Study and Implementation of Algorithms for in flight performance analysis of the

PW4000-100 Turbofan engine for the purpose of Engine Condition Monitoring

João Pedro Rego Freitas
joao.prfreitas171@gmail.com

Instituto Superior Técnico, Lisboa, Portugal

July 2014

Abstract
The first part of this paper describes the implementation of a Flight Data Monitoring (FDM) database in the Analysis Ground
Station (AGS) software for decoding recorded data on an Airbus A310. The results for a selection of recorded parameters are
analysed and the procedure that computes the flight phase is described. The second part of the paper describes the study
and implementation of algorithms for performance trend monitoring of PW4000-100 engines. Two tools were developed using
the R programming language. The first tool is responsible for the acquisition of stability points in cruise from recorded flight
data using specific conditions and criteria and, after validation, it was implemented in AGS. The second one uses the stability
points from the flights processed in AGS to derive engine baseline models and calculate the performance trends with respect
to these models. The engine parameter data is corrected using the inlet temperature and pressure and the performance
characteristics of the engines are displayed for standard day conditions. The engine baselines are composed by linear
regression models adjusted to the data obtained from the flights after the engines return from shop visits. The trend
re and confirmed the potential of the

shifts that might be representative of engine faults.

Keywords: Flight Data, Condition-Based Maintenance, Engine Condition Monitoring, Performance Trends, Corrected Gas
Turbine Parameters, Stability Points

capabilities to enhance safety levels and reduce airline

1. Introduction operating costs, allowing airlines to respond to the harsh
The aviation industry plays an important role in the global economic requirements.
economic activities. In 2013, airline scheduled operations
handled 3.1 billion passengers, up from 2.9 billion 2. The Airbus A310-325 Aircraft
passengers from 2012 [1]. Airbus [2] predicts that air The Airbus A310 is a wide body transport jet airliner
traffic will double in the next 15 years at an annual growth powered by two turbofan engines. The A310 made its
rate of 4.7% in RPK (Revenue Passenger Kilometers). maiden flight in 1982 and during its production run 255
Despite the excellent safety record of air travel, aircraft units have been delivered to different customers. The
accidents always have major economical and emotional A310-325 variant is powered by 2 PW4156A Engines,
consequences. The Flight Data Monitoring (FDM) is a providing 56,000 pounds of thrust each.
program used in the prevention of aeronautical accidents 2.1 On-board Transmission and Recording
that consists in the decoding and analysis of the data from
Protocols
the aircraft recorders on a routine basis. TAP Portugal
was one of the first airlines in the world to use its flight The communication between the main systems in the
data in a preventive way through the implementation of an Airbus A310 is made through the ARINC 429 protocol.
FDM program containing both FOQA (Flight Operations The format in which the data is stored in the different
Quality Assurance) and MOQA (Maintenance Operations recording units on-board the aircraft is specified by the
Quality Assurance) procedures. The objective of FOQA ARINC 717/573 protocol.
procedures is to identify divergent practices amongst crew 2.1.1 ARINC 429 Protocol
members or difficulties to adhere to the established The ARINC 429 is a protocol that defines how the
standard procedures. The information is then compiled different avionics equipment and aircraft systems should
into reports and corrective actions may be taken, leading communicate with each other. They are interconnected by
to improvements in the safety performance of the airline. wires in twisted pairs and the communication is done with
The MOQA procedures are used for maintenance a unidirectional data bus standard.
purposes and are mainly executed as a diagnosis tool, to
2.1.1.1 ARINC 429 Electrical Characteristics
help the various airline maintenance departments to
identify and correct on-going problems and failures in the Each ARINC 429 bus uses two signal wires to transmit
aircraft equipment or to conduct studies. the messages, which are transmitted at a bit rate of 12.5
Flight data can also be used to continuously monitor the or 100 kilobits per second. The transmission is
performance of the aircraft and the condition of its asynchronous, with consecutive words separated by at
systems. In addition to the diagnostic analysis, monitoring least 4 bits with null value (zero voltage), eliminating the
the aircraft/aircraft systems greatly relies on a prognostic need for an external clock signal. The bits are transmitted
analysis, where engineering personnel predict the time using a bipolar return-to-zero (BPRZ) and the information
left until component failure based on its deterioration is retrieved from the voltage difference between the two
levels. This type of analysis is known to reduce the risk of wires. The number of receivers connected to each bus is
potential failures during operation and allows planning for limited to 20 by the protocol.
maintenance in anticipation. 2.1.1.2 Format of the Data Word
The increasing recording capacity of the flight data
The ARINC 429 data words have a length of 32 bits. The
recorders, the development of newer aircraft fitted with
format is shown in figure 2.1 and includes five fields.
dedicated sensors and the prospect of transmitting real-
time information, give FDM programs almost unlimited
1
 transmitted to these units is organized in a frame (or
dataframe). Each frame corresponds to four seconds of
recorded flight data and represents information about the
Figure 2.1 ARINC 429 Word Format [3]
parameters being sampled with time.
Parity: The Most Significant Bit (MSB) is the parity bit. A frame is divided into four subframes and each takes
The ARINC 429 uses an odd parity, which means that the one second to be recorded. The subframes are composed
h 32-bit word should be odd. n+6
by 64, 128, 256, 512 or 1024 12-bit words (2 , with n
This bit is effective in the identification of transmission from 0 to 4). The words in ARINC 429 format are
errors when the number of incorrect bits is odd. converted to the ARINC 717 12-bit format by aircraft data
Sign/Status Matrix (SSM): Bits 30 and 31 contain acquisition units.
information about the conditions of the hardware 4 seconds
equipment, operational mode or validity of the data 1 second
content, in the form of a code.
Data: Bits 29 to 11 contain the data from the parameter(s) SUBFRAME 1 SUBFRAME 2 SUBFRAME 3 SUBFRAME 4
in various formats. There are some standard types of data
and non-standard types specifically implemented by the
manufacturers. Usually, not all the bits are used, only
those necessary to cover the range and resolution of the WORD 1 WORD 1 WORD 1 WORD n+6

