AP-3C Orion Advanced Flight Simulator The Challenge of Performing a Flight Test Data Acquisition and Modelling Program for the First Time

Transcription

1 AP-3C Orion Advanced Flight Simulator The Challenge of Performing a Flight Test Data Acquisition and Modelling Program for the First Time Darren Jones James Keirnan Raytheon Systems Company Australia Building 2, Level 3, 14 Aquatic Drive FRENCHS FOREST NSW jonesdj@raytheon.com.au jkeirnan@raytheon.com.au Keywords: AP-3C, Advanced Flight Simulator, flight test data, modelling ABSTRACT: The Royal Australian Air Force (RAAF) has or is in the process of acquiring a number of high fidelity full-flight simulators. One such device is the Advanced Flight Simulator (AFS) for the AP-3C Orion aircraft. Given that the original P-3A aircraft was developed from the Lockheed Electra in the late 1950 s and early 60 s, and the aerodynamic configuration of the RAAF aircraft is unique, it was recognized that the availability of appropriate and accurate modelling data for development of the AFS would be extremely limited. Furthermore, the Operational Flight Trainer (OFT) currently used for training by the RAAF was assessed as exhibiting poor fidelity in a number of critical operating regimes. It was therefore resolved to acquire a new set of aerodynamic, controls, systems, performance and ground handling data with which a full set of AFS models could be fully developed. Commonwealth ownership of the raw test data such that future AFS development and/or modification could be readily performed was seen as a further benefit of this approach. The paper will describe the process that was employed to acquire the necessary data, and the challenges that presented themselves both throughout that program, and during the subsequent effort to use the data for modelling purposes. It will detail how the aircraft parameter requirements were devised, how an instrumentation system was developed, and how the test data requirements were planned and acquired, all with minimal guidance from a simulator manufacturer. It will then describe how the imperative to have the test program completed as quickly as possible in some instances impacted on the fidelity, amount and type of data that was acquired for model development and checkout, and how that impact had follow-on implications during modelling. 1. Introduction 1.1 Background The Royal Australian Air Force (RAAF) has a requirement to extend the life-of-type of their P-3C fleet to beyond To achieve this, a reduction in the current airframe fatigue accrual rate is required. A three-prong strategy has been implemented to achieve this, namely: a) reduce the empty mass of the aircraft, b) purchase Training AP-3 (TAP-3) aircraft from the USN to reduce the requirement to perform fatigue accruing training sequences on operational airframes, and c) acquire a level 5 simulator to provide high fidelity flight station crew training, and thereby reduce the amount of actual aircraft time required, especially in fatiguing and high risk flight regimes. This paper discusses the program that was implemented to acquire flight test data suitable for development of the AP-3C Advanced Flight Simulator (AFS). It further discusses how that data was used in the development of a model set suitable for integration into the AFS. 1.2 History The current RAAF Operational Flight Trainer (OFT) is based on a USN P-3B device. The RAAF assessed that the poor fidelity exhibited by the device in a number of critical operating regimes, most notably ground handling, provided negative training in those regimes. Continued development of the RAAF P-3C fleet has also resulted in a unique aerodynamic configuration for the aircraft, which was not replicated in the OFT. Due to the age of the aircraft design, the availability of credible data from which a simulator could be developed was limited. As a high fidelity simulator with which as much training as possible could be performed was required, it was determined that a complete set of aerodynamic, performance, controls, systems and ground reaction data would need to be acquired. It was further concluded that collection of such data was on the critical path for AFS development. Therefore, a strategy whereby the flight test data requirement was separated from actual AFS acquisition was implemented. 2. Data Acquisition 2.1 Flight Test Support Contract A Flight Test Support Contract was struck between the RAAF P3AFS Project Office and Raytheon Systems Company Australia (RSCA) in May That contract

