Document
J OURNAL OF A IRCRAFT Vol. 47, No. 1, January – February 2010
Current Methods Modeling and Simulating Icing Effects
on Aircraft Performance, Stability, Control
∗ Thomas P. Ratvasky NASA John H. Glenn Research Center at Lewis Field, Cleveland, Ohio 44135 † Billy P. Barnhart Bihrle Applied Research, Inc., Hampton, Virginia 23666 and ‡ Sam Lee ASRC Aerospace Corporation, Cleveland, Ohio 44135 DOI: 10.2514/1.44650 Icing alters the shape and surface characteristics of aircraft components, which results in altered aerodynamic forces and moments caused by air fl ow over those iced components. The typical effects of icing are increased drag, reduced stall angle of attack, and reduced maximum lift. In addition to the performance changes, icing can also affect control surface effectiveness, hinge moments, and damping. These effects result in altered aircraft stability and control and fl ying qualities. Over the past 80 years, methods have been developed to understand how icing affects performance, stability, and control. Emphasis has been on wind-tunnel testing of two-dimensional subscale airfoils with various ice shapes to understand their effect on the fl ow fi eld and ultimately the aerodynamics. This research has led to wind-tunnel testing of subscale complete aircraft models to identify the integrated effects of icing on the aircraft system in terms of performance, stability, and control. Data sets of this nature enable pilot-in-the-loop simulations to be performed for pilot training or engineering evaluation of system failure impacts or control system design.
effects caused by ice. However, even with these advancements, Nomenclature icing-induced loss of control incidents and accidents have occurred b = wing span and continue to occur on all classes of aircraft, from general aviation C , C = drag coef fi cient d D airplanes [1,2] and business jets [3,4] to transport category aircraft C = aileron hinge moment h;A [5 – 11]. These events typically result from a combination of causal C = aileron control effectiveness L;A factors, but they usually precipitate from the altered aerodynamics C , C = lift coef fi cient l L caused by icing.
C = pitching moment coef fi cient m The continuing problem of icing has been recognized by many C = pitching moment coef fi cient derivative with m organizations including the National Transportation Safety Board AoA, longitudinal stability (NTSB), the Commercial Aviation Safety Team (CAST), the Federal C = normal force coef fi cient N Aviation Administration (FAA), and the National Aeronautics and c = chord length Space Administration (NASA). The NTSB has made numerous k= c = ratio of protuberance height to chord length safety recommendations after icing related incidence and accidents Re = Reynolds number and has ranked icing among its “ MOST WANTED Aviation Trans- , AoA = angle of attack, deg § portation Safety Improvements ” since 1997. CAST, a consortium of = angle of sideslip, deg industry and government organizations working to improve aviation , , , = control surface de fl ection: aileron, elevator, a e r f safety, has developed interventions speci fi c to icing. They have rudder, fl ap, deg also speci fi ed that the implicit icing effects on fl ight envelope protection and high- fi delity simulators for upset recovery training are among their highest ranked intervention needs [12]. The FAA and the I. Introduction NASA Aeronautics Research Mission Directorate have invested IRFRAME icing has been a threat to aviation safety for nearly signi fi cant resources to support these safety recommendations,
A 80 years. Over that time, signi fi cant progress has been made to
develop interventions, and to advance the state of knowledge on the reduce the hazard through the design and implementation of ice icing atmosphere, the ice accretion process, and aerodynamic effects protection systems, improvement of certi fi cation procedures, and of icing.
improved pilot training. The foundation of these advancements The purpose of this paper is to review the methods currently has been research to better understand the icing environment, the available to model and simulate icing effects on performance, physics of the ice accretion process, and the degrading aerodynamic stability, and control. The primary method to model icing effects use wind-tunnel testing of 2-D and 3-D models to develop mathematical Presented as Paper 6204 at the Atmospheric Flight Mechanics Conference expressions or databases of aerodynamic forces and moments. Icing and Exhibit, Honolulu, HI, 18 – 21 Aug. 2008; received 30 March 2009; effects simulation uses these mathematical models or databases in accepted for publication 24 Oct. 2009. This material is declared a work of the conjunction with the equations of motion to explore the inter- U.S. Government and is not subject to copyright protection in the United relationship of various parameters and the modi fi ed fl ight envelope States. Copies of this paper may be made for personal or internal use, on imposed by the iced aerodynamics. Icing effects modeling and condition that the copier pay the $10.00 per-copy fee to the Copyright simulation methods have varied maturity levels and limitations in Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include terms of veri fi cation and validation. Recommendations are made to the code 0021-8669/10 and $10.00 in correspondence with the CCC.
∗ Aerospace Engineer, Icing Branch, 21000 Brookpark Road. Senior continue research in fl ight simulation modeling and real-time Member AIAA.
modeling to continue advancing the state of the art.
† Senior Engineer, 81 Research Drive.
‡ Research Engineer, Icing Branch, 21000 Brookpark Road. Member § AIAA. http://www.ntsb.gov/Recs/mostwanted/aviation_issues.htm 202 RATVASKY, BARNHART, AND LEE with the airfoil. Because these shapes are so different, they have very II. Airframe Icing: How it Forms different effects on the aircraft aerodynamics.
To model and simulate the degraded aerodynamics that results To reduce the aerodynamic penalty of icing, ice protection systems from ice, one should understand the ice accretion process and the (IPS) [14] are typically incorporated into airplanes to prevent or resulting ice characteristics because they drive the aerodynamic remove ice from critical fl ight components such as the wing and tail, effects.
engine nacelles and air intakes, propellers, and others. Ice protection Ice forms on an airplane as it fl ies through clouds consisting of systems that prevent ice from forming are typically called anti-icing supercooled liquid droplets (i.e., liquid water drops that are below systems, whereas those that remove ice are called de-icing systems.