information transmitted.
SDI: Bits 10 and 9 form a two-bit code that gives the
Source/Destination Identifier (or SDI). This is used to 1 2 3 4 5 6 7 8 9 10 11 12 12 bits
identify the receiver to which the information is destined Figure 2.4 Frame, Subframe and Word
when there are multiple receivers. It is also used to
identify the source of the transmission or, in some cases, Bit one corresponds to the Least Significant Bit (LSB) and
for transmission of data. is transmitted first, while bit 12 is the MSB and the last to
Label: Bits 8 to 1 contain a label identifying the data type be transmitted. By having information about the
of the 32-bit word and the parameters associated with it. subframe(s), word(s) and bit(s) it is possible to locate the
A label is assigned to each parameter that is transmitted parameter recordings in the frame. This information,
are together with the parameter ARINC 429 characteristics
listed in the ARINC 429 specification [4] and are and the recorded bits, provides the necessary tools to
represented as octal numbers. compute its value from the recorded flight data.
2.1.1.3 ARINC 429 Data Types 2.1.2.2 Regular Frame
BCD (Binary Coded Decimal): Figure 2.2 displays the Most parameters are recorded at least once every four
general format of a BCD word. The data field in a BCD seconds, i.e., in every frame. The recording rate of a
word contains up to five sub-fields, which represent regular frame parameter is directly related with the
different decimal digits. The SSM field gives the sign. number of times that parameter appears in the dataframe.
If the parameter is recorded once per second (1 Hz), it
appears once per subframe and a total of four times per
frame. The parameters can be output at lower rates of 1/4
Figure 2.2 General BCD Word Format [3] Hz or 1/2 Hz, and appear in one or two subframes,
Four bits are allocated to each digit, except for the most respectively. When a parameter is recorded at rates
significant sub-field, which encodes the digit with 3 bits higher than 1 Hz, there are various samples per subframe
only. and each one is called an instance of the parameter.
BNR (Binary): The BNR format stores the data as a 2.1.2.3 Superframe
binary number. The general format and fields are The superframe parameters have a recording rate of 1/64
represented in figure 2.5. Bit 29 is the sign bit and bit 28 is Hz and appear once every 16 frames. Superframe
the most significant bit of the data field. Negative numbers parameters are recorded on dedicated words within a
are encoded complement format. Bit 28
represents half of the maximum value (range) of the significant bits (12 to 9) of the superframe contain a
th
parameter, bit 27 represents 1/4 of the range, etc. counter that is incremented at the end of each frame.
S
change throughout a flight or that does not change as
Figure 2.3 General BNR Word Format [3] frequently as the accelerations, speeds, etc.
1
Discrete: The Discrete type of data is used whenever the 2.2 ATA 31 Recording Systems: Aircraft
parameter can be encoded with only one bit. The discrete Integrated Data System (AIDS)
data is accommodated in the unused pad bits of data The main function of the Aircraft Integrated Data System
words or in dedicated words, which can store up to 19 (AIDS) of the Airbus A310-325 is to convert the various
different parameters. critical (including the mandatory) parameters into a
2.1.2 ARINC 717/573 Protocol recordable format and to record them on a Flight Data
The ARINC 717 protocol was created to describe the Recorder (FDR). The aircraft from which the data will be
equipment and installation standards capable of meeting decoded and analyzed is fitted with a Basic AIDS
mandatory flight data recording requirements prescribed composed by the following equipment:
by the FAA (Federal Aviation Administration) and other A Digital Flight Data Acquisition Unit (DFDAU);
aviation authorities. A Digital Flight Data Recorder (DFDR);
2.1.2.1 Data Frame Description A three-axis linear accelerometer (L.A.);
A Control Panel (C.P.);
The format in which the data is stored in the different A Quick Access Recorder (QAR).
recording units on-board the A310-325 aircraft is specified
by the ARINC 717 protocol. The stream of data
1
Air Transport Association of America
2
 ARINC 429 BUSES

ANALOG ARINC 717 (573)

