Onboard, Real-time Loads Monitoring Using the Loads Observer – Practical Implementation Aspects

International Forum on Aeroelasticity and Structural Dynamics (IFASD) 2009, Seattle, USA, June 21 – 25, 2009

M. Augello1, L. Bensch2, J. Jusseit3, H. Henrichfreise4

Airbus Operations GmbH
21129 Hamburg, Germany
e-mail: 1michael.augello@airbus.com and 2lars.bensch@airbus.com

DMecS Development of Mechatronic Systems GmbH & Co. KG
51105 Cologne, Germany
e-mail: 3juergen.jusseit@dmecs.de

Cologne University of Applied Sciences, Cologne Laboratory of Mechatronics
50679 Cologne, Germany
e-mail: 4hermann.henrichfreise@clm-online.de


Nowadays, Loads Monitoring (LM) of aircraft structures plays an important role not
only in military, but also in civil aviation with regard to maintenance and inspection activities.
It is integrated in numerous applications like fleet wide monitoring and leads to a reduction of
aircraft on-ground time after an in-flight incident. This paper describes the process taken to
adapt, integrate and test the LM algorithm “Loads Observer” on a prototype system within the
framework of the European AWIATOR (Aircraft WIng with Advanced Technology OpeRation)
research program. In this context, the algorithm was implemented on real-time hardware,
which was installed onboard a flying testbed and fed with signals from the aircraft’s
flight test data acquisition unit. For the real-time operation of the system, the pre-processing
of the aircraft’s input signals was enhanced enabling the on-line synchronization and interpolation
of the data. The pre-processed input signals were provided to the Loads Observer algorithm
for the simulation of the aircraft’s motion including the flexible deformation, thus allowing
the reconstruction of the load distribution on the airframe. The flying testbed,
equipped with strain gauges at dedicated stations, provided loads measurements, which were
used for the validation of the Loads Observer algorithm. In this paper, special focus is put on
the presentation of the deployed rapid prototyping process, which allowed a fast evolution of
the LM algorithm.

Download Publication:

DMecS_LoadsMon_RTImplementation.pdf (382 KB)

1 Introduction

New aircraft developments reveal a strong need for systems enabling the operator to react
faster and more efficiently to events, which possibly ground the aircraft for the time of a required
inspection. In this context, the introduction of techniques that assess the structural
health of the airframe are investigated in order to ensure the safety of the passengers and reduce
the operational costs of the aircraft. Beside these aspects, the economic benefit in terms
of life time extension, maintenance on demand activities and fleet-wide monitoring capabilities
endorse research in these fields. In the framework of the AWIATOR [1] research project,
the “Loads Observer” LM algorithm was developed and tested on-board of a flying testbed,
allowing the reconstruction of structural loads due to both, gust and maneuver excitation. For
the prototypical implementation of the Loads Observer on-board the test aircraft, state-of-theart
rapid prototyping tools were applied, which shortened the implementation time and set the
focus more on enhancing the maturity of the developed algorithm and the system. Tools like MATLAB/Simulink [2, 3] and dSPACE [4] allowed the setup of an environment for the validation
of the prototype system and helped not only to cut the costs for the prototypical implementation,
but also reduce the overall development time frame. As one of the major benefits,
the presented prototypical implementation quickly enabled the identification of erroneous
system behavior and consequently helped to reduce mal-functions in service, which could
reduce the customer’s confidence. This aspect is important from an economic point of view,
as a rework of the system after introduction causes costs, which are usually a magnitude
higher than the costs needed to achieve a reasonable “development saturation” that guarantees
a mature functionality by entry into service.
This paper outlines the practical implementation aspects, which are encountered during the
integration of the Loads Observer algorithm into an onboard system. Special focus is set on
the enhancement and adaptation of existing design tools in order to allow the autonomous
real-time operation of the system. Further, the tool chain used for the integration of the algorithm
into the onboard system and the corresponding hardware architecture of the prototype
are presented. The tests towards integration on-board the aircraft and the subsequent flighttesting
campaign are outlined. Here, especially the capability to quickly adapt the algorithm
during this last part of the testing phase turned out to be a key factor for a mature system
setup and fast validation progress.

2 LM Algorithm Components

The idea for the development of an onboard loads monitoring system has existed within the
Airbus Loads & Aeroelastics community since the early nineties. The AWIATOR research
project finally provided the framework for the realization of an onboard LM system as originally
intended. For the development of the Loads Observer, a modular approach is chosen. It
integrates and adapts tools, which are already available from the standard design process. The
modules are compiled to an algorithm suiting onboard system requirements and can be divided
up into the four main components:

  • Model based reconstruction of structural loads, i.e. internal shear, bending and torque
    quantities at arbitrary output stations all over the aircraft
  • Correction of the aircraft motion within the Loads Observer model, which deviates
    from the recorded motion signals due to external excitation and/ or a set of model parameters
    not suiting the actual setup of the aircraft
  • Real-time pre-processing of input signals
  • Adaptation of the aircraft model to the current flight state

In the course of this chapter, each of the mentioned modules is discussed in more detail. The
Loads Observer requires only input signals, that belong to the set of signals recorded anyway.
Hence, they are available for every commercial aircraft and consequently do not depend on
information provided by additional sensors or equipment.

