this investigation, recommendations for
optimal warming up and cooling down times could be given. Since
the temperature distribution at take off - which is obviously
significant for life consumption - also depends on the temperature
distribution at engine start, an additional investigation was
performed. It could be shown that life consumption of an engine
started under 'warm' conditions (in the investigated example at
temperatures which correspond to a period of 30 minutes between
shut down and engine restart) was up to 30% lower than of a 'cold'
started engine. From this reported example it was concluded that
an algorithm to calculate the temperature distribution at engine
start could be very useful to increase the overall accuracy of
life usage monitoring systems, particularly if the engines are
switched off between two successive flights only for short periods.
Details of the respective procedures have been already presented
4 .
Another interesting observation was that life consumption
of critical areas at the low pressure shaft were considerably
different for aircraft from air bases in the north and in the
south of Germany. Life consumption at these areas was for the
engines from the north nearly twice as high as for engines from
the south. Stress at these critical areas is governed by the shaft
torque which in turn is determined by rotational speed of the
shaft and the air density at the intake. It could be shown that
the different altitudes of the air bases - nearly sea level in
the north and about 500 m in the south - led to different
air densities at take off conditions, subsequently to different
maximum torques in the shaft and eventually different life consumption.
The predicted difference in life consumption came fairly close
to the observed one.
The distribution of life consumption shows significant
scatter as illustrated in Fig. 7. Some reasons for these differences
could be already identified (see section 7.). Combining these
facts, it is possible to construct constellations which cause
either excessively high or very low life consumption. These constellations
are not so extreme that they cannot appear under normal daily
engine operation, and - what seems very important - are not related
to particular mission types. Of course, there are mission types
known which produce very high or significantly low life consumption,
but these are relatively rare and the differences disappear as
consequence of the accumulation effect. The above discussed influences,
in contrast, are systematic and therefore not balanced but amplified
by the accumulating process. Generally, it seems that the differences
in life consumption are better correlated to individual engines
than to mission types or different squadrons.
It must be concluded that the mission type - or at
least the mission type alone - is not the dominant factor in engine
life consumption. Based on the current experience, most of the
scatter must be accepted as random occurrence.
Currently, life usage monitoring results are used
for maintenance and logistic purposes only. The systems are designed
such that the information about life consumption is available
to the maintenance staff immediately after the flight. Since life
expiration does not directly influence flight safety (as long
as not exaggerated) and neither does require any pilot's action,
it has been decided not to inform the pilot about the lifing status
during flight. This could mean additional workload to the pilot
without any beneficial effect.
Nevertheless, the situation can change. So, one should
consider to provide the pilot with some hints which engine manoeuver
in the current situation would be beneficial or detrimental for
life consumption. Furthermore, some interaction between engine
monitoring and engine control could be discussed. This type of
interaction could mean that the engine control system could allow
only for manoeuvering in a lifesaving manner. However, one
should be very, very careful when introducing such interactions,
because flight safety and operational needs must be always the
predominant requirements. Possible implications to the development
of the monitoring system should be also born in mind. Currently,
monitoring systems are developed to a risk class not specified
as 'single mission critical'. But if interaction with the engine
control system is required, then the highest risk class becomes
applicable, and subsequently paperwork and cost of monitoring
system development will increase.
Finally, a commercial aspect should be mentioned,
namely the insurance aspect. One could imagine that aircraft and
engines with monitoring systems - which are assumed to fly safer
than those without such systems - can be insured for lower rates.
Particularly for second hand users benefits are seen. They will
have advantages when purchasing aircraft with realistically logged
life consumption, and subsequently the first hand user can achieve
higher prices since continuous monitoring reduces uncertainties
about previous usage.
In the long term, a user of a life usage monitoring
system will have significant advantages with respect to the total
cost of engine ownership. The investment into the monitoring system
will be compensated in the first years in service by extended
inspection intervals, reduced spare part costs and optimisation
of logistics and maintenance.