0 C). The liquid droplets impinge on the leading edge surfaces, Anti-icing systems typically use heat from engine bleed-air or exchange heat with the environment and aircraft surface, and change electrothermal heaters to evaporate water drops that impinge on the phase from liquid to ice. Initially, the ice forms as a thin, rough layer.
leading edge surfaces or to prevent the water drops from freezing on Over time, the ice thickness increases and reshapes the leading edges the heated areas. De-icing systems debond ice that has already of all frontal surfaces. The rough, reshaped leading edges, in turn, formed on the leading edge through a mechanical deformation of the degrade the aerodynamics of the airframe.
surface (e.g., pneumatic de-icing boots) or through intermittently The ice accretion characteristics (size, shape, extent, roughness, heating the surface (e.g., electrothermal heaters near the propeller and translucency) are related to the atmospheric environment param- root). The design of an IPS on any given airplane may use a com- eters, the fl ight condition, the geometry of the aircraft component bination of anti-icing and de-icing equipment. Some airplanes have being iced, and the time in icing conditions. The icing cloud atmo- bleed-air thermal anti-icing on the wings and engine nacelle leading sphere has been de fi ned using three primary parameters.
edges only, leaving the tail surface completely unprotected. Others 1) Liquid Water Content (LWC): a measure of the amount of liquid have pneumatic de-icing boots on the wing, and on the horizontal and water in a unit volume of space.
vertical tails while having electrothermal de-icing on propellers.
2) Median Volumetric Diameter (MVD): a representative size of Some have a combination of bleed-air thermal anti-icing on the wing the water droplet spectrum. MVD is a value for a given spectrum such and engine nacelle leading edges, pneumatic de-icing boots on the that half the volume of water is in drops with diameters smaller than horizontal tail, and no-ice protection on the vertical tail.
the MVD and half the volume of water is in drops with diameters There are other IPS designs as well, so the list could go on. In the greater than the MVD.
end, the airframe manufacture determines the locations in which ice 3) Static Temperature (Ts): the ambient air temperature.
protection is needed and what type of IPS is required to enable The fl ight conditions consist of true airspeed, angle of attack, and safe fl ight operations in icing conditions and meet certi fi cation altitude. Geometric parameters consist of the size of the leading edge requirements.
radius, the chord length ( c ), single or multi-element wing, and Airframe icing is a highly complex dynamic phenomenon. The ice straight or swept wing or tail. Each of the preceding parameters that forms on an airplane is the result of a combination of icing cloud in fl uences the ice accretion development. The result is a multitude of atmospheric parameters, fl ight conditions, airframe geometry, loca- possible ice shapes.
tion and type of ice-protection systems, and the time fl own in icing To illustrate the resulting geometry change due to icing conditions, conditions. Icing atmospheric conditions are not static but change two ice shapes are shown in Fig. 1. They were accreted on a NACA spatially and temporally. Similarly, fl ight conditions and airplane 23012 airfoil model in the NASA Icing Research Tunnel [13].
geometry change as the phase of fl ight changes from climb, cruise, Figure 1a shows an example of a horn ice shape that forms at tem- descent, to approach and landing. Some ice-protection systems do peratures near freezing, where impinging water droplets can fl ow not remove all ice, but leave residual ice on the protected surface.
before freezing. Figure 1b shows an example of a streamwise ice Some thermal IPS may not fully evaporate the impinging water and shape. This type of ice forms at temperatures well below freezing allow water to run back aft of the protected areas where it freezes to where all of the impinging droplets freeze immediately on contact form frozen rivulets, an ice ridge, or other shapes. The amount of time in icing conditions also plays a signi fi cant role as ice builds up on the unprotected surfaces such as unprotected tail, struts, radome, engine pylons, and fl ap hinge fairings.
Clearly, the inputs to icing formation are multidimensional and result in a wide range of potential ice shapes and ice characteristics.
This presents a large challenge to accurately predict the entire range of possible effects on aircraft performance, stability and control, and handling. Over the past 80 years, many efforts have been undertaken to advance the understanding of ice accretion and the impact on aerodynamics. The next section brie fl y discusses some of these advances.
III. Modeling Icing Effects: Wind-Tunnel Results From Two-Dimensional Wing Models The typical aerodynamic effects of ice on airfoils are increased drag, a reduced stall angle of attack, and a reduced maximum lift. In addition to these primary parameters, icing affects the pitching moment and, if equipped with a control surface, the control effec- tiveness and hinge moment.
To understand and model these effects, research organizations have conducted numerous wind-tunnel tests on a number of 2-D airfoil models and some 3-D wing and tail models. Recent iced aerodynamics simulation studies provide a signi fi cant body of knowledge examining the effects of ice accretion (size, shape, roughness, 3-Dity) on drag increase, lift loss, and stall characteristics of a NACA 23012 airfoil at full-scale and subscale geometry and Reynolds numbers [13]. Figure 2 illustrates the change in perform- Fig. 1 Ice shapes accreted at different conditions in the NASA Icing ance characteristics with different ice shapes at Re 1 : 8 10 .
Research Tunnel: a) Ice accreted at temperature near freezing, b) Ice Increase in drag is observed at the initial onset of icing (roughness), accreted at temperature well below freezing.
