26th ICAF Symposium – Montreal, 1-3 June 2011

Development of Load Spectrum for Full Scale Fatigue

Test of a Trainer Aircraft

Andrzej Leski, Piotr Reymer, and Marcin Kurdelski

Air Force Institute of Technology, Warsaw, Poland

Abstract. PZL-130 „Orlik” is a turbo-propeller engine trainer airplane, entirely

designed and build in Poland. This plane is used in Polish Air Force mainly for
primary training of polish military pilots.
Current modernization of this airplane (which mainly consists of the engine and
wings conversion) is enhanced by major change in the maintenance system. Main-
tenance system change program (SEWST) is carried through by EADS PZL-
Okęcie (manufacturer) and ITWL (Air Force Institute of Technology). One of the
major program task is a Full Scale Fatigue Test of the airframe which will be con-
ducted by VZLU in Czech Republic. This article will present the methodology
which was used to prepare a characteristic load spectrum for this test.
Following sections describe the whole process from statistical analysis of data
recorded in the onboard flight recorders through strain gauge based flight load
measurements to development of a fatigue load spectrum. As a result a character-
istic flight load spectrum block was obtained representing equivalent 200 flight
hours. This spectrum will be used in a Full Scale Fatigue Test which is designed to
estimate fatigue life and critical points in the structure.
After completion of all the above described steps the obtained spectrum was
both statistically consistent and representative for characteristic flight program in
Polish Air Force. This guaranteed that the fatigue test based on this load spectrum
will estimate actual fatigue life for these aircrafts operated in Polish Air Force and
determine critical points in which structural damage may occur during operation.

1 Introduction
The PZL-130 "Orlik" trainer aircraft was designed in Poland in the end of the
twentieth century. In 1994 it was introduced to the Polish Air Force (TC-I ver-
sion). It is a single-engine, two-seated aircraft used for preliminary pilot training
and display flying. These aircrafts are still operated and the users are very satisfied
with it's performance. The main disadvantages of the TC-I version are necessary
frequent overhauls (every 1000 flight hours) and frequent periodic maintenance
performed by user. Such maintenance system was a result of suspension of the
former research program which goal was to determine actual fatigue life under
Polish Air Force operation characteristic flight profile. No Full Scale Fatigue Test
was ever carried through. The fatigue life was estimated using analytical methods.
The experience gathered throughout over 20 years of operation showed, that the
current operation method was not optimal and the need to perform overhaul every
574 A. Leski, P. Reymer, and M. Kurdelski

1000 flight hours results in unavailability of the fleet due to repairs. The TCII ver-
sion was developed in the 1990's. The most important changes were:
- Pratt & Whitney engine (700BHP),
- 5 blade propeller,
- new wing geometry,
- new fin geometry.
The manufacturer has provided the opportunity to upgrade the airplanes from TC-I
to TC-II version. Two airplanes were upgraded to TC-II version in the early
twenty-first century. Positive experiences – mainly a significant improvement in
flight performance - resulted in modernization order for another 14 airplanes. In
the mean time PZL Warszawa-Okęcie, the manufacturer of the aircraft, became a
part of the EADS.
Along with the airplane modernization the Polish Air Force ordered develop-
ment of a new modern maintenance system for the PZL-130 Orlik TC-II. The
main requirement for the new system, were:
- avoidance of overhaul every 1000 flight hours,
- confirmation (and most preferably excess) of the previously estimated total fa-
tigue life of 6000 flight hours,
- development of the structure integrity program suitable for this aircraft.
Realization of these objectives is the subject of SEWST research program per-
formed by EADS PZL-Okęcie and ITWL.