2 was based around acquiring data for a tailored list of parameters developed from the International Air Transport Association s (IATA) Flight Simulator Design & Performance Data Requirements (FSD&PDR) document (Reference 1), supplemented with the International Civil Aviation Organization s Manual of Criteria for the Qualification of Flight Simulators (MCQFS) (Reference 2). In all, over 100 parameters were identified in that list. 2.2 Aircraft Modification An instrumentation system was developed to enable data to be acquired to the agreed specification for each parameter throughout the operational envelope of the aircraft. The modifications to install this system were developed in two parts, a permanent modification (principally wiring and mounting provisions) on to which the sensors and data acquisition system (DAS) could be mounted, and temporary modifications for installation of the sensors and DAS. This strategy was devised to provide for ready re-instrumentation of the aircraft should further flight test data be required once the AFS had been commissioned. The temporary modifications included such sensors as sensitive steadystate and vibration accelerometers, potentiometers to measure displacement of controls, surfaces and struts, and a video recording system. A design anomaly with the existing P-3C pitot-static system also meant that a stand-alone flight test air data sensing system was required. This was achieved by designing an instrumented flight test boom to be installed on the outermost hard-point of the right wing, approximately 1½ chord lengths in front of the wing leading edge as shown in Figure 1. The boom was used to sense free field pressures and temperature, and to measure aircraft angle of attack and sideslip. Figure 1: Boom Mounted on P-3C Aircraft The age of the aircraft design made tapping off of signals from existing aircraft indication systems (to keep aircraft modification requirements to a minimum) a challenge in many instances. The schemes that had to be employed for many of the aircraft indication systems, for instance fuel flow and engine power, were such that innovative approaches to parameter measurement were often required. In some circumstances, direct measurement of system indications was not feasible and work-arounds, such as data derivation from more than one input, was necessary. While development of the modifications and procurement of hardware necessarily took place prior to contracting of the AFS manufacturer, it was expected that the manufacturer would be identified and provide input regarding their data requirements prior to finalisation of the modifications and commencement of aircraft installation. However, a contract with the AFS Supplier was eventually not signed until after the Flight Test Program (FTP) had commenced. Consequently, the Commonwealth, in agreeing the data set that was to be acquired and the accuracy and sample rates for each parameter, assumed a significant degree of risk. Development and installation of the instrumentation modifications was further complicated by a requirement to concurrently develop and install a flight loads instrumentation package (which incorporated over 100 strain gauge bridges). This requirement arose late in the program, and entailed acquisition of loads related data to assist in management of the airframe fatigue life. It resulted in a delay in commencement of the AFS test program of approximately 2 months. 2.3 Flight Test Program In a similar manner to aircraft modification, circumstances dictated that the FTP for acquisition of all ground and flight test data necessary for AFS development be composed in isolation from the AFS Supplier. Although RSCA had significant experience with developmental and certification flight test programs, this was the Company s first experience in simulator data acquisition. Consequently, the test program was developed principally from guidance provided in References 1 and 2. Those references characterized data into two basic categories, namely: design / modelling data, and checkout / validation / proof-of-match. While both documents provided reasonable detail regarding validation data requirements, little guidance was provided on the modelling data necessary for development of a Level 5 simulator. Modelling data is naturally the most important in terms of achieving a high level of model fidelity. The tables in Reference 1 that detail the requirements for such data are structured around simulator data packs compiled by larger aircraft manufacturers (examples of Boeing and Airbus data formats are provided in that document). Such manufacturers typically merge data from a number of sources, including wind tunnel and engineering development simulators as well as flight test, to produce a simulator data pack. One requirement in Reference 1, which obviously cannot be satisfied through flight test alone, was to provide tail-off longitudinal stability data. For a program such as the AP-3C AFS, interpreting the data requirements tables to establish the exact type of data being mandated was difficult. Development of a simulator FTP for the first time provided a further challenge in interpretation of the IATA data requirements. While most of the classic flight test techniques are typically used in acquiring simulator data, more