RATVASKY, BARNHART, AND LEE 203 Fig. 2 Aerodynamic performance comparison for NACA 23012 model with casting simulations (Ref. [13]).
and drag continues to increase as the ice takes shape and continues to Experimental efforts such as these have enabled a comprehensive grow in size. Drag increase is observable at all angles of attack. Ice understanding of the basic fl ow physics associated with various ice affects lift mostly at higher angles of attack in which early stall accretions and helped identify the major aerodynamic penalties occurs. For roughness, this is due to early fl ow separation caused by associated with the particular ice accretion studied. Bragg et al. [16] loss of boundary-layer momentum. For larger ice shapes, early stall provided an in-depth analysis of fl ow fi elds caused by ice geometries occurs because the separation bubble that forms aft of the ice shape categorized as roughness, horn ice, streamwise ice, and spanwise cannot reattach to the airfoil. Before this new stall point, there are ridge ice. Studies such as these are critical in developing and shifts in the lift curve slope and the stall break, and post-stall behavior validating computational fl uid dynamics (CFD) tools to predict iced is clearly affected by the ice. The pitching moment is also affected airfoil and eventually iced airplane, aerodynamic characteristics.
due to the altered pressure distribution over the airfoil, and a pitch up These studies are also bene fi cial in de fi ning ice-shapes character- tendency occurs at signi fi cantly lower angles of attack compared istics that are critical to speci fi c aerodynamic parameters. However, with the clean (no-ice) baseline. to understand the interaction of iced wings and tails for various ice Ice accretions have also degraded control surface effectiveness and shapes on airplane fl ight dynamics, a different approach and test increased the hinge moment loads on 2-D airfoil models. Figure 3 methods need to be employed.
shows the effect of a simulated ice shape on aileron effectiveness ( C ) and hinge moment ( C ). This data are for a forward facing L;A h;A IV. Modeling Icing Effects: Wind-Tunnel Results quarter round, k=c 0 : 0139 at various chordwise locations at Re 6 From Subscale, Complete Airplane Models 1 : 8 10 [15]. The fi rst plot shows the rate of change in lift coef fi cient with change in aileron de fl ection, and the second plot When considering the aerodynamic effects of ice on an airplane shows the rate of change in hinge moment with change in aileron con fi guration, the effects found on 2-D airfoils apply, but are de fl ection. On the clean airfoil, the C value remained relatively expanded upon. In this case, the effects of drag on the airfoils are not L;A constant from AoA 5 deg to 5 deg. As the angle of attack was isolated, but couple into the pitching moment of the airplane. The lift increased from this point, there was a gradual, almost linear, reduc- degradation on the horizontal tail also couples into the airplane tion in aileron effectiveness to AoA 14 deg where the airfoil pitching moment in a static sense but also a dynamic sense by reduc- stalled. At stall, there was a sudden reduction in aileron effectiveness, ing pitch damping. Lift degradation on the wing can affect longi- where C became negative, meaning that de fl ecting the aileron tudinal stability and roll damping. Adding the consequence of L;A trailing edge down at this angle of attack decreased lift. All three ice- reduced control surface effectiveness can lead to an aircraft with shape locations ( x=c 0 : 02 , 0.10, and 0.20) showed reduced aileron substantially degraded stability and control and handling qualities.
effectiveness at AoA 0 deg , when compared with the clean air- These coupling effects are dependent on the airplane design, and foil. However, the reductions were greater when the ice shape was airplane manufacturers are required to demonstrate safe fl ying located at x=c 0 : 10 and 0.20 than when it was located at characteristics with potential ice accretions that may occur in both x=c 0 : 02 . For all three iced cases, the aileron effectiveness started normal operation and failure of the ice-protection system. To manage to decrease at lower angle of attack than the clean model. However, the wide range of possible ice shapes as described in Sec. II, the ice the rate at which it decreased varied with ice-shape location. The shapes tested are de fi ned based on fl ight scenarios, but the onus is on hinge moment data did not show any noticeable trend in the rate of the manufacturer to de fi ne the most critical ice shape for the given change in hinge moment with aileron de fl ection. The C values did scenario. The work that has been and continues to be performed in h;A not appear to vary signi fi cantly with change in angle of attack. Also, aerodynamic studies such as [16] helps identify the critical ice-shape the presence of ice shapes did not appear to have altered the C features and verify applicability of data from subscale testing. In h;A values signi fi cantly either. practice, the ice shapes now commonly tested are 1) ice roughness Fig. 3 Aileron effectiveness and hinge moment on NACA 23012 with simulated ice shapes at various chordwise locations (Ref. [15]).
204 RATVASKY, BARNHART, AND LEE (simulating delayed IPS activation or residual/intercycle ice), 2) runback ice (ice that forms aft of thermal IPS), 3) failure ice (ice that forms on the IPS after a failure, typically large horn ice), and 4) holding ice (ice that forms on unprotected surfaces during a 45 min hold, typically large horn ice).
Two test methods have been used to collect data on subscale, complete airplane models with ice shapes. These are static testing and static and dynamic testing. Each of these will be discussed in the following subsections.
A. Static Testing of Subscale, Complete Airplane Models Static testing produces databases of airplane force and moment coef fi cients for static conditions ( , , , ice case). These control surface Fig. 6 NASA Langley Research Center 1 = 8-scale twin-engine subsonic databases provide valuable insights into the effects that ice has on the transport model.
airplane stability and control. Two examples of this type of modeling are provided to demonstrate this.
Prompted by icing related incidents, C. L. ’ Kelly ’ Johnson this test are shown in Fig. 7. In this case, the ice shapes on the wings investigated the effects of ice on performance, stability, and control reduced stall angle of attack by about 5 deg and lowered maximum on the Lockheed Electra [17]. He found through wind-tunnel tests lift coef fi cient ( C ) by approximately 16% compared with the L max using arti fi cial ice shapes with a scaled model (Fig. 4) that maximum clean data at the test Reynolds number. The difference between the lift was reduced by 32%, drag was increased by 47%, stall angle of iced con fi guration and the no-ice con fi guration at fl ight Reynolds attack was reduced by 4 , and aileron control effectiveness was number is expected to be greater due to increase in C for the no- L max reduced by 36%. Figure 5 shows the effect of icing on roll control.
ice con fi guration. A positive shift in the pitching moment occurred The results were obtained at Re 1 : 6 10 . These data clearly with the ice cases, but longitudinal stability ( C ) was approximately m show the dramatic reduction in stability and control when the Electra the same as the no-ice case until the AoA reached about 8 deg where was iced up.
stall occurred and the aircraft became longitudinally unstable. No More recently, NASA conducted a static wind-tunnel test on a data were acquired with control surface de fl ections other than fl aps in twin-engine short-haul transport to measure icing effects on perfor- this wind-tunnel test, so that control effectiveness with the ice shapes mance and stability [18] at the NASA Langley Research Center could not be evaluated.