PARAMETERS DFDR
ARINC 717 (573)
DFDAU
DISCRETE
PARAMETERS
ARINC 717 (573) QAR

L.A.
: Basic AIDS

: Optional
C.P.
Figure 2.5 AIDS Architecture
The DFDAU acquires, conditions and processes the
aircraft parameters coming from the several sensors and
on-board computers [5]. The DFDAU then transmits the Figure 3.1 Regular parameter: Pitch Angle (PITCH)
required parameters to the DFDR in the format specified 3.2.2 Superframe Parameters
by the ARINC 717/573 protocol at a rate of 64 12-bit The frame recorded in the QAR also accomodates 4
words per second. The DFDR, commonly known as the superframes. To decode the superframe parameters it
, is designed to withstand the conditions likely was first necessary to define the location of the
to be encountered in an aircraft crash. The DFDR has the superframe words and the corresponding location of the
ability to store data collected during up to 25 hours of counters. In total, 31 of these parameters were added to
flight and when it reaches its maximum capacity the the dataframe and they all have a period of 1/64 Hz,
oldest data is overwritten by the new data. which means that they are only updated every 64
The DFDAU sends a copy of the DFDR data frame to the seconds.
QAR, which records the data on a magneto-optical disk.
The QAR is located in the avionics compartment and not 4. Decoding Results and Flight Phase
in rear of the aircraft close to the tail cone like the DFDR Computation Procedure
and is used by maintenance personnel acquire the
After programming and validating the dataframe in
mandatory parameter recordings.
database version 10079, the flight data from several
3. Dataframe Programming flights performed by the A310-325 aircraft was decoded
and analyzed. Some procedures that compute new
The flight data reading and processing tasks are carried
parameters from the recorded data were also
at computer ground stations by a flight data analysis tool:
implemented to the database, allowing an improved flight
the Analysis Ground Station (AGS) from SAGEM. AGS is
data analysis. In this section, examples of decoding
capable of decoding the data frames recorded in the
results are presented and the implementation of the Flight
ARINC 717/573 format containing the flight parameters.
Phase Computation procedure is discussed.
AGS works on a Database basis, which is the
configuration used to perform the flight data analysis for a 4.1 Results for Recorded Parameters
particular aircraft or fleet of aircraft. A new database Figure 4.1 displays the evolution of the pressure altitude
version was implemented containing the necessary or Standard Altitude (ALT_STD) as a function of the
information to decode the flight data recorded in the QAR current Greenwich Mean Time (GMT). This is the altitude
of the Airbus A310-325. in the International Standard Atmosphere (ISA) [6] with
3.1 Parameter Implementation the same pressure as the part of the atmosphere in which
the aircraft is flying. The profile shows some spikes during
Before the parameters are allocated to their respective
the climb and descent phases of the flight. This situation
word slots, the information regarding their ARINC 429
is corrected via an additional procedure that calculates
characteristics needs to be inserted in AGS. A total of 219
the Standard Corrected Altitude (ALT_STDC), ensuring a
parameters from various signal types were implemented:
smooth profile as depicted in the same figure.
172 coming from the digital buses of the aircraft in the
ARINC 429 format, 37 discretes and 10 analog
parameters. All these parameters are mandatory.
3.2 Dataframe Construction
3.2.1 Regular Parameters
The majority of the parameters are recorded in the regular
frame. The regular frame parameters can be added to the
dataframe structure by means of a dialogue window
similar to the one displayed in figure 3.1. The dataframe is
represented in the bottom part of the window word, bits
and subframes - and allows the user to visualize and
select the location of the parameter in the word slots.
Each parameter can contain up to three parts, which need Figure 4.1 Altitude, Pitch and Angle of Attack results
to be allocated separately. The recording rate for each The evolution of the Pitch Angle (PITCH) and the Angle of
part must be defined. AGS will use the ARINC 429 data Attack (AOAL) is also presented in figure 4.1. During most
bits information for each part to compute the correct of the climb the Pitch Angle is kept at a higher angle than
parameter values. For parameters with several instances the Angle of Attack, resulting in a positive Flight Path
in each subframe, it is only necessary to select the Angle and an ascending trajectory. The opposite occurs
location of one of them. AGS will automatically assign the during the descent. In cruise, the Pitch Angle and the
location of the remaining instances. Angle of Attack have approximately the same value and
during this phase the trajectory is horizontal. The biggest
difference between these two parameters occurs in the
3
final moments of the flight before Landing, when the computed from the Flap and Slat position recordings
Angle of Attack is much higher than the Pitch Angle. selected with the Flap and Slat Control Lever.
4.2 Additional Procedures: Flight Phase The results of the implementation are represented in
figure 4.17, which shows the evolution of the
Computation
FLIGHT_PHASE parameter throughout a selected long-
The Flight Phase Computation procedure is of extreme haul flight. The transitions between the flight phases occur
importance for the flight analysis. This procedure in the correct order defined in table 4.1. The transitions to
computes the Flight Phase based on several recorded higher flights levels in cruise are indicated by the periods
and computed on ground parameters. The Flight Phase where the FLIGHT_PHASE transitions from Cruise (8) to
parameter (FLIGHT_PHASE) is used in a variety of Climb (7) and then back to Cruise (8). The aircraft does
maintenance and flight operations related procedures and not go through a Rejected Take-Off (4) or a Go-Around
is also mandatory for the correct separation of each flight (13), as expected in normal operation. Results similar to
contained in a media. those in the figure below were encountered for the other
4.2.1 Flight Phases Definition flights and demonstrate the correct implementation of the
The procedure was implemented based on the definition procedure and its usefulness to the analysis.
from SAGEM that includes a total of 14 flight phases [7],
similarly to what happens in the databases for TAP fleet.
Nr. Phase Nr. Phase
1 Engine Stopped 8 Cruise
2 Taxi-Out 9 Descent
3 Take-Off 10 Approach
4 Rejected Take-Off 11 Final Approach
5 2nd Segment 12 Landing
6 Initial Climb 13 Go-Around
7 Climb 14 Taxi-In
Table 4.1 SAGEM Flight Phases