3 unconventional techniques are also employed. Further complication is introduced by variation in actual technique used between model development agencies. For instance, while one such agency preferred a control input when performing aircraft axis excitation manoeuvres, another used a input. Given that little guidance was forthcoming from the AFS manufacturer on their data requirements until the FTP was well underway, further risk was assumed by the Commonwealth in the technique that was selected by RSCA. To compensate for delays in contracting with the AFS Supplier, the FTP was split into 3 phases. The Phase I program was used to undertake all safety of flight testing to ensure that all instrumentation had been satisfactorily integrated, and that suitable quality data was being acquired throughout the flight envelope. The Phase II program was developed to include the majority of the required modelling data such that basic model development could be initiated. That program was centred around testing at a mid Centre of Gravity (CG) loading condition, with a limited test set being repeated at both forward and aft CG conditions. The Phase II program was commenced in early Jul 98, and completed in early Oct 98. A total of 52 sorties and 139 flight hours were required to complete the Phase II program. Due to the delays in program progress previously discussed, the decision was taken to accelerate test conduct part way through Phase II testing. Consequently, a second crew was introduced for test operations such that two flights per day could be achieved. However, in reality only a modest increase in test achievement rate was realized, mainly due to aircraft serviceability and weather related issues. Maintaining consistency of test technique between the crews was also problematic. The Phase III program was commenced in early Mar 99, after completion of the Flight Loads Test Program and an aircraft scheduled servicing. That program consisted mainly of checkout and validation data testing and tests entailing a greater degree of risk, including edge of the envelope manoeuvres (such as stalls, V MCA, V MCG and ground effects). It was finalised in late May 99, and required 5 ground sorties and 45 flight sorties for a total of 126 flight hours to complete. Some of the more challenging flying, and deployment to a number of airfields around Australia to find optimum test operating conditions was necessary during Phase III. A picture of ground effect testing being performed at Woomera is shown in Figure 2. For each sortie a DAS Operator / Flight Test Engineer (FTE) would occupy the DAS station. That person was responsible for; ensuring that the aircraft had been established at the required test conditions, recording data during the test manoeuvre, and reviewing that data airborne to assess whether the test objectives had been achieved. The DAS had been developed to allow display of up to 20 bar (slider) chart parameters and six strip chart (time history) outputs. Display of derived as well as direct measurement parameters was possible. DAS displays were typically set up to show the primary channels of interest for the manoeuvre under consideration as strip charts, and less important channels as slider charts. The time history displays were similar to those shown at Annex A, Figure 2. Figure 2: Ground Effect Testing at Woomera 2.4 Preliminary Data Analysis Post-flight data analysis was performed on a ground station. That analysis consisted of a preliminary quicklook review of all channels to ensure that they were all within their expected operating range and that there were no data drop-outs or spurious artefacts throughout the flight. A typical quick-look display is shown at Annex A, Figure 1. A more detailed analysis of each test manoeuvre was then performed to ensure that each of the critical parameters for that manoeuvre responded as expected, was within the expected signal range and met the contracted accuracy requirements. A typical display (Analyse-9 screen) is shown at Annex A, Figure Model Development 3.1 General A modelling contract with the AFS Supplier was eventually signed in late Nov 98, around six months after flight testing commenced. This meant that a degree of data review catch-up was required prior to commencement of model development. The structure of any flight model used as part of a high fidelity simulator is based on the six classical flight dynamics first order Differential Equations (DE), an example of which is shown at equation 1 below (yaw moment equation). Each equation defines the force or moment along or about three axes. The aerodynamic equations are constructed in a non-dimensional coefficient form, and are effectively computed in the flight simulator in this state. It is important to note that the aerodynamic model equations do not contain any mass or inertia values: these are computed using a separate mass/inertia model. The mass/inertia model may be constructed as a wire frame fuselage representation, with distributed point masses placed at appropriate intervals. The final force or moment value is computed by taking the output of the mass/inertia model and combining it with the relevant force/moment coefficient.