(LaRC) 14 by 22 ft wind tunnel. These tests were conducted on a 1 = 8 - scale model (Fig. 6) at Re 1 : 7 10 . Representative results from B. Static and Dynamic Testing of Subscale, Complete Airplane Models Although static testing provides insights into the effect ice has on stability and control, it does not address the effects on dynamic motion. To better understand and model the full range of an iced aircraft fl ight dynamics, data from forced oscillation and rotary balance tests need to be gathered.
To that end, static and dynamic wind-tunnel tests were conducted under various projects on scaled DeHavilland DHC-6 Twin Otter, Cessna business jet [19], and Lockheed S-3B Viking models at the Bihrle Applied Research Large Amplitude Multi-Purpose (LAMP) facility in Neuburg, Germany (Fig. 8). The scale of the models were 6.5, 8.3, and 6.5% and they were tested at Re 0 : 13 , 0.15, and 0 : 20 10 , respectively.
The primary purpose of these research efforts was to understand the effect of airframe icing on fl ight dynamics. This was accomp- lished through the creation of fl ight simulation models that used the aerodynamic databases derived from wind-tunnel test results. This process was used to explore the utility of iced fl ight simulation in pilot training applications using the Twin Otter aircraft. Simulation Fig. 4 Subscale Lockheed Electra with arti fi cial ice shapes (Ref. [17]).
models were developed for the no-ice (clean) and two IPS failure ice-shape con fi gurations. The successful implementation of this approach led to a similar effort using a Cessna business jet. In this case, four con fi gurations were considered: no-ice (clean), ice rough- ness, wing IPS failure ice, and wing runback ice. Lastly, since NASA John H.Glenn Research Center at Lewis Field is modifying a Lockheed S-3B Viking for icing fl ight research, there is a need to understand the impact of potential ice accretions on its fl ight dynamics. In this case, data were acquired for the no-ice baseline con fi guration and icing con fi gurations consisting of large horn ice shapes on the wing and tail leading edges to represent a hold in icing conditions, and spanwise ridge shapes aft of the thermal IPS on the wing and horizontal tail to represent runback ice.
1. Challenges Associated With Subscale, Complete Airplane Model Testing The geometric scale of these models and the associated Reynolds numbers required the development of methods to account for 1) the premature stalling characteristics typical of low Re and 2) ice Fig. 5 Effect of ice on aileron control: full aileron de fl ection (Ref. [17]).
RATVASKY, BARNHART, AND LEE 205 Fig. 7 Effects of ice on longitudinal aerodynamics of twin-engine transport 40 deg (Ref. [18]).
f accretion scaling to represent full-scale iced aerodynamics. To identify the size and position of the arti fi cial ice for the 6.5%-scale develop fl ight simulation models that are representative of full-scale Twin Otter model [20]. The full-scale ice shape for the wing and fl ight characteristics from the low- Re wind-tunnel data, the clean no- horizontal and vertical tail are shown in Fig. 12.
ice wind-tunnel data are projected out along a trend line to the angle For the Cessna business jet, these tests used a full-scale, a 41.7%- of attack in which stall is anticipated at fl ight Reynolds numbers. This scale and an 8.3%-scale wing panel model (Fig. 13) to identify the angle of attack can be determined from other sources such as high arti fi cial ice for the 8.3%-scale business jet model [21]. The full-scale Reynolds number wind-tunnel data and fl ight tests. This method of ice shapes that were tested are shown in Fig. 14.
data extension is shown in Fig. 9 for the Twin Otter and Fig. 10 for the These tests provided valuable insight into the aerodynamic scaling Cessna business jet. Another challenge with this type of wind-tunnel relationships for arti fi cial ice shapes. For example, the failure IPS ice data is the reduced effectiveness of fl aps for high lift (Fig. 9). As fl ap shapes could be geometrically scaled to obtain representative aero- angle increased, the change in lift associated with that fl ap de fl ection dynamics with the subscale model at low Re conditions. The was not as great as seen in fl ight. In this case, corrections were made roughness and runback ice cases for the Cessna business jet proved to based on fl ight data. be greater challenges. For these ice shapes, geometric simulation did The small geometric scale of the complete airplane test articles with ice not produce acceptable results because the full-scale C L max required methods for de fi ning the size and position of arti fi cial ice shapes were greater than the subscale no-ice baseline C . This led L max shapes to represent full-scale aerodynamics. Although research on to determining the size and position of arti fi cial ice on the subscale subscale 2-D airfoils with large ice shapes has shown iced aerody- model that had similar offsets to those observed in the full-scale tests.
namics to be relatively insensitive to Re effects [16], these complete Those results were then projected to the fl ight Re condition similar to aircraft model tests were conducted at much lower Re , and further the no-ice baseline simulation models.
research was warranted. To that end, a series of wind-tunnel tests on full-scale, midscale, and small-scale test articles were conducted.
2. Aerodynamic Model Results The small-scale test articles were representative of the subscale complete aircraft models. For the Twin Otter, this consisted of testing The aerodynamic scaling tests provided con fi dence that the a full-scale, a 42%-scale and a 7%-scale horizontal tail (Fig. 11) to measurements made in the LAMP facility could be used for Fig. 8 Scale models of DHC-6 Twin Otter, Cessna business jet, and S-3B Viking.
206 RATVASKY, BARNHART, AND LEE Fig. 11 Depiction of Twin Otter horizontal tail at various scales.
the most signi fi cant functional dependencies. For the Twin Otter these were angle of attack, angle of sideslip, and fl ap de fl ection.
Incremental coef fi cient data tables were then generated to provide the effects of control de fl ections, dynamic damping, and power effects.
