DESIGN AND MANUFACTURE OF AN INTEGRAL TWIST-ACTUATED ROTOR BLADE *

John P. Rodgers: MIT Department of Aeronautics and Astronautics 77 Massachusetts Avenue, Cambridge, MA 02139, Graduate Research Assistant, Active Materials and Structures Lab, Student Member AIAA

Nesbitt W. Hagood: MIT Department of Aeronautics and Astronautics 77 Massachusetts Avenue, Cambridge, MA 02139, Associate Professor, Active Materials and Structures Lab, Member AIAA

Douglas B. Weems: Boeing Defense & Space Group Philadelphia, PA 19142, Technical Specialist, Helicopters Division

* Revised edition of paper presented at the 38^th AIAA/ASME/AHS Adaptive Structures Forum, Kissimmee, FL, 1997, Paper No. 97-1264

Keywords: Helicopter, Piezoelectric, Servo-flap, Rotor, Vibrations, Active fiber composites

ABSTRACT

The objective of this research is to develop an actuated blade for use in the control of rotor vibrations. Active fiber composites are integrated with the composite spar to induce shear stresses within the spar laminate and thus a distributed twisting moment along the blade. This paper details the design of an active blade model based on a 1/6th Mach scale Chinook CH-47D. The design goals are ±2° of blade tip twist with a maximum of 20% added blade mass. The fabrication of a model blade section is included in the paper as well as preliminary structural and actuation testing.

I. INTRODUCTION

Helicopter rotor blades experience significant vibration levels as a result of variations in rotor blade aerodynamic loads with blade azimuth angle. Actively controlled rotor blades are currently being investigated as a means of reducing these detrimental vibrations. A reduction in the vibration levels will improve pilot effectiveness and passenger comfort, and reduce maintenance and operating costs. Performance benefits may also be achieved through increased payload and cruise speed.

Active twist control of the rotor blades may be used to reject the aerodynamic disturbances affecting the blades. These disturbances are the most severe during transitional and forward flight, manifesting in the form of N/rev vertical hub loads. Individual blade control (IBC) of blade twist or angle of attack enables the implementation of active vibration control strategies.

Active fiber composites offer a means of directly twisting the rotor blade. Anisotropic active plies may be embedded within the composite spar of the blade to induce shear stresses which create the twist as shown in Figure 1. In addition, the dense electroceramic active material is positioned near the quarter-chord of the blade section, thus minimizing the weight penalty. The benefits of this concept are the significant actuation authority and high bandwidth. This concept is an alternative to trailing edge flap concepts which induce an aerodynamic twisting moment to deflect the blade. The integral actuation concept eliminates the need for a complex, highly efficient actuation amplification mechanism.

Currently, several other means of reducing rotor vibrations are being pursued. Integral twist-actuated rotor blades using monolithic directionally applied piezoelectric wafers are also being developed¹. Most other research in IBC involves flap actuation. At MIT, a new high-efficiency discrete flap actuator is being developed which relies on piezoelectric stacks². Another flap concept utilizes both piezoceramic and shape memory alloy materials in a mechanism for both trim control and higher harmonic control³. A blade-mounted, piezoceramic stack-driven, hydraulic stroke-amplification system is also being developed for a discrete flap actuator⁴.

Several previous studies have formed the background for the current research. In order to evaluate the feasibility of the integral twist actuation concept, a Rehfield single-cell composite beam model was developed⁵. Interdigitated electrode piezoelectric fiber composite actuators were selected and used in a 1/16 scale benchtop twist demonstration. Active fiber composites have been investigated extensively as a means of anisotropic structural actuation. Recently, a series of characterization tests have demonstrated the integrity of the actuators in a simulated rotor stress environment⁶. A more advanced rotor dynamic analysis of the integral actuation scheme was later performed⁷. This included a systems-level cost-benefit analysis and demonstrated the potential impact of the integral actuation concept.

II. ACTUATOR DEVELOPMENT

The Piezoelectric Fiber Composite (PFC) is an anisotropic, conformable, composite actuator which was developed for embedding within composite laminates. The actuator consists of continuous, aligned, electroceramic fibers in an epoxy-based matrix which is sandwiched between two layers of polyimid film which have a conductive inner surface for applying the driving electric field. More recently, the performance of the PFC system was greatly improved with the change to an Interdigitated Electrode (IDE) pattern, which orients the applied electric field along the active fibers, enabling the use of the primary piezoelectric⁸. A diagram of the actuator is shown in figure 2.

The active fiber composite material system can be tailored to achieve desired performance for a specific application. In this case, performance is measured in terms of induced stress capability of the embedded actuators. Additional consideration is given to the performance in the rotor operational environment.

Several variations of the material system have been investigated for this and other related studies. First, several compositions of the active fibers have been tested, all within the PZT family. All prior publications have centered on the PZT-5H composition. Recently, other compositions have been evaluated which show promise for higher induced stress capability and thus greater actuation authority in the blade. In addition, these compositions offer higher compressive depolarization stress tolerances, which increase the survivability of the actuators. An additional fiber consideration is the geometry. Extruded round fibers and cut square cross-section fibers have been compared. The performance may also depend upon differences in the quality of the fibers resulting from the respective fiber manufacturing process.

The matrix material surrounding the active fibers has also been investigated for performance improvements. The typical hybrid matrix consists of an epoxy resin doped with high dielectric PZT-5H powder and dispersing agents. This increases the relative dielectric of the matrix, thus improving the distribution of the electric field into the active fibers. A combination of conductive fillers and high dielectric powders have been studied to increase the strain capability of the actuators for a given applied voltage⁹. Important manufacturing trade-offs are also factored into the decision including viscosity, and void content and compression in the cured composites. Other factors are the mass and toughness of the matrix system.

Two types of interdigitated electrodes have been investigated for the actuators. The baseline system consists of an etched copper pattern on a kapton substrate produced using a photolithography method. In order to overcome size restrictions and improve production time, a silk-screened electrode has been developed.

A series of characterization tests was performed on the active fiber composites in order to evaluate performance under realistic operating conditions⁶. In addition to providing mechanical, electrical, and actuation properties of the actuators, the tests qualify the actuators for mechanical strain levels within the active blade. The performance of the baseline PZT-5H fiber system with copper/kapton electrodes was shown not to significantly degrade under the following conditions:

• 3000 microstrain static load

• 2.3e8 electrical fatigue cycles

• 1e7 mechanical fatigue cycles at 1250±900 microstrain

with gradual degradation at more severe levels as shown in reference⁶.

