NASA/TM—2016-218925 AIAA–2015–3987 Investigation of Asymmetric Thrust Detection With Demonstration in a Real-Time Simulation Testbed Amy K. Chicatelli, Aidan W. Rinehart, and T. Shane Sowers Vantage Partners, LLC, Brook Park, Ohio Donald L. Simon Glenn Research Center, Cleveland, Ohio January 2016 NASA STI Program . . . in Profi le Since its founding, NASA has been dedicated • CONTRACTOR REPORT. Scientifi c and to the advancement of aeronautics and space science. technical fi ndings by NASA-sponsored The NASA Scientifi c and Technical Information (STI) contractors and grantees. Program plays a key part in helping NASA maintain • CONFERENCE PUBLICATION. Collected this important role. papers from scientifi c and technical conferences, symposia, seminars, or other The NASA STI Program operates under the auspices meetings sponsored or co-sponsored by NASA. of the Agency Chief Information Offi cer. It collects, organizes, provides for archiving, and disseminates • SPECIAL PUBLICATION. Scientifi c, NASA’s STI. The NASA STI Program provides access technical, or historical information from to the NASA Technical Report Server—Registered NASA programs, projects, and missions, often (NTRS Reg) and NASA Technical Report Server— concerned with subjects having substantial Public (NTRS) thus providing one of the largest public interest. collections of aeronautical and space science STI in the world. Results are published in both non-NASA • TECHNICAL TRANSLATION. English- channels and by NASA in the NASA STI Report language translations of foreign scientifi c and Series, which includes the following report types: technical material pertinent to NASA’s mission. • TECHNICAL PUBLICATION. Reports of For more information about the NASA STI completed research or a major signifi cant phase program, see the following: of research that present the results of NASA • Access the NASA STI program home page at programs and include extensive data or theoretical http://www.sti.nasa.gov analysis. Includes compilations of signifi cant scientifi c and technical data and information • E-mail your question to [email protected] deemed to be of continuing reference value. NASA counter-part of peer-reviewed formal • Fax your question to the NASA STI professional papers, but has less stringent Information Desk at 757-864-6500 limitations on manuscript length and extent of graphic presentations. • Telephone the NASA STI Information Desk at 757-864-9658 • TECHNICAL MEMORANDUM. Scientifi c and technical fi ndings that are preliminary or of • Write to: specialized interest, e.g., “quick-release” reports, NASA STI Program working papers, and bibliographies that contain Mail Stop 148 minimal annotation. Does not contain extensive NASA Langley Research Center analysis. Hampton, VA 23681-2199 NASA/TM—2016-218925 AIAA–2015–3987 Investigation of Asymmetric Thrust Detection With Demonstration in a Real-Time Simulation Testbed Amy K. Chicatelli, Aidan W. Rinehart, and T. Shane Sowers Vantage Partners, LLC, Brook Park, Ohio Donald L. Simon Glenn Research Center, Cleveland, Ohio Prepared for the 51st Joint Propulsion Conference cosponsored by AIAA, SAE, and ASEE Orlando, Florida, July 27–29, 2015 National Aeronautics and Space Administration Glenn Research Center Cleveland, Ohio 44135 January 2016 Level of Review: This material has been technically reviewed by technical management. Available from NASA STI Program National Technical Information Service Mail Stop 148 5285 Port Royal Road NASA Langley Research Center Springfi eld, VA 22161 Hampton, VA 23681-2199 703-605-6000 This report is available in electronic form at http://www.sti.nasa.gov/ and http://ntrs.nasa.gov/ Investigation of Asymmetric Thrust Detection With Demonstration in a Real-Time Simulation Testbed Amy K. Chicatelli, Aidan W. Rinehart, and T. Shane Sowers Vantage Partners, LLC Brook Park, Ohio 44142 Donald L. Simon National Aeronautics and Space Administration Glenn Research Center Cleveland, Ohio 44135 Abstract The purpose of this effort is to develop, demonstrate, and evaluate three asymmetric thrust detection approaches to aid in the reduction of asymmetric thrust-induced aviation accidents. This paper presents the results from that effort and their evaluation in simulation studies, including those from a real-time flight simulation testbed. Asymmetric thrust is recognized as a contributing factor in several Propulsion System Malfunction plus Inappropriate Crew Response (PSM+ICR) aviation accidents. As an improvement over the state-of-the-art, providing annunciation of asymmetric thrust to alert the crew may hold safety benefits. For this, the reliable detection and confirmation of asymmetric thrust conditions is required. For this work, three asymmetric thrust detection methods are presented along with their results obtained through simulation studies. Representative asymmetric thrust conditions are modeled in simulation based on failure scenarios similar to those reported in aviation incident and accident descriptions. These simulated asymmetric thrust scenarios, combined with actual aircraft operational flight data, are then used to conduct a sensitivity study regarding the detection capabilities of the three methods. Additional evaluation results are presented based on pilot-in-the-loop simulation studies conducted in the NASA Glenn Research Center (GRC) flight simulation testbed. Data obtained from this flight simulation facility are used to further evaluate the effectiveness and accuracy of the asymmetric thrust detection approaches. Generally, the asymmetric thrust conditions are correctly detected and confirmed. I. Introduction Asymmetric thrust is cited as the cause of several loss of control aviation incidents and accidents (Refs. 1 to 3). As noted in the propulsion system malfunction plus inappropriate crew response (PSM+ICR) reports, the detection of unintended asymmetric thrust conditions needs to be performed in a timely manner in order for there to be sufficient time for recovery. This effort was part of a feasibility study that evaluated three asymmetric thrust detection methods. This paper presents the results from that study which was conducted in simulation to demonstrate that asymmetric thrust conditions can be detected. In addition, to test effectively any proposed asymmetric thrust detection method, a realistic flight simulation environment is necessary. The NASA GRC flight simulation laboratory provides an opportunity to evaluate the effectiveness of the proposed methods with a pilot-in-the-loop. However, additional work would be necessary to address whether this information should be annunciated to the flight crew. The remainder of the paper is organized as follows. The paper begins with a section that provides the motivation for this work and describes the significance and improvement over state-of-the-art. Following that, the description of three developed asymmetric thrust detection methods is provided. Next, results from a simulation-based sensitivity study and the application of these methods in a real-time flight simulation laboratory are presented. Finally, a summary that outlines the effectiveness of the asymmetric thrust detection approaches is presented along with generalized concluding remarks about the research effort. NASA/TM—2016-218925 1 II. Motivation A. Significance As previously noted, asymmetric thrust is cited as the cause of several Loss of Control aviation incidents and accidents. The sequence of events that lead to a PSM+ICR event can be described as follows. In a typical scenario, the aircraft’s autopilot is being used by the crew and the autothrottle control is engaged. Due to a failure, an unintended asymmetric thrust condition occurs while the autopilot and autothrottle are engaged. Due to the automated flight control systems, the asymmetric thrust condition is initially managed, but it progressively increases because the flight controls reach their limits of effective control. When the autopilot disengages and the crew takes over control, there may not be enough control authority for the crew to sufficiently control the aircraft. In addition, the crew’s response may be inappropriate and exacerbate the situation. If the flight crew is aware of the developing thrust asymmetry, they should be able to identify the cause and take preventive or corrective action. Therefore, the detection of the asymmetric thrust condition is needed, and the annunciation of asymmetric thrust may be beneficial in warning the flight crew. However, as indicated in the Federal Aviation Administration (FAA) report (Ref. 4) for indications of propulsion system malfunctions, there are risks associated with thrust asymmetry annunciation and the only time when annunciation may be practical is when the autopilot is engaged. Many of the reported incidents and accidents of thrust asymmetry by themselves were not so significant that they were not recoverable. In addition, aircraft are required to be controllable in extreme asymmetric thrust conditions throughout typical flight profiles. For the purposes of this effort, the asymmetric thrust condition that leads to aircraft upset and to loss of control is of interest. Typically that condition occurs due to the lack of reaction or from the inappropriate action of the flight crew. It is in this way that an asymmetric thrust condition can become unmanageable. B. Improvement Over State-of-the-Art The flight crew is responsible for recognizing and responding to unintended asymmetric thrust conditions. Today, this is done by the pilots monitoring available engine cockpit instrumentation such as rotor speeds or engine pressure ratio, which provide an indication of engine power. If an imbalance in the power produced by the aircraft’s engines is observed, the pilots must take appropriate action to address the issue. It is possible, when the autopilot is being used, the thrust imbalance can actually be exacerbated. If the flight control system is continually correcting for thrust asymmetry, the limits of control authority could be approached. In this case, recovery by the flight crew will be challenged by the suddenness of the event and the reduced margin for corrective action. Depending on its design, the autopilot will disengage at some point as it becomes unable to maintain the desired input flight conditions (i.e., heading, airspeed, altitude). When the autopilot disengages, the flight crew is given control of an aircraft that may have developed an unusual flight attitude with controls at or near their maximum deflection from the autopilot. In these situations, recovery depends on the proper response and skill of the flight crew. As noted above, asymmetric