2 Sewst Program Definition

The whole SEWST program covers many different topics. The most important
part is the confirmation of the total fatigue life without the need of overhaul. The
main objective of the whole program, as it comes to the aircraft structure, is the
Full Scale Fatigue Test. As stated above aircrafts intended for the Polish Air Force
will be the modernized version of the aircraft previously operated as TC-I. During
the modernization aircrafts are fitted with a completely new wing and the fin's ge-
ometry is slightly changed. It was therefore decided that the fatigue test will be
performed for an equivalent structure. For the Full Scale Fatigue Test purpose
EADS PZL-Okęcie has performed a partial modernization (from TC-I to TC-II
covering the structural changes only) of an aircraft already withdrawn from opera-
tion (No. 015).
ITWL is a research institute subordinate to the Polish Ministry of Defense and
executes all kinds of aviation investigations for all types of Polish Armed Forces.
One of the regular ITWL tasks is monitoring of flight load on all kinds of aircrafts.
No major changes are going to be introduced in the training course for TC-II ver-
sion. Hence data gathered from operation of TC-I version of the training aircraft,
which has been recorded throughout many years, were the basis to determine the
severity of flight load spectrum characteristic for Polish Air Force and furthermore
to determine the test flights schedule. The test flights were performed by military
pilots according to a precisely scheduled program which included the most common,
Development of Load Spectrum for Full Scale Fatigue Test of a Trainer Aircraft 575

statistically determined, types of sorties. Basing on data recorded during test flights a
characteristic flight profile and the test load sequence were determined.
The contractor of the Full Scale fatigue Test is the VZLU Praha in Czech Re-
public. After finalization of the test a Teardown Inspection is scheduled.

3 Flight Program Analysis

All the aircrafts operated in Polish Air Force were equipped with digital flight re-
corders. The main objective of aircraft instrumentation was to analyze the training
process. Unfortunately no usage monitoring system, based on the gathered data,
was initially developed. Thanks to ITWL flight data from over 40000 flight hours
were gathered and could be used for preliminary flight load spectrum determina-
tion and analysis [1]. These spectra are represented by nz vs time characteristic.
The first problem that had to be solved during the recorded data analysis was to
determine time of takeoff and landing. The recorder was not able to show whether
the airplane was on ground or in the air. Solution to this issue was developed by
analyzing all the available flight parameters, e.q. velocity, height, engine RPM etc.
This enabled separation of ground and airborne states.
The main parameter taken into consideration in analysis was the Nz factor (g
factor). A detailed statistic of Nz exceedances was determined.
Since PZL-130 "Orlik" aircrafts are used for pilot training most of the flights
are sorties from training program. The number of the sortie that is going to be per-
formed during each flight is given to the recorded before each flight during de-
briefing. This additional information was important to determine the number of
performed types of sorties per statistical hour of flight. During the analysis of re-
corded data it appeared that many flights (about 30%) have the sortie type number
entered incorrectly. In majority those flight were performed not during the regular
flight school program, including display flying which are significantly more se-
vere than sorties performed during training.
Example results of analysis described above are presented in the table 1 below.

Table 1 Percentage of distinguished flight phases [1].

Basic Intermediate
Loiter Spins Route flights Other
maneuvers maneuvers
21% 13% 8% 4% 23% 31%

The mean number of landings, per hour of flight, was determined:

landings
nl = 2.31
fth (1)
More touch and go landings were recorded than those ending with full stop.
Among the flight phases mentioned above spins and stalling cause severe loads
to the structure. Those maneuvers are not distinguished in the Nz factor analysis.
Hence in order to determine how often are they performed per statistical hour of
576 A. Leski, P. Reymer, and M. Kurdelski

flight the training program sorties definition along with sorties statistics were
used. Mean number of spins per hour of flight was estimated to be equal 0.16.

4 Flight Test Program

Instrumentation

In order to determine the actual loads acting upon aircrafts structure during flight
and on-ground manoeuvres a series of carefully designed test flights were per-
formed by trained military pilots. One of the PZL-130 "Orlik" TC-II trainer air-
crafts operated by Polish Air Force was designated for flight tests purpose
(No.37). The design and installation of measurement system was carried out by
AFIT with close cooperation with EADS PZL-Okęcie.
Overall thirteen measurement sections were chosen in which 86 measurement
points were installed (due to required redundancy). The total number of measured
loads was equal to 27. Three sections were determined in each wing where bending
moment, and shear force were measured. In addition torque in wing was measured
in section 1. In horizontal stabilizer as well as in fin bending moment along X axis
and shear force were measured in only one section each. Moreover there were two
fuselage sections (Frame 1 and Frame 9) where bending moments along X and Y
axis were measured (in addition torque in Section 9). Two last sections were located
on the main gear, where bending moments along X and Y axis were monitored.