The normal force coef fi cient can be represented by Eq. (1). The C lookup table is based on static inputs of , , . The C N f N BASIC ROT lookup table of increments of C is based on inputs of , N b= 2 V SGN , j j , . The C lookup table of increments of f N Fig. 9 Comparison of Twin Otter no-ice wind-tunnel normal force DE C is based on static inputs , , . Lastly, the C lookup table of coef fi cient data with simulation model values for fl ight Re .
N e f N Q increments of C is based on inputs , q c= 2 V , . The result from N f each of these lookup tables is added to arrive at the total normal force coef fi cient.
modeling full-scale fl ight dynamics. The data acquired at the LAMP consisted of forces and moments along all three axes for static 1. Normal Force Coef fi cient Model for No-Ice Baseline conditions ; ; ; ; ; ; forced oscillations in roll, pitch, and a e r f yaw axes; and steady rotation about the wind vector. These data were C C ; ; C ; b= 2 V acquired for the no-ice baseline and for the iced con fi gurations. The N N f N TOTAL CLEAN BASIC ROT no-ice baseline data were shifted to a full-scale Re stall AoA as SGN ; j j ; C ; ; f N e f DE described earlier. To illustrate key effects that ice had on the aerodynamic characteristics, select aerodynamic model results from C ; q c= 2 V; (1) N f Q the Twin Otter effort [22,23] are presented.
where SGN 1 depending on the sign ( ) of . The differences in the no-ice and ICE02 (IPS failure ice on wing, C. Normal Force horizontal and vertical tail) data sets drove the modeling effort to create separate databases for each con fi guration. Equation (2) shows The normal force coef fi cients were measured over a wide range of that the model structure for ICE02 is the same as the no ice, but the angles of attack, sideslip angles, elevator de fl ections, rotational values within the tables are different.
velocities, and pitch rates during forced oscillations. Data from these measurements were analyzed to determine the model structural dependencies and to insure the preservation of all nonlinear effects.
2. Normal Force Coef fi cient Model for All-Iced (ICE02) The initial de fi nition of the basic force model ( C ) was built from N BASIC C C ; ; C ; b= 2 V N N f N TOTAL ICE 02 BASE I2 ROT I2 SGN ; j j ; C ; ; f N e f DE I2 C ; q c= 2 V; (2) N f Q I2 As an example of the C lookup tables, the normal force N BASIC coef fi cient data for the Twin Otter with fl aps at 0, 20, and 40 deg are presented in Fig. 15 for the no-ice baseline and the ICE02 con fi gurations. Comparing these results, the effect of ice on the normal force is observed mostly at the angle of attack near stall, where there is a reduction in the maximum normal force for all fl ap settings. This effect is most pronounced with the fl aps setting of 20 . The reduced normal force is similar to fi ndings with 2-D f airfoil research, but note that the stall break and post-stall character is similar to the no-ice baseline. This small change is due to the rather benign stall characteristics of the Twin Otter wing and the size and shape of the wing ice tested.
D. Pitching Moment As with the normal force, the pitching moment coef fi cient data were measured over a wide range of angles of attack, sideslip angles, elevator de fl ections, rotational velocities, and pitch rates during forced oscillations. Data from these measurements were tabulated into fi ve databases so that the pitching moment coef fi cient can be represented by Eq. (3). Similar to the C equations, there is a N corollary equation and set of lookup tables for the ICE02 con fi guration. There is an additional term in the C equation for the m thrust effects: C . This database was developed using existing m Fig. 10 Comparison of business jet no-ice C and C wind-tunnel data CT L m NASA Twin Otter fl ight data. with simulation model values for fl ight Re .
RATVASKY, BARNHART, AND LEE 207 Fig. 12 Full-scale ice-shape pro fi les for Twin Otter wing, horizontal tail and vertical tail.
1. Pitching Moment Coef fi cient Model for No-Ice Baseline moments throughout the AoA range with or without ice. However, the ice does reduce the amount of pitching moment affected by the C C ; ; C ; b= 2 V elevator in the 10 < AoA < 10 deg range. This means that larger m m f m TOTAL CLEAN BASIC ROT elevator de fl ections are required to trim the iced aircraft. One thing to SGN ; j j ; C ; ; f m e f DE note is this effect is more pronounced with greater fl ap de fl ections. As fl aps are de fl ected, the in fl ow angles at the horizontal tail increase C ; q c= 2 V; C ; C ; (3) m f m T f Q CT causing separation bubbles to extend and reducing the effectiveness of the elevator. As seen in Fig. 18, the effect of the ice is greater in the As an example of the C databases, the pitching moment m BASIC 40 case and extends to lower angles of attack than when the coef fi cient data for the Twin Otter with fl aps at 0, 20, and 40 deg are f presented in Fig. 16 for the no-ice baseline and the ICE02 fl aps are not de fl ected.
con fi gurations. The effect that ice has is in fl uenced by the fl ap de fl ection . With 0 deg , there is a general positive shift in f f E. Model Equations for Remaining Force and Moments C for AoA > 7 deg , indicating the download from the hori- m For completeness, the remaining force and moment equation zontal tail is reduced for the given elevator de fl ection. Also, the models for the no-ice baseline Twin Otter are shown below in static longitudinal stability ( C ) is reduced signi fi cantly for AoA < m Eqs. (4 – 7). Like the normal force and pitching moments, each of 7 deg , indicating longitudinal instability at this range. With f these equations has identical model structure for the iced case but 20 and 40 deg, the differences caused by the ice are insigni fi cant in different values within the lookup tables.
the linear range, but reduced static longitudinal stability occurs for AoA > 8 deg and for AoA < 4 deg . The wing ice causes the 1. Axial Force Coef fi cient Model for No-Ice Baseline reduced C at positive AoA, whereas the ice on the horizontal tail m reduced the C at the negative AoA.
m C C ; ; C ; A A f A e; f TOTAL CLEAN BASIC e To illustrate the data in the C lookup tables, Fig. 17 is m DE C ; q c= 2 V; C ; C ; (4) provided. Overall, the elevator is effective in creating proper pitching Aq f AC T f T Fig. 13 Full-scale, 41.7%-scale, and 8.3%-scale business jet wing panel models.