4 C N C N C N b C N b C N = C N0 + δr+ β+ r + p δr β r 2U p 2U C N exampleofa stability derivative coefficient β 3.2 AP-3C Influences The essential difference between a high fidelity and low fidelity model is the mechanization of the stability derivative coefficients for each degree of freedom. In order to conveniently extend a simple first order DE into the non-linear regions required to represent a real aircraft, the stability derivative coefficients are constructed as non-linear functions of certain aerodynamic variables, the most influential of which are typically: Mach number, dynamic pressure, angle-ofattack (AOA), angle of side-slip (AOSS), configuration (flap and gear position), and thrust setting. The stability derivatives are generally implemented as look-up tables, accessed via a linear functional interpolation routine. Most of the stability derivatives are a function of at least three variables. During AP-3C model development, certain stability derivatives were found to be functions of up to five variables. For example: C N / β = function (AOSS, AOA, Flap Position, Rudder Angle, Dynamic Pressure) One of the main objectives of the AP-3C AFS Flight Test Program was to acquire sufficient data to enable high fidelity construction of the stability derivative coefficients, which are the building blocks of the model and are therefore crucial to defining aircraft behaviour. The difficulty in compose a flight test program that will enable characterization of the stability derivatives throughout the aircraft operating envelope is that any number of stability derivatives will be influential to aircraft behaviour for a given flight manoeuvre. The ideal flight test program would be one that isolates the effect of each stability derivative, piece by piece, independently of all others. However, this is only possible, in certain instances, in wind tunnel testing or Computational Fluid Dynamic (CFD) analysis. Nevertheless, flight test still provides the most credible (and arguably accurate) source of data from which to base a flight simulator model. In the case of the AP-3C, it was also the only plausible source available to the Commonwealth. 3.3 Preliminary Model Development In practice, the flight model is constructed in a number of stages, the first being a baseline linear model, which serves as a starting point for later detailed analysis. The baseline linear model for the AP-3C was developed using linear Parameter Identification (PID) techniques (in this case, the Modified Maximum Likelihood Estimator, or MMLE). PID evaluation requires an analysis of aircraft free response to a set of known excitations. For the AP-3C AFS programme, the excitations were provided by a close sequenced series of rudder, aileron and elevator step inputs as depicted in Figure 3. The PID evaluation routine was used to estimate a set of linear stability derivatives based on the eqn 1 spectral response of the aircraft s rotational rates and accelerations to these control inputs. Figure 3: Control Input Sequence The next stage in model development was manoeuvre exploration. For most manoeuvres in a flight test program, a minimum of two derivatives were noted to interact. For example, a rudder step will cause the aircraft to establish a yaw rate, which will then damp out: the aircraft response then enables an exploration of both the rudder power derivative C N / r, and the yaw damping derivative C N / r. Thus, a rudder step was used as the basic manoeuvre to enable evaluation of these two stability derivatives, with the manoeuvre being repeated at a range of configurations, weights, centre-of-gravity locations, Mach numbers, thrust settings and dynamic pressures, to enable all the dependent variables that were likely to influence aircraft behaviour to be investigated. The art of developing a rigorous flight test program is in ensuring that sufficient data is made available to provide for complete definition of all the stability derivatives throughout the flight envelope. The difficulty in the case of the AP-3C was that the AFS Supplier was not identified until the test program was well underway. Certain check manoeuvres are also executed as part of the flight test programme to verify correct implementation of the aircraft dynamic equations for each degree of freedom. For instance, since the Dutch Roll response is affected by all lateral-directional stability derivatives, it is a good basis for assessing the performance of the integrated model after the individual derivatives have been extracted and combined. 3.4 Model Development Challenges The principal challenge in constructing a high fidelity simulator model, based on the flight test data, is working with the accuracy limits of the data. The simulator model is accredited to a level of fidelity, based on a Proof-of -Match (POM) demonstration against a certain standard. For the AP-3C project, the requirements of the UK Manual of Criteria for the Qualification of Flight Simulators (MCQFS) at Level II (which is equivalent to the US FAA Circular B Level D, or Australian FSD1 Level 5) were mandated. These simulator fidelity standards specify

