Accepted Manuscript

Design and production of filament-wound composite square tubes

Lei Zu, Hui Xu, Bing Zhang, Debao Li, Bin Zi, Bingzhan Zhang

PII: S0263-8223(18)30469-0
DOI: https://doi.org/10.1016/j.compstruct.2018.02.069
Reference: COST 9419

To appear in: Composite Structures

Received Date: 31 January 2018

Revised Date: 13 February 2018
Accepted Date: 22 February 2018

Please cite this article as: Zu, L., Xu, H., Zhang, B., Li, D., Zi, B., Zhang, B., Design and production of filament-
wound composite square tubes, Composite Structures (2018), doi: https://doi.org/10.1016/j.compstruct.2018.02.069

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Design and production of filament-wound composite square tubes

Lei Zu a, Hui Xu a, Bing Zhang a,*, Debao Li a, Bin Zi a, Bingzhan Zhang b

a
School of Mechanical Engineering, Hefei University of Technology, Hefei 230009, China
b
School of Automotive and Transportation Engineering, Hefei University of Technology,
Hefei 230009, China

Abstract:

Composite square tubes have gained increasing attention as energy absorbers due to their

high specific energy absorption capacity and long stroke. One of the key important issues for

producing filament-wound composite square tubes demands both windability and uniform

coverage of winding patterns. Based on the analytic geometry, the spatial relation between the

feed eye and the mandrel was outlined and the kinematic equations for coupling the motion of

the mandrel and the feed eye were derived. Consequently, a design method for small-angle

winding of composite square tubes was proposed, taking the non-slippage condition of

winding trajectories into account. A periodically geodesic winding theory was presented and

its winding error for various initial winding points was analyzed. The designed fiber patterns

were then applied to the practical production of a composite square tube with small winding

angles. The results show that the present design method for filament-wound square tubes is

accurate and reliable. The obtained kinematic equations and motion laws of the feed eye and

the mandrel satisfy the basic winding principle and manufacturability of filament-wound

composite square tubes. The present method is able to provide a useful tool for design and

production of composite square tubes.

Keywords: Composite materials; Filament winding; Square tube; Small-angle geodesic;

Windability

*Corresponding author. Tel.: +86 551 62901137; Fax: +86 551 62901137
E-mail address: [email protected] (B. Zhang).
1. Introduction

The major challenge for manufacturing square cross-sectional tubes for aircraft frame

structures is to make them light-weight so as to reduce the operational cost and improve the

overall system requirements to make them an attractive option for aerospace industries.

Filament winding has been widely used in producing and manufacturing polymeric composite

structures [1-5]. Through the relative motion between the mandrel and the feed eye, the fiber

yarn winds to the mandrel in accordance with certain rules to create a composite component.

Filament-wound products have been widely used in many industrial fields because of their

high specific strength & stiffness, high production efficiency, excellent vibration and

corrosion resistance, and good fatigue behavior. The winding parameters of the

manufacturing process can be deliberately designed to maximize the fiber strength according

to the stress conditions of products.

Aerospace frames made of composite structures possess significant crashworthy

characteristics accompanied by weight and cost effectiveness. Composite square tubes have

become a key component of load-carrying structures and energy-absorbing members in

aerospace, civil and automobile applications due to their high specific energy absorption

capacity, long stroke, and high specific bending stiffness. Moreover, tubes with square

cross-sections possess better compatibility for connecting with other frame structures. Various

methods and theories have focused on energy absorption and crashworthy characteristics of

composite square tubes. Zhang et al. [6] investigated both experimentally and numerically the

applicability and effectiveness of adopting buckling initiator for axially loaded square tubes.

Oshkovra et al. [7] carried out a comprehensive finite element simulation to evaluate the

response and analyzing energy absorption capacity of natural silk/epoxy square tubes.