Figure 4.3 Flight Phase Results

5. Engine Condition Monitoring
In this section, the fundamentals of Engine Condition
Monitoring (ECM) are introduced.
5.1 The Condition-Based Maintenance
Philosophy
Over the last decades, aircraft maintenance has evolved
according to the existing aviation safety standards and the
economic requirements from the air transport industry.
The Federal Aviation Administration (FAA) recognizes
three Primary Maintenance Processes [8]:
Hard-Time (HT);
On-Condition Maintenance (OC);
Condition-Monitoring (CM).
The Hard-Time is a preventative maintenance process
that defines a limit for the parts in terms of flight hours,
Figure 4.2 Flight Profile with the SAGEM Flight Phases [7] cycles or calendar time. The On-Condition maintenance is
4.2.2 Implementation also a preventative process where an appliance or part is
periodically inspected or checked against some
The conditions for the transition between the flight phases appropriate physical standard to determine whether it can
were modified and adjusted according to the limitations of continue in service. The On-Condition maintenance tasks
the recordings in the QAR. The N2 (high pressure rotor range from visual inspections to workshop and laboratory
speed of the engine) was substituted by the EPR (engine tests. Condition Monitoring is a predictive approach where
pressure ratio) to make the detection of the Take-Off. This the components/parts can actually fail in service. The
is because the N2 parameters are recorded in the target of condition monitoring is to define improvements in
superframes and most of the times they are not updated the maintenance programs by increasing the aircraft
during the take-off runs. This resulted in abnormal availability and reducing costs.
transitions from the Taxi-Out to the Climb phases, for Condition-based Maintenance is a concept for maintaining
instance. The use of the EPR, recorded at 1 Hz, ensures gas turbine engines that is gaining acceptance between
the normal transition from Taxi-Out to Take-Off and airlines and engine manufacturers [9]. The maintenance
nd
afterwards to the 2 segment and subsequent phases. needs are determined based on the actual condition of
The vertical speed (IVV) is not recorded and is computed the engine rather than on a preset schedule [10]. The
from the variation in the altitude. The HEIGHT parameter objective
computes the altitude above ground level (or AGL overhaul only when the engine needs major maintenance
altitude). At low altitudes up to 2500 ft. - it corresponds work. With the Condition-based maintenance concept,
to the value given by the radio altimeter. Thus, the true aircraft engines are subject to control by the three Primary
vertical distance to the ground is used. At higher altitudes Maintenance Processes [8]. In order to have a regular
overview of the proper functioning of the engines, Engine
therefore, the HEIGHT parameter consists of an estimate Condition Monitoring is applied. It consists of a wide
that is computed from the corrected standard altitude range of activities where the health of aircraft engines is
(ALT_STDC) and the altitude standard of the origin and assessed and followed on a routine basis, from the
destination runways. The Flaps and Slats configuration is moment the engine is put on-wing until its removal [11].
4
The information is then used to ensure that preventive in the experienced conditions. The difference between the
action is taken at an early stage before safe operation is measured data and the reference model is called the
affected and that the root causes of the problem are trend delta or parameter delta.
identified. 5.4.1 Cruise Performance Trends
5.2 Systems for Data Collection and Analysis The most widely used method in the industry for cruise
The most basic system for collecting engine data consists trend monitoring is to compare parameter data from each
of manual recordings by the flight crew. The crew is presented in
2
generally requested to take into account some standard day conditions [13]. The baseline models are
constraining stability criteria. One of the disadvantages is generally developed by the engine manufacturers based
that the parameters measured are limited to those with upon flight-test data and/or in-service experience. For the
cockpit instruments. Modern commercial aircraft are comparison between the raw recorded data and the
equipped with data acquisition systems that automatically
record data for monitoring purposes. They are capable of to correct the engine parameter data so that the
recording a much larger number of engine parameters, deviations are determined. The correction procedure and
coming from an increased number of sensors in the the expressions are covered in the next section.
engines. Gradual vs. Rapid Performance Shifts: There are two
major types of movements that can be identified in a trend
Automatic Data Acquisition [8]: Step Shifts and Slow Drifts.
Manual Data Acquisition

Manually
Recorded
Data

Powerplant Engineering
ARINC TAP Server
Department
Satellite G/S
Radio G/S

Figure 5.2 Slow Drifts vs. Step Shifts [8]

Figure 5.1 Engine data acquisition processes at TAP As the engine accumulates flight cycles its performance
deteriorates due to several reasons: dust/dirt ingestion
The most efficient method for automatic data acquisition and accumulation on the turbine and compressor blades,
consists of transmitting the in-flight recorded data to increasing tip clearances on the compressor and turbine
ground stations via ACARS (Aircraft Communications blades, erosion of the airfoils, hot section oxidation, etc.
Addressing and Reporting System). Although the cost of Slow Drifts represent the gradual deterioration of the
using ACARS is higher when compared to the other performance of the engine that is caused by these factors.
methods, there is the added benefit of making the Step Shifts indicate mechanical changes in the engines
information immediately accessible to the engineers, who that may propagate to failure and lead to the occurrence
can conduct real-time/near real-time assessments and of incidents, such as in-flight shutdowns and aborted take-
plan in advance any required maintenance actions to be offs. The effective monitoring of the cruise trends helps to
undertaken on the aircraft, thereby optimizing the time on minimize the risks associated with unexpected engine
the ground. The major disadvantage of manual data failures, which result in unscheduled engine removals with
acquisition processes is that it can take a long time from additional maintenance costs to the airline, and to avoid
the moment the recordings are executed to the moment
when they are analyzed. checking the performance delta parameters against
Once the data is available it is interpreted by ECM potential problem limits.
software, typically provided by the Original Equipment
Manufacturers (OEM) of the engines. The primary 5.4.2 EGT Margin
function of the software tools is to perform parameter To meet aircraft performance requirements, engines are
trend monitoring. This is the primary process used by designed to provide a constant Thrust up to a designated
. Flat Rate Temperature (FRT). Below this temperature the
Thrust is limited by software. When the Outside Air
5.3 Engine Parameter Measurements Temperature (OAT) is lower than the FRT, the EGT is
There are two types of parameters used for monitoring less than the limit. This limit is called the EGT Red Line
gas turbine engine condition [12]: the mechanical and is demonstrated during endurance tests required for
parameters and the performance parameters. The first engine certification [14]. When OAT is higher than FRT,
group of parameters includes the engine vibrations and oil the fuel flow has to be reduced in order to keep the EGT
temperature and pressure. These parameters are not below the limit and protect the turbine hardware.
significantly influenced by the flight conditions and engine The EGT Margin is an estimate of the difference between
thrust setting, contrary to the performance parameters. the certified EGT Red Line and a projection of the engine
Today, most of the airliners in operation are equipped with EGT to full (non-derated) take-off reference conditions.
turbofan engines. They produce lower noise levels than The observed/recorded peak EGT during the take-off is
earlier generation jet engines and have considerably projected to the reference condition of full take-off power,
improved fuel economy. The key performance parameters on a FRT day at sea level. This projected temperature
in a turbofan engine are the EPR, the Low-pressure rotor represents the expected EGT if the take-off actually
speed or Fan speed (N1), the High-pressure rotor speed occurred with the reference conditions.
or Core speed (N2), the Exhaust gas temperature (EGT) The EGT Margin is routinely used to monitor the health of
and the Fuel Flow (FF). the engines, together with the trends from the cruise
5.4 Parameter Trend Monitoring performance parameter deltas. EGT Margin can be used
to forecast the remaining time of the engine on the wing
Parameter trend monitoring is the process in which the in-
flight results are processed and then compared to a 2
Temperature: 288.15 K, Pressure: 1013.25 hPa (Sea-level ISA
baseline model of how the engine is expected to perform atmospheric conditions)
5
by predicting the point at which the margin will be 6.2.1 Observation Window and Stable Frame
completely eroded. A period of time identified as a stability point is called in
6. Study of Algorithms for ECM Trend the A330-223 aircraft a Stable Frame and has a duration
of 100 seconds. Before searching for stable frames, there
Monitoring are some conditions that need to be verified first. They
The main objective of this work is the development of are used to avoid report triggering in flight phases where
algorithms for performance trend analysis of engines. The the parameters are of no interest and are called the Basic
methodology presented in this section and the next relies Conditions. When the Basic Conditions are met, the DMU
only on recorded flight data to derive the engine baseline searches for stable frames. It looks to the parameter data
models and determine its deterioration levels. The in each 100 seconds of flight and then computes the
objective is to use this approach in complement to the difference between the maximum and minimum values
trend monitoring tools s software. and compares it to .
6.1 Case Study: PW4168A Engine This first method of searching stable frames consists of
performing computations for each individual 100 seconds
The studies were conducted on the PW4168A engine, observation window in the flight. Individual means that
-223 aircraft. It is a dual rotor each window is independent from the others, covering
(two spool), axial flow turbofan engine with separate data from different periods of time. The second method
primary and fan duct exhaust systems. The engine uses a gliding window, where the observation window
delivers a maximum static take-off thrust of 68,600 lbs. at advances seconds. If is small (e.g. 1 second) the
sea level conditions and has a bypass ratio of 4.9. number of observation windows considered for the stable
Figure 6.1 identifies the different Gas Path configuration frame search increases significantly. The A330-
areas on the PW4168A engine and the Engine Stations. DMU searches for the best stable frame using a gliding
The latter correspond to location points in the engine gas window method with an advancing front of 20 seconds
The ( ). In the algorithm developed, the advance in the
EPR is the ratio of the turbine discharge total pressure gliding window is equal to the period with which the flights
(Station 4.95) to the compressor inlet total pressure were exported from AGS.
(Station 2). The EGT is the total temperature of the low
6.2.2 Quality Number
pressure discharge gas flow (Station 4.95).
LOW PRESSURE
If several stability points are identified it is necessary to
COMPRESSOR
have a consistent procedure that selects the best of them.
Stability points are selected based on the quality number
[15]. The Engine Quality Number (QE), is the sum of each
of the individual quality numbers of the stable frame
parameters and is computed using the following formula:

(6.1)

Where is the weight factor, is the tolerance or the

maximum variation of the parameter allowed in the
observation window and is the variance.
6.2.3 Algorithm for Extraction of Stability Points
The algorithm was developed using the R programming
Figure 6.1 PW4168A Turbofan engine language and consists of two main parts: the Basic
Engine Cruise Report (01): The main component of the Conditions and the Stable Frame Criteria.
Aircraft Condition and Monitoring System (ACMS) is the 6.2.3.1 Basic Conditions
Data Management Unit (DMU). One of the tasks The conditions in table 6.4 must remain true during the
performed in real time by this unit is the generation of the observation window before stable frame calculations for
the maximum and minimum values are performed and
conditions. Within the scope of this work, the interest is on stability points encountered.
the Engine Cruise Report (01), used for engine trend
Logic for True Condition Parameter(s)
-223 fleet. It consists of a Nacelle Anti Ice Status Off FNAI1=1, FNAI2=1
collection of both aircraft and engine information. Wing Anti Ice Status Closed WAO1=1, WAI1=1, WAI2=1, WAO2=1
20000ft.<Altitude<41100ft. ALT_STD
6.2 Stability Points 0.6<Mach Number<0.86 MACHR1
To monitor the performance of an engine and to evaluate 70<N1K<120 N11, N12, TAT
its deterioration, it is important to collect data that is N1K=N1/(?2^0.5)
?2=(TAT+273.15)/288.15
representati The Engine
Cruise Report, in particular, contains information about Symmetrical Engine Bleed EG_PRV_POS=O_ENGPRV_POS,
Configuration EG_HPV_POS=O_ENGHPV_POS
Cross Feed Valve Closed XBV_POS2=0
during the Cruise flight phase, where a set of conditions APU Bleed Valve Closed APUBV_O=0
and stability criteria were respected during a certain Table 6.1 Logic for True Basic Conditions and Parameters
period. A point in these conditions is also called a Stability
When available, the parameters that are corrected in AGS
Point. This section presents the development of an
using additional procedures - ALT_STDC, N11C, N12C -
algorithm for extracting stability points from flight data
were used. This holds true for the stable frame
recorded in the DAR (Digital ACMS Recorder). First, the
parameters.
stability criteria and trigger conditions employed by the
DMU were used. Then, a new set of more restrictive 6.2.3.2 Stable Frame Criteria
stability criteria was applied, leading to a reduction in the Table 6.6 contains the stability criteria initially
number of stability points for each flight. The objective of implemented in the algorithm.
this study was to acquire stability points with better quality
and use these points for a trend monitoring analysis.
6
 IALT 100 ft. MACHR1 0.008 stable frame. When the cycle reaches the end of the
GS 5 Knots N11, N12 1.8% flight, a file with the recorded data and the results is
ROLL 0.8º PT21, PT22 0.05 PSIA written. If stability points are encountered two additional
TAT 1.1ºC FF1, FF2 200 kg/hr files are written: one containing the data from all the
N21, N22 1.0% EPR1, EPR2 0.05 stability points found and another containing the best
EGT1, EGT2 22ºC HPT1, HPT2 5% point from each search period.
GVRTI 0.05 g LPT1, LPT2 5%
Table 6.2 Aircraft Stability Criteria 6.2.4 Results
6.2.3.3 Filtering of Parameters The algorithm was tested with a selection of flights
performed by different aircraft. After this implementation, it
The on-board systems of the aircraft also filter the was found that stability points were not being generated
parameter data. The formula used to filter these readings from approximately 50% of the flights. This was because
is the following: in many of these flights the anti-ice discretes maintain the
) (6.2) ON status and therefore the Basic Conditions were never
Where: met. However, this is an expected situation during normal
NFV = New Filtered Value t = Samples per Second operation.
OFV = Old Filtered Value T = Time Constant
NRV = New Raw Value
: Waiting Times
The division is constant for the filtered parameters
and equal to 1/3.
Basic Parameters ALT_STDC, MACHR1
ROLL, TAT, N21C, N22C, EGT1C,
Stable Frame
EGT2C, GVRTI, N11C, N12C, FF1C,
Parameters
FF2C
Table 6.3 List of Filtered Parameters
6.2.3.4 Description
Figure 6.2 contains the flowchart for the algorithm, which
is essentially divided in two parts. The first part is
responsible for reading and processing the files Figure 6.3 Flight Results obtained with the algorithm
containing the flight data and the second part writes the Figure 6.3 illustrates some of the results obtained from a
results to dedicated files. The flights were exported from processed flight. The altitude (ALT_STDC) is also
AGS in .csv format with a parameter rate of 8Hz to take represented to illustrate the climb, cruise and descent
into account all the recorded values. flight phases. The flag_cruise variable indicates that the
The algorithm starts by loading the flight data in the file. aircraft is in Cruise from the moment it finishes its initial
Afterwards, there is a cycle that reads every line of data. climb up to a stable cruise altitude until starting the
LOAD FLIGHT
descent for landing. The exception is made to some
periods where the aircraft transitions between flight levels.
READ FLIGHT DATA The basic_conditions variable indicates when the Basic
YES WRITE FILE WITH RECORDED
Conditions are met. For this flight, Basic Conditions exist
END OF FLIGHT?
DATA AND RESULTS during most of the cruise, except for the initial 15 minutes
NO
FILTERING OF THE PARAMETERS
after cruise is reached for the first time. The counter
NO
NR. OF STABILITY
POINTS FOUND > 0? variables for the search and waiting times are illustrated
NO FLIGHT PHASE = CRUISE
(FLIGHT_PHASE=8)?
YES
with red and black lines, respectively. After the basic
NO
YES
WAITING TIME > 15
WRITE FILE WITH ALL THE
STABILITY POINTS
conditions are met for the first time, a search period of 4
MINUTES?
YES
hours (144000 eighths of second) for stability points
NO ARE THE BASIC WRITE FILE WITH THE BEST begins. The search time is incremented whenever the
CONDITIONS MET?
YES
STABILITY POINT(S)
basic conditions are true and in cruise. After 4 hours,
SEARCH FOR BEST STABILITY
POINT
there is a waiting time of 15 minutes before starting a new
TO NEXT FLIGHT OR END
ANALYSIS search period. The stability_point variable indicates when
NO ARE THE STABILITY
CRITERIA SATISFIED?
YES
CALCULATES MEAN AND VARIANCE
corresponding search period (1 or 2, in this case).
Aircraft Stability Converged
Parameter
CALCULATES QUALITY NUMBER
Criteria Stability Criteria
IALT 100 ft. 35 ft.
Figure 6.2 Algorithm for Extraction of Stability Points GS 5 Knots 2 Knots