More recently, data were collected in order to compare fiber compositions and geometry. The existing data on PZT5H round fibers was augmented with data on PZT5A round fibers, PZT5A square fibers, and PZT4 square fibers. The tests performed were free actuation, laminate actuation, and actuation under static tensile loading.

Testing Procedures

A standard 0.5 inch by 3.5 inch test article configuration was used for the experiments⁶. The interdigitated electrode pattern covers only the center 2.5 inches, leaving space for loading tabs at the ends. The test article is illustrated in figure 3.

Standard manufacturing procedures were used for the samples⁶. The matrix material used for all samples in this study consisted of Epon 9405/9470 epoxy resin doped with 60 weight percent PZT5H powder. Following the cure of the AFC, the sample is poled at twice the coercive field for 20 minutes in an 80°C oil bath (Note: PZT4 may not be fully poled at this level). After aging at least 24 hours, the representative work cycle strain is measured. The representative work cycle consists of a 1 Hz sinusoid which is 600 VDC plus 3000 Vpp (-900 V to 2100 V).

Following these tests, the sample is laminated, and the representative work cycle strain is measured. All strain data to this point is collected with an interferometer. Next, loading tabs and a symmetric pair of strain gages are bonded to the laminate for the actuation under load tests.

For the actuation under load tests, designed to quantify the survivability of the actuators, a constant static tensile load is applied to the test article while an actuation voltage is applied. The representative work cycle is applied (1 Hz sinusoid, 600VDC+3000Vpp) and the peak-to-peak strain is recorded from the opposing strain gages. The measurements are averaged to eliminate any effects of bending from the test. A load is applied in steps each equivalent to 500 microstrain of static strain in the laminate. The actuator induced strain is measured at each step. After reaching a predetermined maximum static tensile strain, the sample is unloaded, and then stepped up again to evaluate damage effects and residual actuation and stiffness properties. For this study, actuation at a 3000 microstrain static load level was selected as the primary metric for operation in the centrifugal field of the blade. This value represents an approximate average strain level of the active plies in the blade. The performance of each sample was measured at 3000 microstrain in the first load cycle. In general, no measurable change was found in a repeated cycle to this load level in any of the samples tested. Some samples were brought up to either 4000 microstrain or 6000 microstrain. For samples that saw this strain level, the performance of the sample was recorded at 3000 microstrain in the second cycle and is compared with that of the first to quantify the accumulated damage in the composite at elevated strain levels.

Results

A summary of the results for the preliminary tests on the samples and for the actuation under load tests is described and presented in the figures below. Figure 4 presents a summary of the actuation tests performed on each active fiber composite sample, before and after lamination. The ratio of the laminate to free actuation provides a relative measure of the induced stress capability for each active composite type. The representative work cycle simulates a possible operating cycle for actuators in the integral blade.

From the data in figure 4, the 5A square doped configuration stands out as having the highest laminate representative work cycle strain of the doped samples at 369 microstrain. This is not a significant increase over the baseline configuration. The PZT 4 square fiber systems showed a higher Rep Cycle ratio (Lam/Free) than the other fiber systems. The PZT 4 square system produces much lower strain values even though it showed the highest ratio, because of low saturation strain levels. The square fiber samples used in the tests had approximately 10% more active material as a result of fiber cross section dimensions. This may account for some of the advantages shown in the laminate actuation levels. The standard deviations for the free and laminated sample data are given in Table I.

Composition	Number of Samples	Free s	Laminate s
5H circ	4	63	18
5A circ	4	75	19
5A sq	3	92	28
4 sq	3	51	22

Table I. Standard Deviation Data for Representative Work Cycle Tests (Units: microstrain)

Actuation under load tests for the active laminates are summarized in figure 5. For each sample type, the representative work cycle strain was for different loading cases at a static strain level of 3000 microstrain. The strain at the second load cycle to 3000 microstrain was measured, following a first cycle loading to either 4000 or 6000 microstrain. The plotted ratios represent a normalization by the initial 3000 microstrain actuation level.

From these data, the 5A square system stands out as having the highest ratios after loading to 4000 or 6000 microstrain. For example, 93% of the representative work cycle actuation was retained after loading to 4000 microstrain static tensile load. The 5H circular undoped also performed well in the tests up to 4000 microstrain, while the 4 square system did not stand out. Each data point was based upon either one or two samples of each type so no standard deviation data is available.

Another major factor in the selection of the fiber composition is the compressive stress depolarization limit. Table II lists the approximate maximum stress levels achievable before significant depolarization of the piezoceramic results. The values are taken from experimental studies at Penn State¹⁰ on bulk ceramic. The significance of the compressive strain limits depends on the expected maximum strain levels in the hover tests of the integral blade. The depolarization phenomenon is in reality dependent on both the stress and the electric field applied to the ceramic. Thus, the combination of hover loads, applied field, and actuator induced stress and strain complicate the prediction of depolarization conditions for the active plies.

Composition	Compressive Stress Limit	Compressive Strain Limit
PZT 5H	30	270
PZT 5A	50	450
PZT 4	150	1100

Table II. Compressive Stress Depolarization Limits (Units: MPa, microstrain)

Material Selection

Based on the characterization test results and predicted loads for the active plies in the hover testing (to be described in Section IV), the PZT5A composition was selected. While the square fibers performed the best, especially in damage tolerance, both square and round 5A fibers have been used in the preliminary pack manufacture for the section test. Performance of the blade-size packs has shown that the circular fibers are currently the best choice for the material system. Manufacturing difficulties and damage to the copper electrode pattern have precluded the use of square fibers. For the matrix, data collected have shown that the advantages of high dielectric fillers are outweighed by the increase in viscosity and the resulting gap between the electrodes and the fiber surface. Samples having an undoped epoxy matrix had the better performance than doped samples. Although actuators with the screen printed electrodes have demonstrated improved actuation performance as a result of a closer proximity of the conductor to the active fibers within the composite, the screened electrodes have had problems with ink cracking within the active fiber composites. Thus, the copper/kapton electrodes have been selected for the blade packs.