5 the various tolerances on allowable differences between simulator performance and flight test data: various flight parameters must meet certain tolerances, depending on the type of manoeuvre being tested. For example, the roll rate output by a simulator model must be within +/- 2 deg/sec of flight test value for aileron step inputs. If the measured value deviates from the true value by more than the (POM) tolerance, then constructing a model to pass Level II standards is not feasible. An important activity during data acquisition, therefore, is data verification and validation. This effort is aimed at determining whether the true value and measured value of parameter are within specified accuracy bounds and that the data is suitable for use in modelling. Theoretically, after the flight test instrumentation system (FTIS) has been installed, calibrated and tested, it should perform to specification, within the allowed calibration interval. This, unfortunately is not always the case, as systems become unserviceable or drift out of calibration, sometimes in insidious ways. Thus, continuous verification of data against accuracy specifications and validation against modelling requirements must be performed. Otherwise, in the worst case, the data may not be suitable for high fidelity modelling. During the AP-3C project, ATS (now RSCA) were contracted to perform only high level data health and feasibility checks. As data verification was outside of the work scope required by the Commonwealth, it was not performed. Partial validation was accomplished by evaluating both the trim condition and execution of each manoeuvre against the requirements of the agreed flight test technique and any associated IATA Flight Simulator Design & Performance Data Requirements (FSDPDR) requirements. As previously stated, in order to characterize a stability derivative, a certain set of isolating manoeuvres must be executed. If the manoeuvres are not accomplished to within certain tolerances, then use of such data in the modelling process may prove unfeasible (even though the data is within accuracy bounds). Data validation was therefore performed to ensure that data was within the required tolerance levels. 3.5 Data Review and Repair Prior to using the data for modelling, ATS and KSR were require to complete a more rigorous data review process. This process set out to verify accuracy and complete data processing. Data processing involved such steps as filtering/smoothing noisy signals, identifying and removing spurious artefacts and shifting data to remove any definite zero offsets. A significant part of this effort could have avoided if a more data analysis were accomplished as part of the data acquisition process. Data verification was therefore necessarily performed as an initial part of the model development effort. It was a complex process, involving a number of techniques that had to be accomplished in parallel with model development. The principal method used to establish data accuracy for rates and acceleration parameters was to compare static before-flight and after-flight data, to ascertain whether any significant variation had occurred. Where necessary, static condition offsets were then applied to the flight test data to enhance the accuracy. For example: static pitch rate should be 0 deg/sec; if it is consistently found to be 0.1 deg/sec, then this value may be subtracted from the flight data. Other parameters such as AOA and AOSS are more problematic. In such cases, various techniques to establish model consistency are employed. For example, in a trimmed wings-level flight, the AOA and pitch attitude should follow each other with some constant offset (depending on FTIS calibration); if this is not the case, then one of the parameters is now known to be in error. Consistency checking of AOSS is more problematic and may be accomplished by the use of advanced state-estimation analysis and prediction software. Signal data processing is usually accomplished as a companion activity to data verification. The intent of data processing is to clean-up all signals prior to modelling to ensure that any spurious noise, spikes, data drop-outs (where a signal may have a spurious value for a short duration) do not effect the model development and POM process. If a parameter contains a spurious artefact, then the POM routine will expect the model to simulate the artefact. Consequently, the model developer requires a reasonably clean signal to avoid this situation. This process may become problematic when a measure of uncertainty exits as to whether a signal is measuring true aircraft behaviour or a spurious artefact has been introduced. For example, this situation can occur with low frequency oscillations observed on rotation rates and airflow angles that may either be signal/sensor problems or a valid aircraft modal response. Consequently, data processing must be kept to the absolute minium necessary, and must be accomplished by personnel who are familiar with aircraft flight dynamics. 3.6 Slip Stream Influences As approximately 55% of the AP-3C wing area is under the influence of the propeller slip stream, the complexity of the modelling effort for the AP-3C was significantly increased. Power effects naturally had a large and complex influence on the behaviour of the wing and hence the flying qualities of the aircraft. This power influence was modelled through the use of a Coefficient of Thrust (C T ) parameter. C T is a nondimensional parameter that is effectively proportional to the strength of the propeller slip stream. The modelling technique required that all likely stability derivatives were investigated for functional dependency on C T. Note, the types of manoeuvres conducted in the flight test programme have a significant impact on whether the modeller is able to isolate the influence on each stability derivative of such dependencies. In the case of a large turbo-propeller aircraft, the personnel constructing the flight test programme should be anticipating that power effects will be present and tailor manoeuvres accordingly (guidance may not be provided in the IATA FSDPDR). For example, in the AP-3C