Mamalis et al. [8] studied the compressive properties and crushing response of square carbon

FRP tubes subjected to static axial compression and impact testing using the LS-DYNA3D
explicit finite element code. Kalhor [9] tested hybrid square tubes made from S2 glass/epoxy

composites and 304 stainless steel with different fiber orientation, stacking sequence, and

thickness under quasi-static loading. Palanivelu [10] presented an experimental investigation

on the progressive deformation behavior of uni-directional pultruded composite square and

circular cross sectional tubes subjected to an axial impact load. Shin et al. [11] investigated

energy absorption capability of axial crush and bending collapse of aluminum/GFRP hybrid

square tubes. Mamalis et al. [12] reported the crashworthy behavior of square frusta of

fibreglass composite material subjected to axial compression at various strain rates; they [13]

also evaluated the behaviour and crashworthiness characteristics of square composite tubes

subjected to static and dynamic axial compression exerted by a hydraulic press and a

drop-hammer, respectively.

However, previous research has a major defect in that manufacturing of composite square

tubes was based on unidirectional pultrusion process, and the fiber alignments in the

transverse direction of the tube were thus considered negligible. Despite the fact that

pultruded square tubes subjected to uniaxial tension/compression do show great performance

and manufacturability, the absence of fiber stiffness/strength in the transverse direction of the

tubes proved to significantly confine their load-carrying capacity. A typical example of this

restriction is the limit for designing a composite square tube under transverse bending

moment or shear forces. Hence, filament winding process has emerged as an attractive

alternative to improve the transverse structural performance as well as to reduce the tube

weight. Many researchers have so far concentrated on design and manufacturing of filament

wound products. Zu et al. [14-15] developed non-geodesically overwound toroidal and domed

pressure vessels; they also determined the optimal shapes for filament-wound bellow-shapes

[16] and pressure vessels with unequal polar openings [17]. Vasiliev et al. [18] presented the

optimality conditions for a pressure vessel and derived the optimal dome profiles for various
anisotropy parameters. Liang et al. [19] outlined the optimal dome contours based on the

maximum shape factor and evaluated the effect of the dome depth on the structural

performance. Teng et al. [20] investigated the effect of the fiber band width on the stability of

the winding patterns and carried out the optimal design for an ellipsoid dome. Vafaeesefat et

al. [21] presented a multi-level strategy for the optimization of composite pressure vessels

with geodesic and ellipsoidal heads. Hojjati et al. [22] assessed the effect of mechanical

properties of composite materials on the dome profiles and proved that the matrix properties

have a major role in the dome design. Fukunaga et al. [23] obtained optimal meridian shapes

using several failure criteria and outlined an analytic approach for the design of dome

structures. Braun et al. [24] analyzed the process parameters of carbon fiber overwound

seamless-aluminum-lined composite pressure vessels. Seereerem et al. [25] investigated the

all-geodesic manufacture of a T-shaped form.

Nevertheless, very few investigations have been devoted to design of filament-wound

composite square cross-sectional shapes. Up to date a complete theory for designing and

obtaining the winding patterns and the feed eye movement of the square tube remains scarce.

Moreover, since the load in the transverse direction of the tube is generally much lower than

that in the longitudinal direction, the small-angle fiber orientations referred to the longitudinal

axis are required for achieving a rational and optimal stress distribution. The small-angle

winding of square tubes is more likely to cause fiber slippage on the tube mandrel as

compared to the winding of circular cross-sections due to non-rationally asymmetric

cross-sectional shape of the square tube that leads to uneven motion trajectories of the feed

eye and the mandrel [26]. Based on the elaboration of the relative motion of the feed eye and

the contact points on the mandrel, we formulate a non-slippage-based design theory and

successfully achieve the geodesic winding on the mandrel of the square tube. The kinematic

equations of the feed eye and the mandrel are then derived for the whole winding circuit. In
addition, a special yarn-hanging device is installed to ensure that the small-angle fiber

trajectories will return back at the end of the tube and smoothly transit to the next winding

circuit. Finally, a composite square tube is produced using the small-angle winding to verify

the feasibility and the accuracy of the present design method.

2. Kinematic equations of the feed eye with small-angle winding

Assuming that the mandrel does not move and the feed eye spirally moves around the

mandrel, we determine the kinematic equations of the feed eye using the theory of relative

motion.