208 RATVASKY, BARNHART, AND LEE Fig. 14 Ice shapes tested on full-scale business jet wing panel.
2. Side Force Coef fi cient Model for No-Ice Baseline PC-based simulation environment that fully supported both the Twin Otter and Cessna business jet simulation models development, as C C ; ; C ; b= 2 V well as the analysis and validation activities.
Y Y f Y TOTAL CLEAN BASIC ROT One key objective of these research efforts was to enable real-time, SGN ; j j ; SGN C ; j j ; f Y a f a pilot-in-the-loop simulations to demonstrate icing effects on fl ight dynamics to pilots and engineers. D-Six provided the simulation SGN C ; j j SGN a Y r r r environment to accomplish this because it permits the dynamic C ; pb= 2 V; C ; rb= 2 V; (5) Y f Y f p r linking of other object modules that can control everything from the simulation integration scheme to external graphics and network communications with no requirement to edit the source code. Pilot- 3. Rolling Moment Coef fi cient Model for No-Ice Baseline in-the-loop simulations will be further discussed in the Using Flight Simulation Models V.Csubsection.
C C ; ; C ; b= 2 V l l f l TOTAL CLEAN BASIC ROT SGN ; j j ; SGN C ; ; f l a f a B. Validating Flight Simulation Models C ; ; C ; pb= 2 V; l r f lp f r New fl ight test data were required to validate the fl ight simulation C ; rb= 2 V; (6) l f models. For the Twin Otter, a fl ight test program was conducted with r the no-ice baseline, arti fi cial ice on the horizontal tail only (ICE01 4. Yawing Moment Coef fi cient Model for No-Ice Baseline con fi guration), and arti fi cial ice on the wing, horizontal tail and verti- cal tail (ICE02 con fi guration) [24]. Flight test maneuvers included C C ; ; C ; b= 2 V control doublets, idle-power stalls, steady-heading sideslips, thrust n n f n TOTAL CLEAN BASIC ROT transitions, throttle sweep, wind up turns, and simulated approach SGN ; j j ; SGN C ; j j ; f n a f a and missed approach.
Flight data from these maneuvers were then imported into D-Six SGN C ; j j ; SGN a n r f r r for analysis and validation of the simulation models. Avalidation tool C ; pb= 2 V; C ; rb= 2 V; (7) n f n f p r called Overdrive enabled the validation of the simulation aerody- namic database against fl ight-extracted data using the process illus- trated in Fig. 19. At each time slice, Overdrive extracts aerodynamic V. Simulating Icing Effects moment coef fi cients from the fl ight-recorded time histories, as A. Implementing Flight Simulation Models shown on the right side of Fig. 19. Angular rates are numerically For the purposes of the research, to understand the icing effects on differentiated to obtain the angular acceleration of the vehicle. After fl ight dynamics, the fl ight models described earlier were implem- the removal of the inertial effects, the remainder is nondimension- ented using D-Six, (Bihrle Applied Research, Inc.) a commercial alized to generate the aerodynamic force and moment coef fi cients off-the-shelf product from Bihrle Applied Research, Inc. D-Six is a experienced during fl ight.
Fig. 15 Twin Otter normal force coef fi cient for various fl ap settings Fig. 16 Twin Otter pitching moment coef fi cient for various fl ap with no-ice baseline and ICE02 con fi gurations. settings with no-ice baseline and ICE02 con fi gurations.
RATVASKY, BARNHART, AND LEE 209 element (i.e., pitching moment due to elevator, etc.) is stored and summed as prescribed in the aerodynamic model. By over-plotting the model predicted coef fi cients with the fl ight-extracted total coef fi cients (Fig. 20), differences can be easily identi fi ed. Correlat- ing the discrepancies with the excitation of the individual elements and parameters from the fl ight time history aids in isolating potential weaknesses in the aerodynamic model.
Overdrive was used with the Twin Otter fl ight data to validate the models. The validation effort was reported on [25] but is currently not published.
C. Using Flight Simulation Models One key objective of these research efforts was to enable real-time, pilot-in-the-loop simulations to demonstrate icing effects on fl ight dynamics to pilots and engineers. Although the simulation models Fig. 17 Twin Otter pitching moment coef fi cient increment for no-ice could be run on a laptop or desktop PC, a portable fl ight training baseline and ICE02 due to elevator de fl ection ( 0).
f device called the Ice Contamination Effects Flight Training Device (ICEFTD) (Fig. 21) was developed to be a more effective tool [26].
The ICEFTD consists of a raised platform and framework that supports a pilot seat, a control yoke, rudder pedals, a twin turbo-prop throttle quadrant, a fl ap selector, three fl at-panel monitors for out- the-window graphics, and two additional fl at-panel monitors for instrument panel graphics. The control column is connected to a programmable loader for longitudinal force feedback, whereas the yoke (lateral) and rudder pedals force gradients are provided by spring resistance. Two PCs using D-Six host the simulation models and control the graphics. A third PC, the control loading computer (CLC), controls the electromechanical loader to simulate represen- tative column forces. These PCs are mounted under the fl oor of the device, and the control loader device is mounted to the frame forward of the pilot ’ s feet. A curtain surrounds the ICEFTD to isolate the pilot from external visual distractions. The design is well suited for mobility and usability at various settings, from labs or of fi ces to class rooms or hangars.
Both the Twin Otter and Cessna business jet fl ight models have Fig. 18 Twin Otter pitching moment coef fi cient increment for no-ice been implemented on the ICEFTD and were used for pilot education baseline and ICE02 due to elevator de fl ection ( 40).
f and training, as well as pilot evaluations of the simulation models.