Fig. 1 Location of measurement sections [2].

Strain gauge configurations used are shown on Fig. 2. Due to access problems
mainly Poisson's half bridges were used with full temperature compensation. For
strain measurement in the main spar due to bending separate channels were desig-
nated for the upper and lower flange.
Development of Load Spectrum for Full Scale Fatigue Test of a Trainer Aircraft 577

a)

b)

c)

Fig. 2 Strain gauge configurations for different loads (a) bending moment/tension b) torque
c) shear force ) [3].

a) b)

Fig. 3 Aircraft instrumentation. (a) strain gauges (b) recorder modules [2].
578 A. Leski, P. Reymer, and M. Kurdelski

For majority of the measurement sections main and backup strain gauges were
installed. High level of redundancy was applied to maximize the reliability of the
measurements. Measurements and recording were performed with KAM-500 re-
corder. During majority of flights sampling frequency of 25 Hz was used. For re-
cording spins and stalling the sampling frequency was set to 400 Hz for strain
gauges located in the empennage region.
In addition the KAM-500 recorder was gathering signals from accelerometer
and flight parameters, like velocity or pitch/yaw/roll angles, from the onboard
flight recorder. A laser rangefinder was attached to the left wings lower surface in
order to measure descent speed. Signal from this device was also recorded in
KAM-500. Exactly the same strain measurement array will be installed by VZLU
on the Full Scale Fatigue Test specimen (No.15) according to detailed technical
documentation [2].

Calibration

After instrumentation and preliminary flight, performed in order to eliminate pos-

sible hysteresis in the strain gauges output signals, calibration was carried out. The
idea of this process is to exert known loads upon the aircraft's structure and to si-
multaneously record the response strain in measurement points [4]. Loads were
applied by means of jacks and belts tensioned with belt stretchers and monitored
with dynamometers. To distribute the loads evenly and make sure to introduce
them in the ribs surfaces as well as to prevent possible dent and damage to the
structure a number of specially designed clamps were manufactured for the cali-
bration purpose only. Figure 4 shows how loads were applied to the wings and fin.
Recorded signals from right wing are presented on the figure 5.

Fig. 4 Clamps and load application during calibration (wing and fin) [5].
Development of Load Spectrum for Full Scale Fatigue Test of a Trainer Aircraft 579

Fig. 5 Calibration of the right wing gauges. Systems response to upward bending moment
caused by force F4 [5].

Data recorded during calibration were carefully analyzed in order to verify that
the strain gauges work correctly and react to the applied loads in the predicted
manner. Changes to the measurement system could be introduced only at this
stage, since after any modification the calibration process should be repeated what
would be impossible after forwarding the airplane to the Air Force.

Experimental flights

Flight loads were recorded in three phases [6]. In phases I and II the aircraft was
fully instrumented as shown on Fig. 1. For phase III, which is meant to last till the
aircraft is withdrawn from operation, only 8 most crucial strain gauges were left
(two in each wing, one in left part of the horizontal stabilizer and fin and two in
the tail boom). The phases are described below:
Phase I – experimental flights performed according to detailed program,
Phase II – routine flights performed by the user (autonomous, full instrumentation),
Phase III – continuous recording (autonomous, reduced instrumentation).
Test flights with full instrumentation were performed according to carefully de-
veloped flight program (Phase I). During definition of the program most care was
taken to include all the characteristic elements that occur in the regular operation.
Sixteen flights were planned and performed. Among scheduled flights we can dis-
tinguish: performance flights, mixed sorties flights and two specially designed for
spin and stalling loads monitoring. The flights were carried out throughout sum-
mer in 2010 by a professional Air Force instructor pilots.
In order to determine the following flight stages a special marker signal was in-
troduced which could be triggered by the pilot. This signal was recognized by
the KAM-500 recorder and enabled to determine the beginning and end of
each planned maneuver. This signal had greatly helped in interpretation of ob-
tained results.
580 A. Leski, P. Reymer, and M. Kurdelski