The first condition is based on the Flight Phase parameter ROLL 0.8º 0.6º
TAT 1.1ºC 0.5ºC
(FLIGHT_PHASE). Stability Points can only be generated N21, N22 1.0% 0.2%
in Cruise, when FLIGHT_PHASE=8. Next, a condition for EGT1, EGT2 22ºC 3ºC
the Waiting Time was implemented. The Waiting Time GVRTI 0.05 g 0.04 g
corresponds to a period of 15 minutes when stability MACHR1 0.008 0.0035
N11, N12 1.8% 0.3%
points cannot be searched or generated that occurs in two
PT21, PT22 0.05 PSIA 0.04 PSIA
situations: FF1, FF2 200 kg/hr 50 kg/hr
st
1) After Cruise is attained for the 1 time; EPR1, EPR2 0.05 0.05
2) After a search period of 240 minutes. HPT1, HPT2 5% 5%
LPT1, LPT2 5% 5%
Finally, the Basic Conditions in the observation window Table 6.4 Aircraft and Converged Stability Criteria
are checked using the logic presented in table 6.1. If they
are true, the algorithm will search stability points for a For the flights studied there were thousands of stability
maximum of 240 minutes. If the differences between the points encountered. This showed the potential to study a
maximum and minimum values of the stable frame new set of more restrictive criteria. Table 6.4 displays the
parameters are lower than the tolerances in table 6.2, a tolerances for the Converged Stability Criteria, together
stability point is generated and the quality number for the with the tolerances from the aircraft. The new criteria are
7
the result of an iterative process: each time the tolerances
were reduced, the flights were processed and the number (6.8)
of stability points checked. After this iterative process,
only approximately 10% of the exported flights did not (6.9)
generate any stability points.
6.3 Engine Parameter Corrections The EGT is converted from degrees Celsius to kelvins.
The theta correction factor, , is equal in both engines
Ambient conditions have a significant impact on the
because it is computed from the TAT. The delta correction
various parameters along
factor, , is computed from the respective total pressure
flows, temperatures, pressures, speeds, etc. These not
only vary with the power condition, but also with the
temperature and/or pressure at inlet [16]. A 7. Baseline Model Definition and Trend
change in these conditions contributes to an attendant Monitoring Results
difficult to characterize the aero-thermodynamic The engine baseline model and trend monitoring results
relationships between the different gas turbine engine are presented in this section. The source of engine
parameters (even at a constant engine operating point) parameter data are the stability points that are acquired
unless the ambient conditions are accounted for. with the algorithm described in the previous section after it
Generally, these relationships are determined by using was adapted and implemented in AGS.
corrected engine parameters. 7.1 Engine Baseline Model
6.3.1 General Formulation This section presents the study that was undertaken with
A change in the inlet temperature ( ) and pressure ( ) is the objective of defining the cruise performance
accompanied by an attendant change in any downstream characteristics of the PW4168A engine. The engine
gas path parameter , The corrected parameter parameter data in the baseline models is corrected with
represents the value that the parameter would have at a the correction factors and the equations from section 6.3.
fixed reference inlet condition ( , ). It is a common 7.1.1 Engine Selection for Baseline
practice to select standard day atmospheric conditions at As a starting point, it was necessary to access the records
sea-level as the reference condition. The most widely -223
used parameter correction formula has the following form: fleet. This was done in order to select a set of engines to
(6.3) be used for deriving the baseline models.
REMOVED ENGINE INSTALLED ENGINE
Where and are the relative temperature and relative A/C Tail Pos. EGT Date of EGT Date of
pressure, respectively. They are defined as Margin Removal Margin Installation
and . Generally, the exponents CS-PWA 1 15ºC 09-09-2013 42ºC 10-09-2013
and will depend on the engine and cycle. However, CS-PWA 2 34ºC 26-09-2013 30ºC 26-09-2013
there are some classical corrections and approximations CS-PWB 1 -6ºC 07-03-2013 23ºC 07-03-2013
that are commonly used for all gas turbine engines. Table CS-PWB 2 -8ºC 21-02-2013 34ºC 22-02-2013
6.9 summarizes some of these corrections [16]. Table 7.1 Engine Installations after Shop Visits3
Corrected The two aircraft listed in table 7.2 have been selected for
Parameter
Parameter (K) deriving the engine baseline models. This selection was
Rotor Speed (N) 0.5 0 mainly based on the EGT Margin of the installed engines.
7.1.2 Bleed and Pack Selection
0.5
Fuel flow (FF) 1 Bleed air can have a major effect on the performance
(classical)
1 curves of an engine and needs to be accounted for in
Temperature (T) 0
(classical) order to provide accurate estimates. To take into account
Table 6.5 Common Gas Turbine Parameter Corrections the bleed air effects in the ECM analysis, an Aircraft
6.3.2 Application to the Case Study Engine Bleed Code (ABC) and an Air Conditioning Pack Code
The computation of the theta and delta correction factors (PKS) are computed from the recorded data using the
is done using the total temperature and the total pressure ENGINE BLEED P/b switch and the PACK FLOW
The Selection switch discretes.
computation is done with the average values of the TAT, 7.1.3 Results
PT21 and PT22 parameters from the stability points. The points used in the definition of the baseline models
were collected over a period of three months after the
(6.4)
respective engines were placed on the wing. This period
was found to be a good compromise between the number
(6.5) of points, the range of conditions experienced and the
The free stream total temperature (TAT) is used for the gradual deterioration the engine experiences.
total temperature at both engine inlets. 7.1.3.1 Baseline Model with EPR
In the PW4168A, there are four key engine performance The results in figure 7.3 represent the variation of the
parameters being monitored: EGT, Fuel Flow, N2 and N1. corrected EGT, as a function of the EPR, for the two
The values contained in each stability point are corrected engines on the CS-PWA aircraft. The data from each
for changes in the inlet condition using the equations engine was collected using with the Converged Stability
below. Criteria in table 6.4, over the periods of time indicated,
with bleed air being extracted from the engines (ABC=1)
(6.6)
and a normal pack flow selection (PKS=2).

