A Unifying Framework for Longitudinal Flying Qualities Criteria Robert F. Stengel* Princeton University, Princeton, New Jersey ‘The objective ofthis papers to interpret existing Mying qualities eitera ltivariable format based spon state vectors and murices of stabilly and coatrl derivatives. Sttespace methods are shown to be early ‘compatible wit conventional fying qualities analysis. Furthermore they lead to definition of the wellknown control anticipation parameter that Kas potential wility in the evalua z Introduction -XISTING flying qualities erteria are based upon modal characteristics and single-input/single-ourput_ trans Functions, and while these parameters are adequate to describe the more important aspects of conventional aircraft response, they tend to mask the multivariable nature of the motion, particularly for “superaugmented” configurations whose dynamic modes are revised by strong feedback control. ertheless, scalar criteria have been and continue to be useful. The relationship between existing criteria and proposed criteria must be understood by a wide audience, for the essence of a criterion is not only that it be right but that it Serve as a common basis for evaluation and decision-making ‘With these thoughes in mind, the objective of this paper is to interpret existing criteria jn'a multivariable format based upon sate vectors and matrices of stability and control crivatives, implicitly or explicitly, stability-and-control engineers have been Norking with vectors and matrices for decades.' The state-space approach is little more than @ method of accounting, and it is equally applicable to time domain (differential equation) or frequency-domain (Fouriee/Laplace transform) models of dynamic systems (for example, Ref. 2) There are at least four advantages to using state-space models. The first is that a large number of seelar equations ccan be expressed as a small number of vector-matrix uations; hence, the dysamic system can be described in concise yet rigorous fashion. The second ig that the vector= mairix expression is independent of the dimension of the multivariable system. Consequently, it is easy to extend simple low-order} concepts to complex (high-order) systems, ‘The third is that the linearization of she aircraft equations of motion, which are inherently nonlinear, is straightforward. There s no ambiguity regarding what is the nominal and what, Is the perturbation, and potentially difficult coupling due 10 nnon-straight-and-level light is handied readily. The fourth is that appropriate reduced-order models, including equivalent systems, can be generated easily. The procedures for ob- taining these models are relatively transparent in the state- space format. A Few Busie Rules The state vector, x(7), contains the aircraft's motion ariables, e.g, the Four longitudinal variables, axial velocity, Preanicd at Paper SLI889 at the AIAA Atmospheric Fight hanies Conference, Albuquerque, N. Mex., Aug” 19-2, 1981; ‘ebmted On 5, IAI, revision receted Dee 8, 1982. Copstigh © American Intute of Acronaucs and Arona, In 1981. All erected, *Profecce of Me ical end Aecospace Engineering, Plsht [Esboruiors. Asrosiste Fellow ATA. OF higher-order aireraft/controt normal velocity, pitch rate, and pitch angle: 4a) win) x= a, a an) In this example, the control vector, 5(0), may contain throttle position, flap setting, and elevator displacement: ar Bens | ar) @ EW) Neglecting disturbance inputs, the aircraft equations of ‘motion can be written as a single, nonlinear vector differential equation of appropriate dimension: #0) fet.) @ ‘The state, x(2), may be observed as the output ofthe dynamic system, but other quantities may be observed as well. For ‘example, the outputs may be in a different reference frame (8, body-axis dynamic equations with stability-axis out pus) or they may be functions of the state and control (€-8., accelerometer measurements). Consequently, the output vector, (1), is a (possibly nonlinear) function of the state and contra vectors, yt) 2h[x00.800) ® and its dimension may be larger or smaller than that of the state vector. ‘The aircraft is in equilibrium if the states are unchanging, or #(1) =0 [Eg (3)}. Denoting this “rimmed” condition by a zero subscript, the states and controls for trim ate defined by xeD8e(0 1 o Perturbations from trim can be characterized by a linear model, which is obtained by a Taylor series expansion of Eq. G) (vith g~0). Using SC) to denote the perturbation variables, the lincerized dynamic equation is Sk) = Farin + 65000) © | |MARCH-APRIL 1983 inearized