Phase II was recorded in automatic mode. The recording started and ended
autonomously without pilots interference. The preliminary SEWST program
schedule assumed long term recording in Phase II. Unfortunately from reasons be-
yond the Air Force Institute of technology this phase has been strongly reduced.
Four flights were recorded (1 flight route and 3 display flights).
Phase III is focused on constant signal recording from a group of selected strain
gauges during regular operation. A substantial part of the measuring system was
removed before phase III started. The results obtained in Phase III were not used
for Full Scale Fatigue Test load spectrum determination. The carried out research
proved the constructors prediction, that significant vibration occurs just before
stalling in the empennage area (Fig.6).

Fig. 6 Flatter in empennage preceding stalling. Presented signal corresponds to the fin's
bending moment.

Since these vibrations are characterized by relative high frequency, they cannot
be implemented in the standard block of frequency 0,5 Hz. Therefore a separate
block was created in which all the jacks, except ones in the tail section, will be
stopped at determined neutral position while fin and horizontal stabilizer will be
loaded with higher frequency. This will enable to determine fatigue damage that
may occur in the rear section due to recorded vibration.

5 Development of Load Sequence

Regression equations

During the test flights signals from all the installed strain gauges were recorded.
After completition of flights in Phase I and II a preliminary signal selection was
Development of Load Spectrum for Full Scale Fatigue Test of a Trainer Aircraft 581

done taking into account their performance. The main cryterion of selection was
the stability and repeatability.
Linear regression equations determining flight loads by means of registered
strains were developed on the basis of data recorded during calibration [7]. The
equations were determined using advanced linear regression module in STATIS-
TICA (GRM). In each case the load applied during calibration was chosen as the
dependant variable and the corresponding strain signals as the independedt ones.
In the first approach signals from all the strain gauges present in the analyzed
section were taken into consideration and with use of the best set method and
mean square root method the signal with the best correlation was chosen. Due to
the chosen calibration methodology only one strain gauge was chosen for deter-
minining each load. However equations for the remaining strain signals corre-
sponding to the analyzed load were determined as well for comparison purposes.
Since high level of redundancy was one of the main assumptions it was possible to
choose for each load the strain signal with the highest level of correlation with
high level of excitation at the same time.
Secondly the linear regresion equations coefficents were determined. The ob-
tained value of loads were compared with the recorded ones using statistical meth-
ods available in the software. This was an additional source of information to
determine whether chosen signals describe the loads correctly within te whole
range. Below some of the results used in equation verification are presented.
Figure 8 shows how the determined linear regression equation (represented
with the trend line) corresponds to the measured values of load (depicted with
markers along the line).

Fig. 7 Expected versus observed values of MxSL1 bending moment [7].

582 A. Leski, P. Reymer, and M. Kurdelski

Figure 9 shows a histogram of raw residuals which depicts how the obtained
values differ from the measured ones. Since most of the samples are concentrated
near the zero value the obtained results seem to be representative and the achieved
difference is within reasoneable range.

Fig. 8 Hiatogram of raw residuals [7].

Data filtering

The raw data obtained from flight tests needed to be verified for any data loss or
strain gauge malfunctions. Since a high level of redundancy was applied it was
possible to compare obtained load values from different gauges in order to check
whether they are similar and correspond to manufacturers predictions for given
values of Nz. After assuring that signals from the chosen set of parameters is reli-
able throughout the whole research further post processing was undertaken. The
goal was to determine characteristic flight load spectrum for approximately 200
Simulated Flight Hours. The duration of load block during the fatigue test was
also determined within reasonably boundaries. Hence data reduction leading to
achieving about 150 load lines per hour of flight had to be done.
Firstly the peak and valleys values were determined using numerical algorithm.
Although 27 loads were monitored not all had to be taken into consideration
for extremes definition, since generally shear forces extremes will correspond to
the bending moments in common sections. This allowed to achieve further data
reduction.
Development of Load Spectrum for Full Scale Fatigue Test of a Trainer Aircraft 583

Secondly filtering of amplitude of 10% was performed. Since higher magnitude

cycles introduce relatively more fatigue damage to the structure, the low ampli-
tude cycles could (and had to due to block restrictions) be neglected.