thrust conditions can be masked from the flight crew when the automatic controls are being used. In order to add an alert for asymmetric thrust conditions, the thrust imbalance must be detected and confirmed when these controls are being used. The results of the research presented in this paper, will show that the detection and confirmation of asymmetric thrust conditions is possible. The simulation studies and pilot-in-the-loop tests that were conducted under this effort will show that reliable automated real-time detection of asymmetric thrust conditions is feasible. NASA/TM—2016-218925 2 III. Asymmetric Thrust Detection and Confirmation For this research effort, three asymmetric thrust detection approaches are developed and evaluated. The first two are based on producing an estimate of the engine thrust for each engine and comparing those values with each other in order to determine a mismatch in the thrust. The first two approaches include a Kalman filter-based thrust estimation approach and a two-dimensional table lookup thrust estimation approach. The third method takes a fundamentally different approach. Instead of a monitoring for a mismatch in estimated thrust between engines it monitors for a mismatch between the commanded and actual power produced by an individual engine. This is done by monitoring the primary engine control parameter (typically either corrected fan speed or engine pressure ratio (EPR)), which is a proxy for the amount of thrust produced by the engine. If an engine is detected to be producing more/less thrust than commanded, that serves as confirmation that an asymmetric thrust condition is likely. The three detection approaches are described in the following sections. A. Kalman Filter Approach A Kalman filter (KF) is an optimal linear estimator designed to estimate the unknown states of a dynamic system. It incorporates a dynamic model of the system and is designed to recursively update estimates by processing acquired system measurement data. Accounting for measurement noise and model uncertainty, the Kalman filter is designed to minimize the mean squared error in the estimated parameters. It is well suited for aircraft engine applications and several previous efforts have reported on the application of Kalman filters for onboard real-time aircraft engine performance estimation (Refs. 5 and 6). In this study, an asymmetric thrust detection strategy based on Kalman filter estimation technology is considered. Here, a Kalman filter is designed and applied for each engine installed on the aircraft. To account for the nonlinear behavior inherent in an aircraft gas turbine engine, a piecewise linear Kalman filter design is applied. Individual linear Kalman filters are designed spanning the entire engine operating envelope and then combined and scheduled applying interpolation to account for changes in engine operating condition. In addition to estimating the dynamic states of the engine, the Kalman filter is also constructed to estimate states reflective of turbomachinery performance deterioration (Refs. 7 and 8). In this fashion, that Kalman filter is able to account for deterioration induced performance changes in the engine. Additional details on the Kalman filter formulation and implementation for asymmetric thrust detection are provided in the subsections below. 1. Kalman Filter Formulation The nonlinear model of an aircraft engine can be represented by the following equations x  fx,u,h y gx,u,h (1) z  g x,u,h z where x and u represent the vectors of engine state variables and control command inputs, respectively. The vector h represents health parameters, such as efficiency or flow capacity, reflective of performance deterioration within the major modules of the engine. For given input values, the nonlinear functions f, g, and g generate the vectors of state derivatives x, sensed engine outputs y, and unmeasured engine z outputs such as net thrust denoted by z. By linearizing the engine model at a given operating point, the following state-space equations are obtained: NASA/TM—2016-218925 3 x  Axx Buu Lhh  trim trim ref x u h x  AxBuLh yy  Cxx Duu Mhh  trim trim trim ref y x u h (2) y CxDuMh zz  Fxx Guu Nhh  trim trim trim ref z x u h z FxGuNh Here, A, B, C, D, F, G, L, M, and N are the state-space matrices reflecting system dynamics. The trim vectors, denoted by the subscript “trim,” reflect the values of the state variables, commands, and measured and unmeasured outputs when the model is at steady-state (i.e., x = 0) at the given operating point. The vector h represents a reference health condition specified by the system designer. In ref Equation (2), parameter deviations relative to trim or reference conditions are denoted by the delta symbol (). Through algebraic manipulation, Equation (2) can be re-written to shift the health parameters to become state variables as shown in Equation (3): x A Lx   Bu h 0 0h x   y C M  Du (3) h x   z F N  Gu h Since engine performance deterioration evolves slowly in time, the health parameter states in Equation (3) are modeled without dynamics. Once the health parameters are augmented with the state variables, they can be estimated by applying a Kalman filter as long as the system is observable. However, a necessary condition for observability given the Equation (3) formulation is that there are at