(6.7) 3
Due to TAP confidentiality policies, the A/C tails correspond to
fictitious registrations, which do not exist at the date this work
was done.
8
 from the measured and the baseline points yields the
delta parameter.

Parameter
Cruise
Performance Measured Point
Parameter (Corrected Data)
Baseline Point
Corrected
Corrected Baseline
y=b0+b1x (Adjusted for Bleed and Pack Selection)

N1K

Figure 7.3 Delta Parameter Calculation

The equations used in the results of this section are
Figure 7.1 Baseline Results: EGTK vs. EPR written below.
From the analysis of these results, it becomes clear that (7.1)
the EPR is restricted to certain values. This occurs (7.2)
because the EPR parameters are recorded with a
resolution of 0.0625, meaning that the points were
probably acquired when the actual EPR was different (7.3)
from the recorded value.
7.1.3.2 Baseline Model with N1 Smoothed delta parameters are calculated using equation
5.2 with a smoothing coefficient of 0.2. The calculation of
The N1 was selected as the thrust reference parameter
the cruise performance trends is one of the main tasks
instead. The results in figure 7.2 represent the expected
executed by a dedicated program written in R. The
performance characteristics of the PW4168A engine for
program receives the stability points from AGS and
the corrected EGT. The data was collected from the CS-
corrects the EGT, FF, N1 and N2 values from each point.
PWA aircraft in the same conditions as those in figure 7.1.
Then, the Engine Baseline Model is derived from the
corrected engine data of the stability points for the
specified engine and time interval. The raw deltas are
calculated next with equations 7.1 to 7.3 and smoothed.
The cruise trends are output in the form of plots.
Stability Points