output equation is (=H, 8x() + H,8060) [) where F. G. Hy, and H, are “Jacobian” matrices of partial Gerivatives evafusted at the trim condition. F contains the aircraft's stability derivatives plus gravitational and kinematic terms, G contains the aircraft's control derivatives, and H, and H, depend on the definition of the output, All of the elements of thes. 2s are real, scalar variables. With the assumption that the aircraft is trimmed, it can be assumed that the matrices are constant. The initial response to control is a state rate, which, from Eq. 4) with Ax(0)=0is, Ai(0) =G40(0) @ “The corresponding initial output rate is found by substituting this in Eq. (7). ‘Constant perturbations in the control lead to retrimming at a new state. Denoting these constant perturbations by ()*, the new trim condition is approximated by Eq. (6) with Ax) "ax" + GAS" 0a) Fax =~ Gas" @) Assuming that F is a nonsingular matrix, that is, its inverse Arts -F1Ga8" ao From Eq. (7), the corresponding steady-state output per turbation is ay HEG+H,) a8" ay The matrices multiplying AS* in Eqs. (10) and (11) represent all the “de” (zero frequency) gains of the state and output transfer funetions described below. ‘The transfer function matrices are derived from Eqs. (6) and (7). Neglecting initial conditions, the Laplace transform of the linear, constant-coefficient vector dynamic equation is sx(s) = Fax(s) +628(s) (22a) Ax(s)=(6I-F -!6a8(5) a2 (sI~F)'G is the (nx) transfer funetsion matrix relating the m controls to the m states, and it can be written as, Adi(sI-F)G _N,(s) shee ak dere eel) ay where d(s) is the scalar characteristic polynomial whose factored form contains the n poles (or eigenvalues) of F. N,(5) is the (ax) numerator matrix containing the 2er0s of the system's state transfer functions. ‘The output transfer function matrix, 3¢(s), is found by g the Laplace transform of Eq. (7) and substituting Eq. (120) for Ax(s). Thea Nas) a 31S) =H AIF) IG +H, = ay sdsorder roodaling simplifies the analysis of systems with fast and slow dynamic modes, Suppose that fast and slow state variables, Sr, and 4x,, can be identified in Eg. (6), LONGITUDINAL FLYING QUALITIES UNIFYING FRAMEWORK as If there is litle interaction between the fast and slow modes, ive,, if FE and Ff are negligible, then Ea. (15) can be trun- ted, and the motions can be considered separately, Aig = Fors +0545 a6) ip =Fydtp G88 an With moderate interaction, the slow modes tend to be modified by the fast modes but not the reverse, and Ea. (15) can be resigualized. On the time scale of the fast modes, the effects of the slow modes are negligible, so Eq. (17) can be ‘considered to be an adequate descriptor of the fast modes. On, the slow time scale, stable fast modes appear to be in- stantaneous—so quick that they are alweys in steady state. To, approximate the effects of the fast modes on the slow modes, itis assumed that Az,=O, allowing the lower half of Eq. (15) to provide an algebraic solution of Ax, in terms of Ax, and. 5: : Axp= FF! (FGats +G,A5) as, Substituting Bq. (18 agi, the remaining solution for Ax5 Eq. Aay= (Fy FFF !F ARS + (Gs FIFF'G A185 (19) ‘The assumption of instantaneous fast modes can be relaxed. The dynamic lags associated with these modes can be modeled as a single effective time delay, r, or rise time, T. that is independent of the slow modes. Wich an equivalent time delay, Axp(s) =axpe™™ Ang (9) = I~ (FS FEF SFE) H1G,— FIFO Gye" 1ab(s) en ‘while with an equivalent rise time (0 95% of Final value), at = ° sse-f- [teen DB «(6- LRG Jao @ If there is strong interaction between the modes, neither truncation nor residualization is appropiate. Longitudinal Equations of Motion Atterate State Vectors Three definitions of the longitudinal state vector ate of particular interest: xs lu w gq 6)? es mV ye)? es, nal¥o gal? 26)‘on, the aircraft's translational and angular ¢ expressed in body axes, and the piteh angle relates the body axes to an inertial (or, in this case, Earth- relative) frame. In the second version, the translational velocity is expressed in an Earth-relative frame, with velocity magnitude, V, and flight path angie, y. In the thitd case, the vehicle's attitude is measured with respect to the Might path angle rather than an inertial axis, The nonlinear equations which relate x, w, V, a, and to each other are a wt] en =! (WV) 28) yet-a ay An exact st of nonlinear equations of motion can be wer wth any ofthe arate sate vectors, When body. ts tanlainal velochies are usd, iy convene tp ex pres foresters Of ania and normal forse cocfcets {yand Cy 01 Cy and Cy), whereas and drag (Cy and Gf tre bic apo Tor Earnie velosines The pitching moment coeficent Gay te the same tn al twee For the mott per, we are intrested inthe lndariaed equations of motion bere, ba k's useful to considera pa ‘iular nonlinear output variable that does not appear tn any Othe aerate sete vectors The normal aceeleeaion a that is sensed at an arbitrary polnt, Ua Jara). 