least as many measurements as health parameters (Ref. 9). To construct a reduced-order state space system of appropriate dimension to enable Kalman filter formulation, consider a transformation matrix, V*, that maps the health parameter vector, h, to a tuning vector of lower dimension, q, such that: qV*h (4) An approximation for h based on q can be calculated using the pseudo inverse of V*: hˆV*qˆ (5) NASA/TM—2016-218925 4 Then, substituting Equation (5) into Equation (3) produces the following reduced-order state space system: x A LV*x B     u q 00q 0 xxq Axq xxq Bxq  x yCMV* qDu (6) Cxq  xxq  x zFNV* qGu Fxq  xxq The choice of the transformation matrix is a design decision made prior to constructing the Kalman gains. In this study a technique, referred to as “optimal tuner selection,” is employed to produce a transformation matrix that is a linear combination of all health parameters, and constructed such that the mean squared estimation error in the parameters of interest are minimized (Ref. 7). In this case, the transformation matrix was selected to minimize the estimation error in net thrust. Given Equation (6), a linear Kalman filter at a given operating point can be formulated as:    xˆ  A KC xˆ B uKy xq xq xq xq xq yˆ C xˆ Du (7) xq xq zˆF xˆ Gu xq xq After individual linear Kalman filters are designed spanning the entire engine operating envelope, they are combined and scheduled applying interpolation to form the piecewise linear Kalman filter. 2. Kalman Filter Implementation for Asymmetric Thrust Detection A block diagram of the piecewise linear Kalman filter implementation to estimate the thrust produced by an individual engine is shown in Figure 1. The Kalman filter requires engine sensed measurements (y) and actuator inputs (u). Parameter correction is applied to improve the interpolation between grid points in the piecewise linear Kalman filter. Trim and matrix information corresponding to the current operating point are retrieved applying a three-dimensional interpolation scheme using altitude, Mach, and corrected fan speed as the scheduling parameters. The Kalman filter produces estimated corrected output deltas from trim consisting of state variables (xˆ ), sensed measurements (yˆ ), and net thrust (zˆ ). Corrected xq c c net thrust is produced by summing the estimated delta in net thrust, zˆ , and net thrust at the trim c condition, z . trim The Kalman filter estimated net thrust values for each engine are compared following the asymmetric thrust detection and confirmation logic as shown in Figure 2. The absolute difference in estimated corrected net thrust between the two engines is calculated and then converted to an absolute percent of maximum thrust. This absolute percent error signal is then compared to a pre-established detection threshold. When an exceedance of this threshold is detected and then persists for an established time duration, the asymmetric thrust condition is confirmed and annunciated. NASA/TM—2016-218925 5 From Engine xˆ xq y y + y c c Cor ‐ Kalman  yˆc u rec uc + uc Filter t zˆ ‐ c x Altitude xq,trim u Mach Trim  trim y Interp trim zˆ z + c trim + (Estimated Altitude A,B,C,D,F, Corrected Mach Matrix  G,L,M,N,K Net Thrust) Interp Corrected Fan Speed Figure 1.—Kalman filter-based thrust estimation. Absolute Absolute error Engine 1 Percent Estimated + error |u|  detection yes persistency yes Annunciate Corrected 100 threshold threshold thrust Net Thrust - ÷ exceeded? exceeded? asymmetry Engine 2 Estimated Max no no Corrected Thrust Net Thrust No thrust No thrust asymmetry asymmetry detected detected Figure 2.—Asymmetric thrust detection and confirmation logic. Relative to the other asymmetric thrust detection approaches considered in this study, the primary benefit offered by the Kalman filter is the estimation accuracy it enables. It is designed to account for transient engine behavior and turbomachinery deterioration induced changes in engine performance. However, it is a relatively complex solution in terms of processing requirements compared to the other approaches described below. B. Table Lookup Approach This asymmetric thrust detection approach estimates the net thrust of each engine applying a two-dimensional table lookup approach based on corrected fan speed and Mach number as shown in Figure 3. The lookup table data is created from steady state data generated using an engine model reflecting mid-life (i.e., 50 percent deteriorated) engine performance. The model is run at operating points over the entire flight envelope of the engine spanning a range of corrected fan speeds, Mach number, and altitude settings. Then, for each corrected fan speed and Mach number combination, corrected net thrust results generated over the range of altitudes considered are averaged to produce the two-dimensional lookup model. An estimated corrected net thrust value is produced for each engine, then these estimated thrust values are processed applying the same asymmetric thrust detection and confirmation system as shown previously in Figure 2. As compared to the Kalman filter approach, the table lookup method is simpler in design, which is a benefit for a flight software application. However, unlike the Kalman filter, the table lookup method does not account for engine performance deterioration or engine transient dynamics. NASA/TM—2016-218925 6