• Home
• Conference Presentations
• Preliminary Program
• General Information
• Program at Glance
• Symposium Topics
• Plantema Memorial Lecture
• Jaap Schijve Award
• Best Paper Award
• Paper Submission
• ICAF Delegates
• Important Dates
• Venue
• Sponsors and Exhibition
• Registration
• Hotel Accommodation
• Contact

13:30   Structural Health and Structural Loads Monitoring I Chair: Antoni Niepokolczycki 13:30 30 mins PILATUS PC-21 – A DAMAGE TOLERANT AIRCRAFT Lukas Schmid Abstract: The PC-21 Advanced Trainer was designed and certified as a damage tolerant aircraft. The PC-21 is certified to the acrobatic category of FAR 23 Amendment 23-54. Some major certification aspects are addressed in this paper. The design fatigue spectrum was created based on pilot input in a deterministic manner. The distribution of vertical acceleration was shown to result in fatigue damage similar to FALSTAFF. The aircraft inspection intervals were defined by crack growth analyses, whereas results of the FSFT were taken into account. The Full Scale Fatigue Test (FSFT) was performed by considering a life scatter factor of 3. First, two lives of durability testing were performed. After introduction of artificial damages another life of damage tolerant testing was run. Finally, a residual strength test campaign was carried out. The certified Fatigue Monitoring System (FMS) allows monitoring the loading history of major structural assemblies by means of strain sensors. Several Fatigue Indices are calculated from the strain signals on a flight-by-flight basis for Individual Aircraft Tracking purposes (IAT). 14:00 30 mins INTRODUCTION TO SERVICE OF AN ARTIFICIAL NEURAL NETWORK BASED FATIGUE MONITORING SYSTEM Steve Reed, Brian McCoubrey, Andy Mountfort Abstract: Individual aircraft fatigue and usage monitoring is an essential element of structural integrity assurance. This is particularly significant in the military environment where the scatter in usage between individual aircraft within a fleet and the difference between design-assumed usage and in-Service usage can both be highly significant. In the UK military, substantiation of the individual aircraft fatigue monitoring or tracking system is undertaken using an Operational Loads Measurement (OLM) programme. Within these programmes, a proportion of the fleet is fitted with a range of strain gauge sensors and associated parametric sensors. One of the key outputs of such programmes is an assessment of the performance of the individual aircraft fatigue monitoring system. Many legacy military aircraft have relatively simple individual aircraft fatigue monitoring systems based upon, for example, exceedence counts from a normal accelerometer (Nz). These data are coupled with additional information, such as aircraft mass and store configuration and some assumptions of usage, such as point-in-the-sky dynamic pressures. This information is combined within fatigue meter formulae to give an indication of fatigue consumption, for Nz-driven critical features, usually on a sortie-by-sortie basis. Substantiation of these systems using OLM data is usually based upon average performance of the fatigue meter formulae in comparison with the OLM strain-based data. It is usual for corrections to be applied to these formulae as a result of this process; often these corrections can be very significant. Sufficiently accurate monitoring of in-Service fatigue life is essential from both a safety and cost of ownership perspective and methods of improving the accuracy of apportioning fatigue life consumption on an individual aircraft basis have significant implications for fatigue management. However, obtaining funding for improvements to equipment in individual aircraft fatigue monitoring systems for legacy aircraft is unlikely to be successful due to higher-priority funding requirements. This paper describes an alternative, cost-effective approach to obtaining a significant increase in individual aircraft fatigue monitoring accuracy, using legacy equipment and ride-along data from an existing OLM programme. In this process, artificial neural networks are used to determine the relationships between input parameters from the legacy equipment, such as Nz counts and the flight-by-flight fatigue damage calculated from strain data captured during the OLM programme. Neural networks are able to determine generalised, non-linear relationships between data in a controlled environment and present an attractive solution for multi-dimensional regression-type problems. Within this paper, the development, verification and validation of the Structural Health and Usage Neural Network (SHAUNN) based fatigue meter formulae for 2 critical wing features for the Tucano TMk1 military trainer is described. The training, testing and independent validation environment used in the SHAUNN framework, using nearly 900 sorties of OLM data from 3 aircraft, is explained. Emphasis is placed upon understanding the response of SHAUNN to data outside of its learning experience and establishing processes to cater for such eventualities. Additionally, processes for the ongoing monitoring of the continued performance of the SHAUNN monitor in Service are illustrated. 