00 the titre measure lative ic center oF BS eg Wau pot (— Gt pz 4+ BHENY4 — B+ G24 = scorbcost 60) Choosing 4 to be ihe pilot's station provides results of particular significance in longitudinal flying qualites.. For symmetric flight and at points in the aircraft's plane of symmetry, the normal acceleration is 1480-44 gost en If it is further assumed that ¢.2 is negligible, the perceived oad factor, 2,, (in ¢ units with “up” taken as the positive direction) is| (qu Gx4) +0088 oa e Neglecting unsteady aerodynamic effects, the F and G matrices associated with the third state vector [Eq. (25)] ean, be expressed as follows: TDy = s00576 1D, 1D, seein caln envio acai ta | My 0 M, My adie, a impli hese pe ony See eee Warrant improvement” are obtained for the values of At, shown in Table 1 The preliminary findings of Table 1 excent were with a single pilot, except as noted. Flight phase B refers to flight path control at altitude. All these Af,. values are well within the level I region of Fig. 2, and flight phase C results, which involved tracking an aircraft carrier approach mizzor, sare more restrictive than the ight phase A result. Pitch rate rise time is reported to correlate well with pilot opinion of low-speed longitudinal maneuvering charac- {ersties in Re. 14. The present results are consistent with this finding and provide additional. gu relationships between rise time, flight speed, a pitch rate to normal acceleration response. railo of Conclusion State-space methods are entirely compatible with con- ventional flying qualities analysis, and they provide an iam proved framework for extending existing single-control techniques to higher-order systems. Relationships between important variables and fying qualities parameters are Gerived easily, reduced-order models are generated, and the mechanisms for addressing more complex problems such as pilot location effects and phugoid/short period coupling are Geveloped. A time-domain definition of the control ax- ticipation parameter is seen to be numerically identical to the frequency-domain definition, and it is somewhat more amenable to the incorporation of time delays through the mechanism of an equivalent rise time. While the state-space approach can be helpful for curreat fiying qualities studies, perhaps its most significant application is yet to come in the development of flying qualities criteria for the multiple ‘control case. Acknowledgment ‘This research was sponsored by the Center, Warminster, Pa., under Cont References !McRuer, Day Ashkenas, I, and Graham, D., Adrerft Dynarmies and Automatic Control, Princeton University Press, Princeton, NJ. i973 "Brogan, W.L., Modern Control Theory, Quantum Publishes, Ing, New York, 1974, Siuhrle, W., rs A Handling Qualities Theory for Precise Flight Copicol," AFFDU-TR-65-198, Wright-Patterson AFB, June 1956. “sltary Specification, Flying Qualities of Piloted Aipblanes,” MIL-F-TS8C, USAF ASD/ENESS, WrightPaterson AFB, Nov 1960, Stobie, H., Eliott, E., and Malcolm, L., “A New Longitudinal Handlina Qualities Criterion,” Proceedings of the 1956 National Aerospace dnd Elecyronies Conference, Dayton, Onio, May 1966, “Stengel, RAF, ““Eguiibrivin Response of Flight Control Systems," Automatica, Vol 18, No.3, May 1982, pp. 385-348 AtHarrah, RC., Lemsnna, W.J., and Hodgkinson, J.» “Today's Specifications Appropriate for Tomorrow's Algplanes Stability ond Control, AGARD-CP-260, Neu ine, May 1979 pp, 23-1232 ‘Ralston, A. and Wilf, H.S., Mashematical Methods for Com ters, J. Wiley & Sons, New York, 1968 Takahashi, ¥., Rabin, M.J., and Auslander, DM. Control and Dynamic Sysiems, Addison-Weiley Pubiening Go., Reading, Mas, a, "Enkin, B.. Dsnamis of Fght-Stabitiy and Contral, 4. Witey & Sons, New York, 1982, Menai, CAR. “Calspan Recon Qualities," NASA CR-159336, April 1980. ‘opt. GoE, and Harper, RP. Je one Ecieaton of Avera Heading Quaiie,” BABA TN DSS, en. D.C., April 1969. suengel, RLF and Miller, G.E.,""Flight Tens of Misroprocetsor Consol System." Journal of Guidance aad Control, Vol. 3 NO. & Now -Dec. 1980, pp. 484-500. Stool, HAs de Boer, W.P., val Air Development ract N62269-80-C-0720, Waions for SCR Flying I27U, Amsterdam, Des. 197 |