The ICEFTD has been used to demonstrate icing effects on Twin At each time step, fl ight-recorded states, such as angle of attack, Otter fl ight dynamics to over 150 pilots at seminars and short courses angle of sideslip, control surface positions, etc., are used to exercise held by the University of Tennessee Space Institute, National Test the aerodynamic model in accordance with the aerodynamic model Pilot School, Flight Safety International, and several conference speci fi cation discussed previously. Each aerodynamic model exhibits [27]. The ICEFTD was also used in conjunction with the Fig. 19 Overdrive process diagram.
210 RATVASKY, BARNHART, AND LEE Fig. 20 Sample Overdrive result from Twin Otter wing stall maneuver 0, all iced.
f fl ight model validation fl ight tests of the Cessna business jet [28]. efforts are needed with 3-D wings and tails with control surfaces to Simulator sessions were conducted before the fl ight tests to identify determine if 3-D ice features such as scallops need to be considered anomalous fl ight characteristics that were predicted by the simulator. as part of the so-called critical ice shapes.
After the fl ight tests and fl ight model updates based on the new fl ight Regarding icing effects on fl ight dynamics, the research efforts records, Cessna fl ight test pilots reevaluated the simulation models with the Twin Otter and business jet simulation models are an on the ICEFTD. For both airplane types, pilots could readily see the excellent start in that they have provided a method to develop and changes in performance, stall characteristics, and the increased validate icing effects simulation models. These fl ight simulation workload to recover from the stall with the iced airplane. models and the fl ight training device are valuable resources for addi- tional research into icing effects on fl ight dynamics, pilot education, and future fl ight training simulator requirements. But the knowledge gained through the efforts with the Twin Otter and business jet cannot VI. Future Directions be applied across the entire spectrum of aircraft. Further efforts are As described earlier, there is a large assortment of ice shapes needed to identify icing effects on other classes of aircraft, such as possible for any aircraft, each having a range of effects on the regional jets, large transports, and future designs such as blended aerodynamics of individual wings, fuselage, empennage, and other wing body.
surfaces. More important, these ice shapes have a wide range of Understanding these needs, NASA is currently using the Twin effects on the aerodynamics of the aircraft as an integrated machine – Otter fl ight simulation models to develop and test new methods for human system. The research with iced 2-D airfoils has helped reduce identifying icing effects in real time to provide envelope protection to the scope of ice shapes that need to be considered by identifying key avoid loss of control. This effort is being conducted through a NASA ice-shape features and the resulting fl ow fi elds. Similar research Research Announcement (NRA) with the University of Tennessee Space Institute and Bihrle Applied Research, Inc.
NASA is also initiating a new icing effects simulation activity using the NASA Langley Research Center Generic Transport Model (GTM). A signi fi cant database has already been developed using a generic twin-engine large transport model con fi guration to develop recovery strategies from loss of control events. To date, the emphasis of the effort has been on damaged aircraft or failure conditions, such as a rudder hardover. The new icing effects effort will provide another scenario under which loss of control can occur. Flight simulation models will be developed from this database and used in conjunction with adaptive control methods development and testing.
VII. Conclusions The general effects of icing on aerodynamics are increased drag, reduced stall angle of attack and maximum lift, and altered pressure distribution over the airfoil surface. In addition, the stall break and post-stall characteristics can be dramatically different in the iced Fig. 21 ICEFTD used for pilot education and training.
RATVASKY, BARNHART, AND LEE 211 Icing Conditions, National Transportation Safety Board, Rept. A-06- cases when compared with the no-ice baseline. These aerodynamic 48/51, Washington, D.C., 10 July 2006.
effects have been studied using 2-D airfoil sections and much has [11] “ Roll Oscillations on Landing, Air Canada, Airbus 321-211, C-GJVX been learned regarding ice-shape features and the associated and C-CIUF, Toronto/Lester B. Pearson International Airport, Ontario, fl ow fi elds that result in these performance degradations.
07 December 2002, ” Transportation Safety Board of Canada, To understand the integrated effects of icing on aircraft perform- Rept. A02O0406, Gatineau, Quebec, Canada, July 2005.
ance, stability, and control research has been conducted using [12] Russell, P., and Pardee, J., “ Joint Safety Analysis Team-CAST subscale complete aircraft models. From this research, fl ight simul- Approved Final Report Loss of Control JSAT Results and Analysis, ” ation models were developed that incorporate the nonlinear nature of Commercial Aviation Safety Team, Washington, DC, Dec. 2000.
icing effects on the forces and moments along all three axes. These [13] Bragg, M., Broeren, A., Addy, H., Potapczuk, M., Guffond, D., and Montreuil, E., “ Airfoil Ice-Accretion Aerodynamics Simulation, ” fl ight models can be implemented into piloted fl ight simulators for AIAA Paper 2007-085, Jan. 2007.
pilot evaluation and loss of control recovery strategies.
[14] “ Aircraft Ice Protection, ” Federal Aviation Administration, Rept. AC Signi fi cant progress has been made in this area of icing fl ight No 20-73A, Washington, DC, Aug. 2006.
dynamics research, but the knowledge gained with the limited [15] Lee, S., “ Effects of Supercooled Large-Droplet Icing on Airfoil number of models may not be applicable to other airplane con fi g- Aerodynamics, ” Ph. D. Dissertation, Department of Aeronautical and urations. Additional research efforts using other airplane models are Astronautical Engineering, Univ. of Illinois, Urbana, IL, 2001.
needed to extend our current understanding of icing fl ight dynamics.
[16] Bragg, M. B., Broeren, A. P., and Blumenthal, L. A., “ Iced-Airfoil Aerodynamics, ” Progress in Aerospace Sciences , Vol. 41, No. 5, 2005, pp. 323 – 362.