The effective properties of the final active ply configuration are presented in Table III. These stiffnesses were determined from a Uniform Fields microelectromechanical model⁸ of the active fiber composite using experimental stiffness data from the composite to adjust the active fiber compliance (1.22 times bulk PZT5A compliance). In keeping with standard practice for composite materials, the longitudinal or '1' axis is aligned with the fiber direction. The density was determined from the measured mass of prototype active fiber packs. All of these properties represent the average expected properties of the active fiber composite packs, including the copper electrodes and Kapton substrate as well as the PZT5A fiber and epoxy resin.

fiber	PZT 5A
resin	Epon 9405/9470
thickness (mm)	0.165
density (kg/m^3)	4060
Q11 (GPa)	33.6
Q12 (GPa)	7.54
Q22 (GPa)	16.6
Q66 (GPa)	5.13

Table III. Active Ply Properties

III. ACTIVE BLADE MODEL

A relatively simple mathematical actuation model was developed to analyze the integral twist concept⁵. The model employed is a modified Rehfield thin-walled, single-celled beam model¹¹ and is used to predict the static twist capability of prospective blade designs. Cross-sectional properties of the beam are calculated and used in the formulation of the finite element model.

The Rehfield beam model was augmented with anisotropic active plies for predicting structural actuation. The following assumptions apply to the single-celled beam model:

• Cross-sectional shape is maintained during deformation

• Wall thickness is small in comparison to other dimensions

• Transverse, in-plane, normal stresses are negligible

• The twist rate varies spanwise and acts as a measure of the cross-sectional warping

There is no restriction on the cross-sectional shape of the beam. The properties of the cross-section may vary around the contour. The actuation stress resultants are developed from the piezoelectric constitutive relations and Classical Laminated Plate Theory applied around the contour.

For a given cross-section, the model is used to predict the effective stiffness, mass, center of gravity, and actuator-induced forces and moments of the single celled beam. In order to account for the effects of the other components of the model blade, lumped stiffness and mass terms are added to the section stiffness matrix. This includes the contributions of the leading edge weights, foam core material, and fairing. This technique of augmenting the simple single-celled beam model has been validated through comparisons with a more complex and extensive multi-celled beam model¹². The single-cell beam model enabled a rapid analysis capability for prospective blade designs.

A beam finite element model was implemented using the cross-sectional properties from the Rehfield model to develop the section stiffness matrix and forcing vector. This also allows for piecewise variations in the cross-section of the beam. In addition, distributed loads may be applied. A static solution for the deformation of the beam yields the effective twist distribution used as a primary metric in the design process. The deformation solution may also be used to find the ply stresses at any point on the beam.

IV. BLADE DESIGN

The approach used to design the active rotor blade was to select an existing rotor blade design as a baseline configuration and then modify it to incorporate active plies. The baseline configuration selected was a 1/6^th Mach scale CH-47D blade developed for wind-tunnel testing at Boeing. This configuration was selected because it was an appropriate size for the anticipated testing and because of the significant experimental data and manufacturing experience available at Boeing.

Baseline Blade

The model CH-47D blade, shown in figure 6, has a span of 60.619 (measured from the center of rotation) and a chord of 5.388 inches. It is designed to be used on a fully articulated hub with a single pin located at 0.15R (15% radius). The blade has built-in 12° linear twist and tapers in thickness from 0.85R to the tip.

The primary structural member of the model blade is a cocured "D" spar , while the aft fairing is added in a secondary cure. This baseline blade design calls for Kevlar fabric, graphite tape, and fiberglass tape composites on balsa and Nomex honeycomb cores. The number, material, and orientation of the composite plies was chosen to produce blade section properties which the full-scale CH-47D blade properties reduced by the appropriate factor for Mach scaling. All of the plies are oriented either aligned with the blade longitudinal axis, to maximize their contribution to axial and bending strength and stiffnesses, or at +45° and -45° to the longitudinal axis, to maximize their contribution to shear and torsion strength and stiffness.

For the active blade design, several of the materials used were updated to reflect current best practices. Most notably, fiberglass fabric replaced Kevlar while Rohacell foam replaced both balsa and Nomex honeycomb cores. The modern materials to be used in the current design are summarized in Table IV along with some typical properties^13,14. A film adhesive is also used around the core and for secondary bonds.

fiber	E-glass	S2-glass	IM7
resin	F-155	SP381	SP381
form	fabric	tape	tape
manufacturer	Hexcel	3M	3M
thickness (mm)	0.118	0.225	0.138
EL (GPa)	20.7	47	165
ET (GPa)	20.7	12.1	8.8
nLT	0.15	0.28	0.34
Distinguishing quality	Durable Fabric	High Strength	High Modulus

Table IV. Passive Composite Materials

Active Blade

The basic design concept of the active blade is to replace some portion of the passive composite materials in the blade spar with active plies, as illustrated in Figure 1. The active plies may be embedded in the upper and lower spar laminates and possibly in the web. Regardless of the position around the spar contour, the active plies will be oriented to induce shear strains which will in turn twist the blade.

Actuators can be manufactured with the piezoelectric fibers aligned at any angle. For this application, the fibers are aligned at a 45° angle to the longitudinal axis of the actuator pack, while the longitudinal axis of the actuator pack is aligned with the longitudinal axis of the blade. In the blade spar, the active plies alternate from +45° to -45°. By actuating the +45° and -45° plies opposite to each other, i.e., extending the fibers in the +45° plies while contracting them in the -45° plies, the actuators will produce a shear deformation of the spar laminate. By then coordinating all the actuators in the spar, the result is a twisting of the entire blade.

Design Objectives and Constraints

In general, the integral blade design was intended to have nominally the same elastic properties as the baseline blade. In updating the materials and lay-ups, the target stiffness and inertial properties of the blade were maintained with the exception of the torsional stiffness. A reduction in torsional stiffness allows for greater blade twist for a given level of actuator authority. Thus the target torsional stiffness was set at 50% of the nominal baseline value. Results from a rotor dynamic analysis of the modified blade suggest that the changes would have no detrimental effects⁷.