Calculate the Correction Factors

and Corrected Parameters

Definition of the
Engine Baseline Model

Calculate the Cruise Trends

Figure 7.2 CS-PWA Baseline Model Results

Plot the Delta Performance
The characteristics are much clearer in comparison with Parameters
those in figure 7.3. The corrected EGT values are higher Figure 7.4 Process of Plotting the Cruise Performance
for Engine 2 than for Engine 1. The latter was installed in Trends
the aircraft with 42ºC of EGT Margin and the first with 7.2.2 Results
30ºC, which allows to conclude that the margin is in fact a
good indicator of the relative performance of the engines. The plots in figure 7.5 contain the results for the results
Simple linear regression models were fitted to the EGTK, for the CS-PWB aircraft from December 2012 to February
FFK and N2K data using the least squares approach. The 2014. The flights from engine nr.2 after the installation in
2
R values showed that the models obtained with both the February 2013 were used for the baseline model.
aircraft and the converged criteria fit well the data. As The results for the delta EGT show very high initial deltas
expected, the number of points obtained with the latter for both engines. When engine nr.1 was replaced in
criteria is lower due to the more restrictive tolerances. February 2013 and engine nr.2 in March 2013, they
Aside from this, the results obtained with the two criteria presented negative EGT Margins. The breaks in the data
were very similar. Although the Reynolds number effects correspond to a period of approximately two months when
are not considered with equation 6.3, it was also the aircraft was out of service for maintenance.
concluded that it was not essential to consider different There are two step shifts, which occur outside the dates
model equations for different altitudes. for the engine removals/installations in figure 7.5. Both
shifts are characterized by an increase in the delta EGT
7.2 Trend Monitoring and delta FF and a decrease in the delta N2K. The two
7.2.1 Calculation of the Cruise Trends situations were detected with the P&W EHM software and
Figure 7.8 illustrates how the delta parameters are were related with faults in the Turbine Case Cooling
calculated. The measured point contains the corrected (TCC) System, which is responsible for reducing the
4
engine value from the cruise performance parameter at a turbine blade tip clearance during take-off, climb and
given corrected thrust (N1K). For this thrust, a baseline cruise operation for better fuel efficiency.
value is calculated using the equation of the model that
was derived. The difference between the parameter value 4
Distance between the blade-tip and the engine casing.
9
 Engine Nr.2
8. Conclusions
Step Shift The current work makes the best use of the engine data
available from on-board recorders of a fleet of aircraft
equipped with Pratt & Whitney engines for tracking engine
operations with performance trend monitoring tools.
The decoding results confirmed the extreme quality of the
Engine Nr.1 recorded flight data on-board the Airbus A310. The
Out for Maintenance Step Shift implemented database can be expanded to include FOQA
and MOQA procedures. However, this is limited to the
recordings in the QAR that became evident when the
flight phase computation procedure was discussed.
The results obtained with the algorithm for the extraction
(a) Delta EGT
of stability points presented motivated the study a new set
of stability criteria. The baseline models derived with
another tool showed that the results acquired with the two
sets of criteria were similar, aside from the difference in
the number of points. The classical parameter corrections
resulted in clear performance characteristics of the
engines that were fitted with linear regression models.
The trend monitoring results confirmed that engines with
higher installed EGT Margins display better performance.
The maintenance records confirmed that the observed
shifts in the results were representative of engine faults.
In addition, they were similar to those obtained with the
and confirmed that the
(b) Delta FF tools that were developed may be used as a complement
to this software to help powerplant engineers in the

9. References
[1] ICAO, "Safety Report - 2014 Edition," 2014.
[2] Airbus, "Future Journeys 2013 2032," Global Market
Forecast, 2013.
[3] Condor Engineering, "ARINC Protocol Tutorial," 2000.
[4] ARINC, "ARINC Specification 429 Part 1-17: Mark 33
Digital Information Transfer System (DITS)," 2004.
[5] Teledyne Controls, "Functional Specification of Digital
(c) Delta N2 Flight Data Acquisition Unit for Airbus A300 and A310
Figure 7.5 CS-PWB: Cruise Performance Trends Aircraft," Los Angeles, California, 1993.
The P&W EHM results for engine nr.2 are illustrated by [6] Mustafa Cavcar, "The International Standard
the Long Term Plot Report in figure 7.6. Having into Atmosphere (ISA)," Anadolu University, Turkey, 2000.
account the manufacturer fingerprints for the PW4168A [7] Sagem, Analysis Ground Station V12 - Method for
engine and the skills built from previous experience, Database Programming. Paris, France: SAFRAN, 2008.
powerplant engineers have ordered the inspection of the [8] General Electric, "Diagnostics Trend Interpretation
TCC system. From this inspection, maintenance Training," 2008.
personnel have found a faulty HPT valve which was
[9] Irem Y Tumer and Anupa Bajwa, "Learning about how
promptly substituted and the engine returned to the aircraft engines work and fail," AIAA Paper AIAA-99-
previous performance as illustrated in figure 7.15. 2850, 1999.
[10] S K Sethiya, "Condition Based Maintenance (CBM),"
Secy. to CME/WCR/JBP, 2006.
[11] R. Vingerhoeds, P. Janssens, B. Netten, and M. Aznar
Fernandez-Montesinos, "Enhancing off-line and on-line
condition monitoring and analysis," Control Engineering
Practice, vol. 3, no. 11, pp. 1515-1528, 1995.
[12] Pratt & Whitney Aircraft, "Malfunction Analysis Through
Gas Generator Comparisons and Mechanical Trends,"
Aircraft Gas Turbine Operation Information Letter 1965.
Figure 7.6 P&W EHM Long Term Trend Plot Report
[13] GE Aircraft Engines, CF-6 - Guidelines for Parameter
The same pattern in the delta EGT, delta FF (WF) and Trend Monitoring, 1993, Transmittal Sheet.
delta N2 parameters can be observed in figure 7.13. The
shifts in figures 7.13 and 7.15 are practically identical in [14] Arnaud Bonnet, "Avoiding high speed rejected takeoffs
terms of their amplitude. In addition, the deterioration due to EGT limit exceedance," Safety First - The Airbus
Safety Magazine, no. 4, pp. 8-13, 2007.
levels of the engine before the failure and after
maintenance work was conducted are similar in both [15] Airbus, Getting to Grips with Aircraft Performance
figures. This might indicate that the performance of the Monitoring, 2002.
engine used to derive the baseline model from the data is [16] Allan J Volponi, "Gas turbine parameter corrections,"
similar to the performance of the engine used by the Journal of engineering for gas turbines and power, vol.
manufacturer software. 121, no. 4, pp. 613-621, 1999.

10