References doi:10.1016/j.paerosci.2005.07.001 [17] Johnson, C. L., “ Wing Loading, Icing and Associated Aspects of [1] Sedor, Joseph M., “ National Transportation Safety Board, Factual Modern Transport Design, ” Journal of the Aeronautical Sciences , Report: Cessna CE-208 Caravan Accident on October 21, 2001 at Vol. 8, No. 2. Dec. 1940, pp. 43 – 54.
Dillingham, Alaska, ” National Transportation Safety Board, [18] Reehorst, A., Potapczuk, M., Ratvasky, T., and Gile-La fl in, B., “ Wind Rept. NTSB ID DCA02MA003, Washington, D.C., 2001.
Tunnel Measured Effects on a Twin-Engine Short-Haul Transport [2] “ National Transportation Safety Board, Safety Recommendation to Caused by Simulated Ice Accretions, Data Report, ” NASA Rept. TM Mitigate the Existing Risk to the Cessna 208 Fleet during the Current 107419, May 1997.
Icing Season, “ National Transportation Safety Board, Rept. A-06-01 – [19] Lee, S., Barnhart, B. P., Ratvasky, T. P., Dickes, E. G., and Thacker, M., 03, Washington, D.C., 17 Jan. 2006.
“ Dynamic Wind-Tunnel Testing of a Sub-Scale Iced Business Jet, ” [3] “ National Transportation Safety Board, Crash During Approach to AIAA Paper 2006 – 0261, Jan. 2006.
Landing, Circuit City Stores, Inc., Cessna Citation 560, N500AT, [20] Papadakis, M., Gile La fl in, B. E., Youssef, G. M., and Ratvasky, T. P., Pueblo, Colorado, February 16, 2005,National Transportation Safety “ Aerodynamic Scaling Experiments With Simulated Ice Accretions, ” Board, Rept. NTSB/AAR-07/02, Washington, D.C., Jan. 2007.
AIAA Paper 2001 – 0833. Jan. 2001.
[4] “ National Transportation Safety Board, Safety Recommendation on [21] Lee, S., Ratvasky, T. P., Thacker, M., and Barnhart, B. P., “ Geometry High-Altitude, Dual-Engine Flameouts on Raytheon Beechjet 400 and Reynolds-Number Scaling on an Iced Business-Jet Wing, ” AIAA Series Airplanes Powered by Pratt & Whitney Canada, JT15D-5 Paper 2005-1066, Aug. 2005.
Turbofan Engines, ” National Transportation Safety Board, Rept. A-06- [22] Gringas, D. R., Dickes, E. G., Ratvasky, T. P., and Barnhart, B. P., 56/59, Washington, D.C., Aug. 2006.
“ Modeling of In-Flight Icing Effects for Pilot Training, ” AIAA [5] “ National Transportation Safety Board, In-Flight Icing Encounter and Paper 2002 – 4605, Aug. 2002.
Loss of Control, Simmons Airlines, D.B.A. American Eagle Flight [23] Barnhart, B. P., Dickes, E. G., Gingras, D. R., and Ratvasky, T. P., 4184 Avions De Transport Regional (ATR) Model 72-212, N401am “ Simulation Model Development for Icing Effects Flight Training, ” Roselawn, Indiana October 31, 1994, ” National Transportation Safety NASA Rept. TM-2003-212115, April 2003.
Board, Rept. NTSB/AAR-96/01, Washington, D.C., July 1996.
[24] Ratvasky, T. P., Blankenship, K., Rieke, W., and Brinker, D. J., “ Iced [6] “ National Transportation Safety Board, In-Flight Icing Encounter And Aircraft Flight Data for Flight Simulator Validation, ” NASA TM-2003- Uncontrolled Collision With Terrain; Comair Flight 3272; Embraer 212114, April 2003.
Emb-120rt, N265ca; Monroe, Michigan, 9 Jan. 1997, ” National [25] Barnhart, B. P., “ An Icing Effects Flight Training Simulator Transportation Safety Board, Rept. NTSB/AAR-98/04, Washington, Aerodynamic Model Development, ” NASA (to be published).
D. C., Nov. 1998.
[26] Ratvasky, T. P., Ranaudo, R. J., Barnhart, B. P., Dickes, E. G., and [7] “ Saab-SF340A, VH-LPI, Eildon Weir, Victoria on 11 November 1998, ” Gingras, D. R., “ Development and Utility of a Piloted Flight Simulator Australian Transport Safety Bureau, Rept. 199805068, April 2001.
for Icing Effects Training, ” AIAA Paper 2003 – 0022, April 2003; also [8] , GE791 “ Occurrence Investigation Report, In-Flight Icing Encounter NASA Rept. TM-2003-212116, April 2003.
and Crash into the Sea TransAsia Airways Flight 791, ATR72-200, B- [27] Ratvasky, T. P., Ranaudo, R. J., Blankenship, K. S., and Lee, S., 22708, 17 Kilometers Southwest of Makung City, Penghu Islands, “ Demonstration of an Ice Contamination Effects Flight Training Taiwan, December 21, 2002, ” Aviation Safety Council, Rept. ASC- Device, ” AIAA Paper 2006 – 0677, 2006; also NASA, Rept. TM-2006- AOR-05-04-001, Taipei, Taiwan.
214233, May 2006.
[9] “ In fl ight Loss of Control due to Airframe Icing Saab 340B, VH-OLM, [28] Ratvasky, T. P., Barnhart, B. P., Lee, S., and Cooper, J., “ Flight Testing 28 June 2002, ” Australian Transport Safety Bureau, Rept. BO/ an Iced Business Jet for Flight Simulation Model Validation, ” AIAA 2002030704, Canberra, Australia, Dec. 2003.
Paper 2007 – 0089, Dec. 2007; also NASA, Rept. TM-2007-214936, [10] “ National Transportation Safety Board, Safety Recommendation to Dec. 2007.
Mitigate the Existing Risk to the Saab 340 Fleet When Operating in