6 Summary
As a result of the described process a load history from each flight was obtained.
Since the final load spectrum corresponds to 200 simulated flight hours the load
block had to be constructed by repeating the recorded flights by a determined
number of times (resulting from the statistical analysis of the historical data) [8].
Secondly the landing loads and ground maneuver loads were input into the
block to simulate the landings between flights (as mentioned above 2,31 landings
per hour of flight). The buffeting loads were defined separately due to high
frequency.
The final load block will have to be additionally tested before the actual Full
Scale Fatigue Test. During the preliminary test the strain measurement array in-
stalled on the test specimen will also be verified. Since it was installed according
to the detailed technical documentation [2] the strain signals recorded during the
fatigue test will be used to verify the exerted loads.
Finally the loads within the block were ordered in such way, that it is expected
to determine fatigue markers within cracks during Teardown Inspection. It is be-
lieved, and was already discussed [9], that due to such markers it can be possible
to determine when the crack initiated and how fast was it propagating.
The whole SEWST program, after a year, is still in it's preliminary stage. The
presented research concerning development of the load spectrum is just the begin-
ning of the whole program. Further results will be presented in time, followed by
discussion of the results.

References
[1] PZL-130 Orlik Flight school program analysis for determiantion of flight spectrum
severity, Report No. 134/31/2010, Air Force Institute of Technology, Warsaw, Poland
[2] Technical documentation of the strain gauge instalaltion on PZL-130 Orlik TC-II
trainer aircraft number 037, Report No. 89/31/2010, Air Force Institute of Technol-
ogy, Warsaw, Poland
[3] Kottkamp, E., Wilhelm, H., Kohl, D.: Strain gauge measurements on aircraft, NATO
Advisory Group for Aerospace Research and Development. AGARDograph 7(160)
(April 1976)
[4] Skopionski, T.H., Aiken Jr., W.S., Huston, W.B.: Calibration of strain-gage installa-
tions in aircraft structures for the measurement of flight loads, NACA report nr 1178
[5] Calibration of the strain gauges installed on PZL-130TC-II Orlik trainer aircraft num-
ber 037, Report No. 127/31/2010, Air Force Institute of Technology, Warsaw, Poland
[6] Development of the flight load measurement procedure for the PZL-130 Orlik TC-II
trainer aircraft, Report No. 57/31/2010, Air Force Institute of Technology, Warsaw,
Poland
584 A. Leski, P. Reymer, and M. Kurdelski

[7] Development of the approximation equations for determining inner loads within PZL-
130 "Orlik" TCII aircraft structure on the basis of recorded strain, Report No.
164/31/2010, Air Force Institute of technology, Warsaw, Poland
[8] Leski, A.: An Algorithm of Selecting a Representative Load Sequence for a Trainer.
In: 2nd International Conference on Engineering Optimization. Lisbon Portugal. CD,
pp.1-8 (September 6-9, 2010), http://www.engopt2010.org/
[9] Molent, L., Barter, S.A., White, P., Dixon, B.: Damage tolerance demonstration test-
ing for the Australian F/A-18. International Journal of Fatigue 31, 1031–1038 (2009)
[10] Air Force MIL-STD-1530C Aircraft Structural Integrity Program (ASIP)
[11] Anderson, I.A., Parker, R.G.: Full Scale Fatigue Test of the Pilatus PC9/A trainer air-
craft. In: 20th ICAF Symposium, Bellevue, Washington, USA, July 14-16 (1999)
[12] Klimaszewski, S., Leski, A., Zurek, J.: The Role of AFIT in the Polish Aging Military
Aircraft Programs. In: Proceedings of 7th Joint FAA/DoD/NASA Conference on Ag-
ing Aircraft, New Orleans, September 9-11 (2003)
[13] Leski, A., Klimaszewski, S., Kurdelski, M.: The Assessment of Fatigue-Life Re-
sources of the PZL-130 Orlik’s Structure. In: Sixth DSTO International Conference of
Health & Usage Monitoring, Melbourne, Australia, March 6-9 (2009)