14:30 30 mins TOWARDS AUTOMATED FLIGHT-MANEUVER-SPECIFIC FATIGUE ANALYSIS Juha Jylha, Marja Ruotsalainen, Tuomo Salonen, Harri Janhunen, Tomi Viitanen, Juho Vihonen, Ari Visa Abstract: Introduction Structural health monitoring of aircraft is a necessary topic especially with aging fleets. In general, it means tracking the structural integrity and keeping the risk of hazardous cracks low. There exist several approaches which operate with different accuracy and scope. Our study relates to F-18 aircraft structural integrity program of Finnish Air Force. Previously, a parameter based fatigue life analysis scheme has been developed in Finland [1, 2]. It provides location and flight specific fatigue life expenditure (FLE) estimates supplementing the fatigue monitoring system of the original equipment manufacturer. This paper discusses the determination of the FLEs of individual flight maneuvers. Manual flight maneuver identification (FMI) is an exhausting task [3]. Recordings from a few hundreds of flights constitute an extensive database implying the use of data mining methods. To expand the earlier work [1, 2, 3], we propose an automated FMI procedure based on the flight parameters. Good applicability to different maneuvers is verified by comparing the results of our automated procedure with those identified manually by an experienced analyst with pilot background. Incorporated with the ability to assess the fatigue life expenditure, the proposed approach allows fatigue tracking inside flight missions of an individual aircraft. The approach is aimed to support pilot training as well as mission planning. Fundamentally, any life expenditure which is not justified by operational or training objectives needs to be excluded. Presented FLE distributions of different flight maneuvers using one hundred flights illustrate the potential of our procedure. Flight maneuver identification procedure The identification process comprises three steps: choosing, modeling, and identifying maneuvers. In practice, we apply so-called model-based pattern recognition here. For every maneuver to be identified, we build a model that we compare with flight parameter data––in other words, we measure the similarity of the model to patterns that exist in the data. As the first step of the process, a representative maneuver is chosen from the flight parameter recordings by an experienced analyst. In practice, many things must be considered here; the abstraction level (time scale) of the maneuvers as well as their sensibility and significance especially from the fatigue life analysis point of view. These kinds of questions demand knowledge about the use, behavior, and structures of the aircraft. The models are built based on the chosen representative maneuvers. Because the amount of available flight parameters is high and only a few of them are relevant in modeling a specific maneuver, expertise is required in choosing the most descriptive flight parameters. For measuring the similarity, we quantize the chosen parameters to three levels to enable a consistent treatment of different parameters. Thus, the quantized, chosen flight parameters within the representative maneuver form the model to be used in the comparison. The last step of the process is fully automatic maneuver identification. In this step, an identification algorithm based on the built models is used to detect maneuvers from the unanalyzed flight parameter recordings. In reality, the same maneuvers can be performed in slightly different ways, and their duration can vary. To cope with this, we use the dynamic time warping (DTW) algorithm to handle the temporal variations. A DTW matrix is calculated between the modeled representative maneuver and the whole flight using the quantized flight parameters that were chosen at the modeling step. Beginnings and ends of the maneuvers are detected from the DTW matrix which provides a similarity values for the patterns. Results Our automated FMI makes it possible to calculate FLE distributions at the maneuver level. The FMI method has been tested to identify maneuvers from approximately 100 flights. The test material provides us good quality flight parameter and strain gauge data. In the first test, three maneuvers (symmetric pull, roll + symmetric pull, and roll) were searched and their damage was calculated for a critical location in the F-18 wing fold. Figure 1 shows the relative maximum and median damage of the maneuvers. In the second test, split-s maneuvers were indentified from the flight set. The FMI method found 18 clean split-s incidents. Figure 2 illustrates relative damage produced by the found maneuvers to the vertical tail stub. The result shows that the split-s causes diverse damage depending on the flying style. Hence, a pilot can be instructed to perform the split-s maneuvers in a fatigue-friendly way. For this specific maneuver, the angle of attack is the key factor in causing the damage. Maneuver-specific FLE distributions that have high variance are interesting, because it might be possible to avoid performing the related maneuvers in the way that consumes FLE the most. Particularly, savings may be substantial when reducing the average FLE of certain frequent maneuvers. [1] J. Tikka, T. Salonen (Patria Aviation, Finland), Parameter based fatigue life analysis for F-18 aircraft, 24th ICAF Symposium, Naples, 16–18 May 2007. [2] J. Jylhä, J. Vihonen, T. Ala-Kleemola, R. Kerminen, A. Visa (Tampere University of Technology, Finland), J. Tikka (Patria Aviation, Finland), Modeling structural vibrations for automated aircraft fatigue monitoring, 24th ICAF Symposium, Naples, 16–18 May 2007. [3] A. Siljander (VTT Technical Research Centre of Finland), A Review of aeronautical fatigue investigations in Finland, 30th ICAF Conference, National Review 13, Naples, 14–15 May 2007.