The design process was directed at achieving a number of goals for the model blade. While the model blade will only be used for hover testing at the MIT Rotor Test Stand Facility, it was designed to meet strength requirements representative of a full-scale service environment in order to demonstrate the viability of this approach. The following is a summary of the design

objectives:

• ±2° tip twist

• <20% mass increase over baseline model blade

• 50% of nominal torsional stiffness of baseline model blade

• Equivalent axial, bending, and shear stiffness of nominal baseline blade

• Section cg at quarter-chord

• Ply strain levels do not exceed allowables

• Provide passive load transfer path in plies surrounding active plies

• Counter crack propagation and delamination failure modes

• Manufacturability

• Electrically insulate active plies

The variables in the design include the amount and placement of the active material and the passive ply lay-up of the spar. Other details of the design that are considered include manufacturability, actuator pack design, power distribution, and interconnections with the hub.

Blade Design Procedure

Candidate configurations for the active blade were developed by adding various numbers of active composite plies to the baseline spar laminate. For each configuration, the passive lay-up for the blade was then modified to match the target stiffness values at a representative section. Leading edge mass was then added to properly locate the center of gravity of the section at

the quarter-chord.

A structural analysis was incorporated into the design process in order to ensure structural integrity of the candidate active blade configuration. This analysis calculates both static and fatigue strains at 13 checkpoints on the cross-section, shown in figure 7, at each of 19 spanwise stations. The checkpoints include locations where chordwise and flapwise bending stresses and torsional shear stresses are high, as well as transition points in the spar laminate. The spanwise stations are well distributed and include critical ply drop-off locations. At each checkpoint, the fiber-direction ply strains are calculated for comparison with established design allowables for each material.

The static loads used in this analysis were estimated from the design loads for full-scale blades, including the assumption of 20% rotational overspeed in addition to a 1.5 ultimate load factor. The fatigue loads were based on analytical predictions from Boeing’s TECH-01 analysis¹⁵ for high-speed forward flight. The centrifugal loading includes the contribution of the added blade mass from the distributed actuators.

Spar Lay-up Design Refinement

The analysis of preliminary integral blade configurations established a general set of requirements for the design of the blade. These provide direction for the design studies in which various parameters are investigated for possible improvements offered to a given design configuration. The requirements, formed from the previously described design constraints, limit the design space and thus simplify the design process.

The results of preliminary analyses showed that two to four active plies could be integrated into the spar and web laminates. The required tip twist set the minimum actuator authority while the mass constraint and constraints on the total spar laminate thickness limited the maximum. Active plies were excluded from the nose of the spar to avoid the high curvature and high strain levels resulting from chordwise bending.

In order to ensure the structural integrity of the blade section, maintain actuation capability after loading, and insulate the actuator electrodes from each other, passive plies are required to surround the active plies. Characterization studies have shown that passive composite plies aligned with and adjacent to active plies increase the effective strength of the active plies by providing a load transfer path around cracks in the ceramic fibers, and may also inhibit the propagation of cracks in the active plies. These neighboring passive plies increase the shear stiffness of the spar laminate and thus reduce the strain levels in the active plies resulting from external

loading.

Chordwise Actuator Placement

Another design parameter is the location of the active plies around the spar contour. Near the leading edge (refer to Figure 6), the increasing curvature of the spar surface restricts the active fiber composites. Excessive curvature can create significant bending stresses in the active fibers. The relatively large diameter of the fibers is the true source of the problem. Another concern near the leading edge is increased loading from chordwise bending stresses. There is no rigid limitation on placement toward the aft of the spar. However, a weight penalty must be considered when placing active material aft of the quarter-chord. Given these constraints, the optimal active ply location is between 0.024c (0.13 inches from the nose) and 0.380c (or the spar heel).

Another consideration in the chordwise placement is the total width of the active ply. Each active ply is composed of segmented active fiber composites. These active plies have a certain dead area associated with the interdigitated electrodes. The percentage of dead area in the active ply is reduced for greater widths. Thus for active plies placed in the web, the relative dead area is significantly greater (roughly 3 times) than for the upper and lower spar. Actuator placement in the web also complicates manufacture in terms of wiring since the passive web plies wrap around the upper and lower spar surfaces. Web actuation will be considered for comparison with upper and lower spar laminate actuation.

Spanwise Distribution

Several factors constrain the spanwise distribution of active plies in the model blade. Adding active material inboard changes the angle of attack over a greater length than adding the same material outboard. In addition, the centrifugal loads at the blade root are less. On the other hand, strains in the active material will be generally higher inboard as a result of centrifugal loads. The spanwise placement of the added actuator mass also effects the dynamic modes of the blade. A uniform distribution of mass has a lesser effect than concentrating the mass in any one region.

From 0.27R inboard to the blade root, the section stiffness of the blade significantly increases due to added spar plies to react the high centrifugal loads. As a result, active plies are only considered outboard of 0.27R. For the outboard extent, the important considerations are the induced aerodynamic forces and the increased centrifugal loads near the root. For a marginal increase in outboard extent, the blade twist is increased from the location of the added active ply to the tip, but the centrifugal load is also increased at the root. The further outboard the marginal increase is considered, the greater increment in the root loads and the lesser the increment in the total length of the blade with increased angle of attack. An increased lift coefficient counteracts this latter effect until approximately 0.98R where tip losses begin to take effect. Increasing the spanwise extent from 0.85R to 0.95R will increase the root loads by 10% while increasing the aerodynamic forces by 20%. For the current design, the maximum outboard extent of 0.95R was selected. Adding active material outboard of 0.95R becomes increasingly difficult because of the taper in the airfoil section.

Another trade study compared two designs with the same amount of active material but uniformly distributed (0.27R-0.95R) in one case, and concentrated over half the span (0.44R-0.78R) in the second case. Nominally, the tip twist should be equal for the two cases. The twist distribution would differ, with the second design achieving the twist inboard of the first design, thus providing better performance. However, the doubling of the active material in a given spar section does not double the induced torsional moment. This is a result of the increase in stiffness due to the active plies. As the relative contribution of the active plies to the total blade stiffness increases, the effectiveness of the active plies decreases as induced stress actuators. The active plies move closer to the limitation of induced strain actuation. In addition, an extra passive ply would have to be added for reliability.

The two designs are compared in figure 8. The induced twist and section lift are shown as a function of span. Clearly, the initial uniform distribution achieves greater performance with greater twist over the entire span, and therefore greater section lift. The total lift for the double actuation case is 77% of the uniform case. In theory, a doubling of the twist rate would have enabled an 8% increase in the total lift. Although other distributions of active material could result in better total performance, constraints on the dynamic modes of the blade and on the complexity of the manufacture support a uniform distribution.