6 flight test programme, aileron step inputs were executed to enable determination of both aileron power and roll damping. It was anticipated that the propellers would have an influence on the ailerons and, as a consequence, some test points were flown at various power settings to help characterise the power effect (the ailerons lost 10-15% effectiveness with the engines at idle power). Specific aircraft types will have flight test requirements unique to that type; an effective generic flight test programme is difficult to develop and will lead to inefficient use of flight test time. Personnel developing the flight test programme should have a comprehensive understanding of the likely characteristics and dependencies of the aircraft, otherwise there is a high probability that the flight test data will be insufficient for a high fidelity model. However, flight testing is a costly exercise and the test program needs to be efficient to minimise those costs. The best way this can be achieved is to tailor the test program as far as possible according to the influence on behaviour of the test aircraft configuration. 3.7 AeroElastic Influences The complexities of dealing with a flexible aerostructure also have to be taken into account when developing a high fidelity flight model. Modelling staff on the AP-3C project discovered that the flight test data for AOA contained a dependency on airspeed (dynamic pressure), which was exacerbated by the large airspeed envelope that the aircraft was capable of operating over. As previously discussed, AOA was measured by a vane located on a wing mounted boom assembly. RSCA flight test staff developed a calibration equation based on Reynolds Number and predicted up-wash effect, to enable true AOA to be derived from the measured value. At FTIS design stage, no data on likely torsional wing flex was available. However, during subsequent analysis of the flight test data, a torsional effect was noted for various Mach and airspeed regimes, which influenced the expected value of AOA. Consequently, modelling staff were able to compare measured AOA with other check parameters and, therefore, a wing flex stability derivative parameter was developed as part of the model. Again, personnel developing the flight test programme need to anticipate the significance of such complex effects and develop suitable manoeuvres that will allow modelling staff to determine the effect on aircraft dynamics. 4. Recommendations A model development agency should be involved from the outset of any flight test program for high fidelity simulator acquisition. This will ensure that: comprehensive data verification and validation is performed during flight test conduct to enable data problems to be dealt with in a timely manner, and the data requirements of that agency are incorporated into the flight test program such that all likely influences on aircraft behaviour as well as particular test techniques are accounted for. 5. References 1. International Air Transport Association (IATA). Flight Simulator Design & Performance Data Requirements, 5th Edition, ISBN International Civil Aviation Organization (ICAO). Manual of Criteria for the Qualification of Flight Simulators (MCQFS), First Edition, 1995 Doc 9625-AN/ Author Biographies Darren Jones has been a Flight Test Engineer and Senior Project Manager with Raytheon Systems Company Australia (RSCA) since His current position is Deputy Director Engineering. Before his time with RSCA, he held a number of FTE positions including a four year stint with Pilatus Aircraft in Switzerland on a variety of aircraft development and certification programs. He earned a Bachelor of Aeronautical Engineering from the Royal Melbourne Institute of Technology (RMIT) in 1983, and a formal FTE qualification from the United States Naval Test Pilot School (Patuxent River) in While he performed a limited amount of simulator flight test data gathering during his time at Pilatus, the AP-3C Program was his first real introduction to the development of a high fidelity six degree of freedom aircraft simulator. Darren has a total of seventeen years of engineering experience, thirteen of these in management or conduct of flight test programs and associated data analysis. James Keirnan studied Science and Engineering at Sydney University and graduated in 1986 with a BE (Elec) and BSc (Physics). He studied as a RAAF Cadet, and was posted to No 486 SQN and RAAF RICHMOND after graduation. Whilst at Richmond, James was involved in the coordination, trial and development of avionics modifications on C-130 Hercules aircraft. In 1990 James was selected to study at Cranfield Institute of Technology (CIT) for a Masters Degree in Flight Dynamics. He graduated from CIT in 1993 after completing a thesis on the application of robust control theory to flight control systems. On return from the UK, James was posted to the RAAF Logistics Command Flight Controls Technology cell. In 1995, he was posted to the Aircraft Research and Development Unit, to head the airborne instrumentation section. In 1997, James left the RAAF and joined Ball Aerospace Australia as a Systems Engineer. In late 1998, James joined Raytheon to manage the AP-3C Advanced Flight Simulator flight modelling project. He is keenly interested in high fidelity flight simulation.

7 Figure 1: Typical Quick-Look (Ground Station) Screen Display Annex A

8 Figure 2: Typical Analyse-9 (Ground Station) Screen Display Annex A