2.1 Aircraft model

The physical model for the structural loads reconstruction is formulated with VarLOADS [5],
which allows the incorporation of parameterization data (e.g. aerodynamics, mass model) and
modeling features from existing design models. Its modular setup enables the adaptation of
the model’s complexity to the actual requirements of the target platform, on which the Loads
Observer is implemented. VarLOADS is implemented in a MATLAB/Simulink environment,
which can be directly connected to the rapid prototyping environment used for the Loads Ob3
server development. Furthermore, the usage of a standard tool allows the embedding of already
available and validated design parameterization data, which reflect the mass setup and
aerodynamic properties of the actual aircraft.
Figure 1 shows the structure of the aircraft model, which consists basically of three modules.
The Rigid body and elastic motion module evaluates the rigid body and flexible response due
to external forces acting on the airframe and provides the input for the calculation of the structural
loads at dedicated monitoring stations. The Engine loads module computes the external
loads due to engine setting and introduces them to the corresponding attachment points at the
airframe. The Aerodynamic loads module evaluates external loads due to the aerodynamic
pressure distribution on the airframe, which is influenced by rigid body motion, elastic deformation
and external excitation, e.g. turbulence.

Figure 1: Schematic aircraft model

The simulation model processes the commanded signals, which are a superposition of pilot
and flight control system inputs. The aircraft properties, such as mass distribution, as well as
the environmental conditions contribute significantly to the load level. In the pre-processing
routines, the aircraft parameterization is chosen according to the actual state of the aircraft for
which the algorithm reconstructs the loads. Prior to the simulation, the initial conditions are
determined by evaluation of the trim setting for an arbitrary maneuver. After each simulation
step, the external and internal loads are superimposed and transformed to the specified monitoring
stations, which coincide in the present case with the flight test measurement stations.
Beside the structural loads, the flight path [6] and the elastic behavior are reconstructed in the
course of the simulation and are available for further processing.

2.2 Loads Observer

For commercial aircraft, no sensors are available for the measurement of spatial gusts. However,
the aircraft’s motion is influenced significantly by turbulent atmosphere. As no signals
exist reflecting this external excitation within the input data set, the VarLOADS aircraft
model alone is not capable of achieving a sufficient match with the measured aircraft motion
in the presence of turbulence. Hence, the model is embedded into a non-linear observer [7] as
shown in Figure 2.

Figure 2: Structure of the Loads Observer

During operation, measured control inputs are fed to the aircraft model. Due to the lack of
knowledge about the atmospheric turbulences encountered by the aircraft, the simulation
leads to a different response than that of the real aircraft. The Loads Observer [7, 8] algorithm
uses these residuals for the stabilization of the rigid body motion and for the reconstruction of
gust components, which are applied as additional aerodynamic pressures at dedicated stations
on the airframe. By this means, the motion of the aircraft model is forced to match the measured
one and the influence of the external disturbance is taken into account for the resulting
structural loads. The final output data set of the Loads Observer consists of the estimated gust
components and the reconstructed loads integrated for the specified monitoring stations. Prior
to the implementation into the prototype system, the Loads Observer algorithm was checked
in numerous tests, in which e.g. synthetic gusts were reconstructed.

2.3 Real-time pre-processing of input signals

Before aircraft signals can be provided as input, a pre-processing must be performed, which
adapts the incoming signals according to the interface specification of the Loads Observer.
The algorithm expects the data to be synchronized and equally sampled to the required step
size. During the AWIATOR project, the flight test raw data were provided by the aircraft’s
data acquisition unit in UDP (Universal Data Protocol) format. The signals were recorded
with a different sampling rate and their own time stamp. Consequently, data pre-processing
prior a provision to the Loads Observer algorithm was mandatory. This task was accomplished
by the implementation of a spline interpolator, which ensured common data properties
in real-time. Figure 3 shows the architecture of the pre-processing module. Before being
processed, the incoming raw data set is checked for validity regarding its time stamp, but also
regarding the value of each received signal in reference to a corresponding plausible band
width. Only in case of valid data, the pre-process is continued. As the on-line interpolation
algorithm [9] requires knowledge about the development of the signal time histories, the data
are written together with their respective time vector into a buffer. The sequential reception of
the input data packages leads to a delay between the recorded signals of one time step. Therefore,
the data are synchronized with respect to the timer value of the pre-processing module.

Figure 3: Architecture of the pre-processing module

The buffered data is passed on to the spline interpolator, which performs the smoothing and
re-sampling of the input data for the Loads Observer. The interpolation and re-sampling process
is depicted exemplarily for one input signal in Figure 4. Here, a short sequence of one of
the input signals is shown. The data arrives sampled with a low frequency and is stored in the
buffer. If the sensor step size is bigger than the recording step size of the data acquisition unit,
the sample and hold mechanism leads to the recording of redundant data points, which are
represented by white squares in the left sub-figure. The remaining data points