Design Comparisons

Four designs were developed with differing levels of actuator authority in the blade. In each, the amount of actuator material around the spar is different. Each design attempted to meet each of the design constraints, while maximizing the induced twist. The four types are: 2 active plies in upper and lower spar walls, 2 active plies in upper and lower spar walls and the web, 3 active plies in the upper and lower spar walls, and finally 4 active plies in the upper and lower spar walls. Figure 9 depicts the spar lay-up for each case. The model-predicted induced moment and twist are compared for each case in figures 10 and 11 using a nominal 1100 microstrain for the active plies. Figure 11 also includes dashed lines representing the initial design constraints.

Table V provides a quantitative comparison of the designs. The section mass, induced moment, and tip twist are compared for each case. The difference between the active mass and total mass of each design is the amount of passive composite material required to meet strength and stiffness constraints, and the amount of nose weight needed to properly locate the center of gravity. The induced twist moves toward an asymptote with increased active material. This occurs as the ratio of active mass to passive structural mass increases. The actuation of the blade spar approaches the induced strain limit of the active material.

	2-ply	2+web	3-ply	4-ply
active mass (kg/m)	0.171	0.196	0.257	0.342
total mass (kg/m)	0.504	0.545	0.568	0.652
actual moment (N-m)	5.44	7.62	7.41	9.20
actual twist (deg)	3.10	3.75	3.88	4.41

Table V. Comparison of Integral Blade Designs

In general, the 2-ply design did not have sufficient authority to meet the desired twist. The addition of active material to the web lay-up was found to be very effective but was limited by manufacturability, increased torsional stiffness, and center-of-gravity related weight penalties. The 4-ply design was most effective, but involved excessive added mass. The 3-ply design nearly meets all of the requirements. One drawback of this design is the unbalanced spar laminate which results in an extension/twist-coupled blade. This was not found to have any significant effect on the predicted blade performance.

Actuator Pack Geometry

The active plies which are distributed in the upper and lower spar laminates are divided into actuator packs in order to simplify manufacturing and increase reliability. In theory, segmenting each active ply into multiple sections does not reduce the overall performance of the blade. However in practice, dead area around the perimeter of each pack slightly reduces the effectiveness. From a manufacturing standpoint, making a pack with a length on the order of 0.15 m is much easier than one which is closer to 1 m. The smaller packs allow for a more reliable manufacturing process. The process is repeated a larger number of times which improves the quality of the product as skills improve. Extra packs can be manufactured at little additional cost so that only those with the highest performance are integrated into the final blade. Reliability is also improved during blade operation since packs can be wired independently. This reduces the effect of a short-circuit failure on the overall performance of the blade.

The fact that the packs will be independently wired creates an additional design constraint. The smaller the actuator segments, the more packs there will be, and the more wires will be required. The total length of each active ply is 1.047 m of which 6 are required. Segmenting each layer into 7 pieces results in a repeated length of about 15 cm and a total of 42 packs. This design maintains a pack size which is manufacturable while increasing the number of leads and internal connections to a high but reasonable level. The electrical component of the design will be discussed in the next subsection.

Blade strength and endurance concerns also drive the actuator pack design. One issue is the transition between the active plies and the adjoining passive plies on the inboard and outboard ends. Rather than dropping all 3 active plies at the same blade station, each ply is shifted by 12.7 mm to stagger the transitions. Another concern is a delamination failure within an active ply which could propagate through the brittle layer of ceramic fibers. Allowing for a gap (5 mm) between adjacent packs will enable glass plies from above and below the active ply to fill the space in between, thus creating a discontinuity in the active material. Figure 12 illustrates the segmented active ply design. Although the gaps increase the effective dead area in the active ply, the reliability of the blade is increased.

Actuator Pack Electrode Design

Another component of the active fiber composite is the kapton electrodes. The electrode layers sandwich the active fibers and matrix material in order to deliver the electric field to the fibers. The interdigitated electrode pattern was designed using the same rules as in previous actuator designs⁸. Figure 13 illustrates the electrode pattern and pack size. The alternating polarity lines (fingers) are spaced 1.14 mm apart and are 0.18 mm wide. The fingers connect to rails which run along the perimeter of the pack. On one side is a positive rail and on the other is the negative rail. The rail is 1.14 mm wide. A 1.14 mm gap is placed between the ends of the fingers and the opposing rail.

Electrical Connections

The electrical connections from the packs to the leads supplying the power are placed along the web of the blade. This arrangement allows direct access to the aft edges of all of the packs. The leads are not embedded within the spar core, which simplifies the assembly but places mass aft of the quarter-chord. Placing the leads along the web also allows for connections to be made after the spar is cured.

A simple method to distribute power to the packs would rely on a positive and negative bus along the web. However, this arrangement does not allow access to the individual packs after the blade is cured. Independently connecting one flap from each pack would enable disconnection of a single pack in the event of an internal short circuit. However, common lead will deliver a constant DC voltage. In the event of a short between the DC voltage and some external conductive material at the edge of a pack, there would be no means of disconnecting the pack.

Individual connections to each electrode flap require 84 leads along the web of the blade. Another option considered involved surface mounted jumper connections for each pack to a main power bus. This concept was rejected because of the high voltages involved (4000 V) and the associated safety concerns. Individual connections allow each pack to be tested in situ, poled in situ if necessary, and actuated independently. Although it is not required for the initial hover testing, independent control of the packs allows for more degrees of freedom in the control of blade vibrations.

The rails are in turn connected to the external leads through the electrode flaps. One flap is connected to each of the two rails. Both flaps can be placed along a single edge in order to constrain all connections to the edge of the pack along the web of the spar. While the flap connecting the forward rail (toward leading edge) is constrained to the end of the pack where the rail terminates, the flap on the other rail may be placed at any point along the length. This degree of freedom allows for placement to avoid interference with the electrode flaps from other packs in other active plies. Figure 14 depicts the relative position of the electrode flaps from the packs in each of the three active plies in the upper and lower spar laminates.

A lightweight, low volume solution for distributing the power to the packs is a flexible circuit. The flex circuit consists of multiple parallel lines of copper arranged in layers with kapton insulating layers in between. Each line of the flex circuit terminates with a solder pad to be connected to a particular electrode flap. The area of the copper cross section was designed to carry the current required to drive a single pack (100 mA), and also to carry the total amplifier current in the event of a short circuit (1 A). The space between lines of copper within a layer was designed to meet the 4000 Vpp voltage requirement. An epoxy-based adhesive separates adjacent copper lines, while layers of kapton separate adjacent layers. Given these constraints, the 1 oz./ft.2 copper with a 10 mil width was selected for the conductor, with two layers of 1 mil kapton separating layers. This allows for 14 lines across the width of the web, with an allowance for solder pads along the upper and lower edges to interface with the electrode flaps as shown in figure 15.

A total of 6 flex circuit layers are required for the blade. Each layer drops off after reaching 14 connection locations such that only one layer reaches the most outboard packs.

At the inboard end of the web, the flex circuit exits the blade and terminates below the hub. Here each flex circuit layer in connected to a common printed circuit board (PCB). The PCB serves as a matrix connector, connecting individual lines from the flex circuit to a total of 5 source wires. The 5 high voltage lines are connected to amplifiers in the stationary frame through a slip ring. Of the 5 lines, one is a common DC signal, one is for the +45° active plies in the upper spar, one is for the -45° active plies in the upper spar, and the remaining two are for the active plies in the lower spar. This enables signals of opposite phase to be used to drive the +45° plies and -45°. In addition, actuation of the upper and lower spar laminates is independently controlled.

Design Summary

A summary of the predicted properties for the design relative to target values is given in Table VI. The final active blade design features three active plies in the upper and lower spar laminates between 0.27R and 0.95R. Figure 16 presents the effect of design actuation on the predicted twist, twist rate, flapwise bending, and chordwise bending distribution during steady hover.

Actuation is shown for maximum positive and maximum negative induced twisting moments. During a representative work cycle, the blade would oscillate between the two curves. Note that these predictions are inaccurate inboard of 0.22R due to the thin-walled beam restrictions of the modified Rehfield model. In addition, the initial pretwist of the blade is not included in this data.

Property	Percent of Target
tip twist	87%
blade mass	98%
EA	107%
EIc	91%
EIf	97%
GJ	107%

Table VI. Predicted Properties of Design

The ultimate design strains predicted for the active plies are 5100 microstrain tension and 3600 microstrain compression, both of which occur at blade station 0.337R. While such strain levels would permanently degrade the performance of the active plies, they would not produce structural failure. Considering the 1.5 ultimate load factor and the limited area over which these high strains occur, this is certainly acceptable for hover testing and potentially for a flying design as well. The most severe fatigue strains predicted for the active plies are 680±660 microstrain, which is below the level at which minimal degradation in 10e6 cycles was demonstrated during the characterization tests.

The peak actuator strains predicted for steady hover testing are 819 microstrain tension and 173 microstrain compression in the extremes. If a load factor of 2 is applied to all loads except CF to account for unsteady effects, the peak actuator strains are 1277 microstrain tension and 685 microstrain compression. Under these conditions, minimal damage accumulation is expected for the active composites. The greatest risk for the hover testing of the model blade is localized compressive stress depolarization of the active fibers (PZT5A depolarization becomes significant above 450 microstrain). Note that the strain predictions do not include the effects of actuation, which will increase the maximum stress levels in the active plies. A more complex analysis including thermal prestress, applied electric field, and temperature effects would provide a more accurate gauge of the state of polarization in the piezoceramic.

V. MANUFACTURE

Both the active plies and the integral blade are manufactured in the Active Materials and Structures Laboratory at MIT. The following subsections provide an overview of the manufacture of the active fiber composites and the process of embedding them as active plies in the integral blade.

Actuator Pack Manufacture

The procedure for manufacturing the active ply segments or packs has been developed from previous experience with active fiber composites. The general concept remains the same: form a composite with a single layer of aligned active fibers in epoxy matrix and interdigitated electrodes. Registration of the top and bottom electrode patterns and compression of the lamina are required to maximize performance. In addition, void content must be minimized to reduce dielectric breakdown risks.

The fibers and electrodes used in the manufacture are prepared in advance. Each batch of fibers is first evaluated at the Materials Research Lab at Penn State University¹⁶. To ensure acceptable fiber quality, the average relative dielectric constant of 10 randomly selected fibers must exceed 850. The accepted fibers are then cut to a length of 65 mm and are divided into 3.88 g groups for each pack. The fibers are then rinsed in acetone to remove any surface contaminants.

For the electrodes, each set is inspected for discontinuities and is checked for pattern distortion. Then a 25 mm thick, 6.4 mm wide strip of copper is bonded to the electrode flaps such that one end of the strip is flush with the inner edge of the rail. The bond is formed with conductive epoxy (Epotek 410E). The electrode flaps on the upper electrode are then covered with GNPT to the outer edge of the rail to enable separation of the upper and lower flaps after the cure. The flaps on the lower electrode are covered after the mold is formed around the perimeter.

The packs are produced using a hot press and vacuum combination. First, the bottom electrode is placed on the cure plate, aligned with marks printed around the perimeter of the kapton. The top electrode is aligned to the top cover, which will mate with the cure plate using alignment pins. A 5 mm wide strip of ±45° E-glass fabric is placed along the inner edge of the longer electrode rails of the bottom electrode. Along the shorter sides, the E-glass is placed 1.125 mm inside the rail, which is the edge of the active area of the electrode pattern. The fabric fills the dead area under the rails in order to provide a more structural interface with adjoining passive plies as well as reduced weight. A mold is formed around the perimeter of the E-glass on the bottom electrode layer using kapton tape. This tape also serves to hold the E-glass in place. The fibers are then placed within the mold by hand, and are adjusted to achieve roughly uniform spacing and proper alignment. This is illustrated in figure 17. The cure plate is then heated to approximately 50°C. The matrix consisting of Shell Epon 9405/9470 epoxy and BYK A530 air release agent (0.5%) is added to the fibers in the mold. Only enough epoxy to coat the fibers and fill the mold is added. Once the epoxy has spread throughout the mold, the fibers are adjusted to form a uniformly distributed, single layer with no crossed-over fibers.

Next, the top plate, which has been preheated to 60°C with the top electrode attached, is positioned with the alignment pins and is suspended 5 mm above the fiber and epoxy on the cure plate. A pair of cross beams bolted to the top plate and supported at the edges of the cure plate supports the top cover. A vacuum bag which was attached to the perimeter of the top cover in advance is then sealed to the cure plate below. A vacuum is then pulled on the sample to degas the matrix and prevent any bubbles from being trapped by the top electrode. A photo of this vacuum/hot press mechanism is shown in figure 18. The support beams are then lowered while the vacuum is applied until the top electrode comes in contact with the fibers and matrix below. Once this has occurred, the support beams are used to apply downward pressure (100 kPa) to the top cover in order to maintain compression in the composite as the vacuum is released. Release of the vacuum allows any voids within the composite to collapse. The temperature of the cure plate is then elevated to follow the cure cycle of 3 hours at 120°C. The cure plate allows for 4 packs to be manufactured simultaneously.

Once cured, the excess kapton and E-glass is trimmed from the perimeter of the pack and the electrode flaps are separated. The GNPT and other material between the flaps is removed. A temporary lead is soldered to the end of the copper strips for poling and proof testing. Afterwards, the flaps are folded to form a 90° angle with the plane of the pack so that the flaps will lay in the plane of the web when embedded in the spar laminate. The copper strip and electrode flap on the outside of the bend are trimmed to a length of 2 mm while the length of the inner is 6 mm. Thus there will be 4 mm of exposed copper to facilitate bonding to the flex circuit after the spar cure.

Proof Testing

Each pack is subjected to a qualification test prior to fabrication of the model blade. First the actuator is visually inspected to ensure minimal void content in the matrix and proper registration of the upper and lower electrodes. The capacitance is checked and compared with an expected value. The actuator is then poled for 20 minutes at 80°C and 4000 V in air. Next it is cycled to a representative work cycle of -1200 V to 4000 V for an analysis of induced strain capability. Any localized dielectric breakdown within the packs can be repaired by removing carburized material and filling with 5-minute epoxy. In order to pass the test, each actuator must surpass a peak-to-peak longitudinal strain level of 1100 microstrain without any recurring breakdown problems. The packs with the best performance are selected for the blade. Thus the qualification tests serve as a means of improving the integral blade actuation system reliability.

Integral Blade

The model blade spar is manufactured using procedures developed by Boeing Helicopters. The details of the procedure are proprietary with the exception of the integration of the active plies developed in conjunction with MIT for this project. In general, the spar consists of a foam core wrapped with composite laminae to achieve the designed lay-up for the nose, upper and lower spar walls, and the web. The active plies are incorporated into this lay-up procedure such that the electrode flaps fold onto the outer surface of the web. Thus the packs are embedded between layers of S-glass and E-glass which end at the web, allowing the electrode flaps to fold over the E-glass web plies. Each of the packs in the active plies is positioned along the aft edge of the upper and lower spar surfaces with the designed 5 mm gap between consecutive packs. This is illustrated in Figure 19.

Once all of the packs are embedded and the lay-up is complete, each of the electrode flaps is covered with GNPT to protect them from flowing epoxy. The entire spar assembly is then positioned in a two-part mold which has a filler block in the fairing portion. The spar is then cured at 120°C for 90 minutes. Following the cure, the mold is opened and the protective covers are removed from the electrode flaps.

Next, the solder pads on the flex circuit are bonded to the corresponding electrode flaps using conductive epoxy. A layer of film adhesive is used between the solder pads to structurally bond the flex circuit to the web. The conductive epoxy is applied sparingly to minimize any flow of excess conductive epoxy during its cure. Pressure is applied to the web through a rubber interface to ensure good bonding of all connections. The epoxy is cured at 100°C for 1 hour.

With the flex circuit attached and connected, each of the packs can be tested in situ. Capacitance checks are used to monitor the integrity of each pack. A photo of the active section of the spar is shown in Figure 20.

VI. SECTION TEST RESULTS

An active blade section was tested in order to evaluate the design and manufacture, and to allow for improvements in the full model blade. The blade section has 12 actuator packs incorporated in the typical model blade section at midspan. Note that only 8 of the packs passed the proof test for free strain actuation. The average for the 12 packs was 1100 microstrain at 1 Hz and a for a 4 kV cycle. The total length of the section was 0.60 m from the root pin. The active length extended from 0.27R to 0.46R, or 0.30 m. Preliminary twist actuation data have been collected for the spar section. Actuation tests also allow for comparisons with the model, but more importantly demonstrate the effectiveness of the design and manufacturing process.

An aluminum tip fixture was cured into the outboard end of the spar section. The composite plies of the spar form a lap joint with the mandrel, which has the same cross-section as the foam core within the blade. Outside the blade, the aluminum is machined to create a flat surface for interfacing with the hydraulic grips for future structural testing. The foam core of the spar was instrumented with 6 full strain gage bridges to be used in future testing.

The twist capability of the active blade section was measured in a benchtop test. One end of the blade was clamped, while the other was free. A pair of laser displacement sensors was used to measure the twist angle at the blade tip as shown in figure 21. The active plies were driven with a 20 Hz sinusoid having a 600 V DC offset and an amplitude of 2550 V peak-to-peak. A 2-channel audio amplifier was used to drive a pair of 25:1 transformers to provide inverse AC components for the +45° and -45° plies, while a separate DC supply provided the offset. The resulting twist performance is plotted in figure 22. The estimated constant twist rate over the active length of the spar is plotted in a hysteresis loop. The total tip twist was 0.38° peak-to-peak for the applied voltage cycle with an average twist rate of 1.26°/m. As a result of imbalances in the distribution of actuation around the spar contour, some bending deflection also was measured. However, the bending deflection at the tip was an order of magnitude smaller than the deflections due to twist.

The intended voltage cycle was 800 VDC + 4000 Vpp for the actuation testing. However, dielectric breakdown occurred at about 2600 Vpp between 2 packs in the spar section. The cause was most likely inadequate insulation between the conductive electrode rails at the ends of the packs. The glass/epoxy plies which separate the active plies may not have provided sufficient insulation between the edges of +45° packs and the -45° packs. Ensuring that the edges of the electrode rails are encapsulated within each pack so that they are not directly exposed will be the first attempted solution to the problem.

For this experiment, the modified Rehfield model predicts roughly 1.69°/m assuming a free strain of 460 microstrain for the active plies in the spar. The estimated pack free strain reflects the effects of reduced voltage and increased frequency. Additional structural tests on the spar will provide information on the efficacy of the model predictions and will quantify the variability in the manufacturing process.

VII. CONCLUSIONS

A 1/6th Mach scale CH-47D model blade has been designed with active twist capability. The design and manufacture of the active model blade has been demonstrated in a benchtop spar section test. A twist rate of 1.26 deg/m was measured in a preliminary test. This was roughly 75% of the predicted value for that level of actuation. Further actuation testing was limited as a result of dielectric breakdown damage in the spar resulting from insufficient insulation between packs. Future tests on the blade section will be used to evaluate the stiffness and strength properties of the spar. The next prototype will incorporate improved insulation and higher performance requirements for the packs in order to achieve the design objectives for the integral blade.

VIII. ACKNOWLEDGMENTS

This work was supported by DARPA under the Smart Structures for Rotor Control contract with Dr. Spencer Wu of AFOSR and Dr. Robert Crowe of DARPA as the technical contract monitors. Special thanks to Aaron Bent, Alessandro Pizzochero, Seward Pulitzer, Paul Bauer, Eric Prechtl, Richard Meyer, and Jaco du Plessis for their assistance with this project at MIT. The project was also supported by Robert Derham, Richard Bussom, Joe Orso, and Ray Gurnee at Boeing Helicopters.

IX. REFERENCES

1. Chen, Peter C. and I. Chopra, “Wind Tunnel Testing of a Smart Rotor with Induced-Strain Actuation of Blade Twist”, AIAA Paper No. 96-1273, AIAA/ASME/AHS Adaptive Structures Forum, pp. 76-90, Salt Lake City, UT, 1996.

2. Prechtl, E. F. and S. R. Hall, “Design of a High Efficiency Discrete Servo-Flap Actuator for Helicopter Rotor Control”, to be presented at SPIE’s 1997 Symposium on Smart Structures and Materials, San Diego, CA, 1997.

3. Straub, Friedrich K. and Ahmed A. Hassan, “Aeromechanic Considerations in the Design of a Rotor with Smart Material Actuated Trailing Edge Flaps”, Proceedings of the AHS 52nd Annual Forum, June 1996.

4. Giurgiutiu, V., Chaudhry, Z. and C. A. Rogers, “Engineering Feasibility of Induced Strain Actuation for Rotor Blade Active Vibration Control”, J. Intelligent Material Systems and Structures, 6(5), pp. 583-597, 1995.

5. du Plessis, A. J. and N. W. Hagood, “Performance Investigation of Twist Actuated Single Cell Composite Beams for Helicopter Blade Control,” 6^th International Conference on Adaptive Structures Technology, Key West, FL, 1995.

6. Rodgers, John P., Bent, Aaron A., and Nesbitt W. Hagood, “Characterization of Interdigitated Electrode Piezoelectric Fiber Composites Under High Electrical and Mechanical Loading”, SPIE Paper No. 2717-60, Proceeding of SPIE’s 1996 Symposium on Smart Structures and Materials, Sand Diego, CA, 1996.

7. Derham, Robert C. and Nesbitt W. Hagood, “Rotor Design Using Smart Materials to Actively Twist Blades”, Proceedings of the American Helicopter Society 52nd Annual Forum, Washington, DC, 1996.

8. Bent, A. A. and N. W. Hagood, “Improved Performance in Piezoelectric Fiber Composites using Interdigitated Electrodes,” SPIE Paper No. 2441-50, Proceedings of the 1995 North American Conference on Smart Structures and Materials, San Diego, CA, 1995.

9. Bent, Aaron A., “Improved Performance in Piezoelectric Fiber Composites using Interdigitated Electrodes,” Active Materials and Structures Laboratory Report #97-1, MIT, 1997.

10. Zhang, Q. M., Zhao, J., Uchino, K., and J. Zheng, “Change of the Weak Field Properties of Pb(ZrTi)O3 Piezoceramics with Compressive Uniaxial Stresses and Its Links to the Effect of Dopants on the Stability of the Polarizations in the Materials”, Submitted to J. Material Research, April 1996.

11. Rehfield, L. W., “Design Analysis Methodology for Composite Rotor Blades”, Proceedings of the 7^th DOD/NASA Conference on Fibrous Composites in Structural Design, AFWAL-TR-85-3094, pp.V(a) - 1-15, June 1985.

12. Weems, Douglas B., model predictions provided using Boeing proprietary anaysis.

13. Hexcel, “F155: A Controlled Flow Epoxy Resin for Lamination and Co-Curing, 250°F Cure”, Hexcel, 1994.

14. 3M, “Introductory Data: SP381 Epoxy Prepregs”, 3M, 1994.

15. Shultz, L. A., Panda, B., Tarzanin, F. J., Derham, R. C., Oh, B. K., Dadone, L., “Interdisciplinary Analysis For Advanced Rotors- Approach, Capabilities, and Status”, Presented at the AHS Aeromechanics Specialists Conference, San Francisco, CA, January 1994.

16. Meyer, Richard, Unpublished data on relative dielectric properties of PZT fibers, Materials Research Lab, Penn State University, 1997.

X. APPENDIX: INDEX OF MANUFACTURERS

Company	Product
3M St. Paul, MN	SP381-S29-UNI-33RC SP381-IM7-UNI-33RC AF163-2U-0.03
Hexcel Pleasanton, CA	E120/F155 Fabric
Cera Nova Hopedale, MA	Piezoceramic Microrods
Staveley Sensors Hartford, CT	Piezoceramic Fibers
Southwall Technologies Palo Alto, CA	Copper/Kapton
Eastprint, Inc. No. Andover, MA	Screen Printing
Epoxy Technologies Billerica, MA	Epotek 410E epoxy
Shell Houston, TX	Epon 9405/9470 resin
BYK Chemie GmbH Wesel, Germany	A530 air release agent
Richmond Aircraft Norwalk, CA	Rohacell 31 IG, 71 IG
All Flex Northfield, MN	Flexible circuits
CHR Furon Division New Haven, CT	K-250 tape
Keyence Woodcliff, NJ	LB11/70 laser sensors
Yorkville Pickering, Ontario	AudioPro 3400
TECO Winnisquam, NH	Transformer
Kepco Flushing, NY	DC Supply

Table VII. Manufacturers and Products.