2.1 Spatial relations of the feed eye and the contact point

Compared with the general cylindrical winding, filament-wound square tubes share a

complex relationship to the feed eye and the contact points; thus, we build a 3D analysis

model. The starting point is fixed on the vertex of the mandrel, and the initial yarn and

mandrel surfaces are on the same plane. During winding, the mandrel rotates and the feed eye

moves horizontally along the longitude of the mandrel. The trajectories of the yarn and the

feed eye at different times are illustrated in Fig. 1.

2.2 Approximate circular method

The so-called approximate circular method is here used for calculating the kinematic

equations of the feed eye, as shown in Fig. 2. The square cross-sectional tube with length a

can be approximately regarded as a circular tube of radius R, both of which have the same

circumference. The length of a chord connecting the both intersections of the approximate

circle and the square edge is a', making a central angle β. The following equations hold:

4a  2 R (1)

a '  2R sin  (2)

 Since the fiber path has an angle α with respect to the parallel circle of the torus, the

relation among geometric parameters dθ/dl, dφ/dl and α, can be given by (see Fig. 2): and the

approximate circle winding angle can be calculated as:

a'
 '  arctan( ) (3)
Z

 2 R
Z   (4)
180 tan 

2 sin   90 tan 
 '  arctan( ) (5)


The relationship of φ and φ' is depicted in Fig. 3.

2.3 Calculation of geometry to prevent yarn slippage

During the winding of square tubes, the winding angle calculated using the approximate

circle method deviates from the true winding angle. Slippage becomes highly likely when the

contact point transits from one mandrel plane to another. Geodesic winding is the most stable

in the winding process. Therefore, to prevent yarn slippage, the triangular envelope for the

projections of various fibers in the X-Y plane should comprise a series of similar triangles. As

shown in Fig. 4, the feed eye moves from position 1, and the triangular envelope for the

projections of the yarn in the X-Y plane is in the plane of the mandrel surface. When the feed

eye reaches position 2, the former and the latter are not in the same plane.

To avoid yarn slippage, the plane including the triangular envelope for the projections of

the yarn in the X-Y plane should be folded to the plane of the mandrel surface while keeping a

certain angle φ, as shown in Fig. 5.

In the X-Y plane shown in Fig. 6, the relatively kinematic angle of the feed eye increases by

90° from β0 to β0'. At this point, the position of the feed eye corresponds to position 1. The

following winding process is repeated until winding ends. We call this repetitive process a
cycle.

In the cycle, Lt is the length of mandrel, a is the side length of the cross section, A is the

distance of the feed eye to the center shaft of the mandrel, φ is the winding angle, and M(x0, y0)

is the vertex of the mandrel. The distance of the feed eye ahead is given by

Z  cos  x  x0 2   y  y0 2  z  z0 2  ctg  x  x0 2   y  y0 2
(6)
 ctg  A2  x02  y02  2 Ax0 cos   y0 sin  

3. Determination of the virtual initial point

We previously calculated the kinematic equation for the starting point of the mandrel

vertex. However, the mandrel vertex is not always the starting point during winding. The

initial turn-around angle can be expressed as β0=arccos(a/2A), β0'=β0+90°. The relation

between the turn-around angle and Y-coordinate of the feed eye is shown in Fig. 7. Y is the

moving distance of the feed eye in winding machine coordinates, and X is the radian for the

turn-around angle.

If the starting points are not from M, the feed eye moves at the first plane of the mandrel

surface and then enters into the circulation. Fig. 7 shows that the turn-around angle is not in

proportional to the motion of the feed eye along the Y-coordinate; this indicates that the

kinematic equations of the feed eye and the mandrel are varied for different initial winding

point. Therefore, formulating a motion control theory for the feed eye for an arbitrary initial

point is the key to the success of the winding.

When winding begins at an initial point different from the mandrel vertex, the oppositely

elongated line of geodesic trajectory and the line of the mandrel edge intersect at the virtual

starting point. As shown in Fig. 8, the geodesic equations for the contact point starting from

point M are equal to the equations after point M, which starts from point M'. We then achieve

geodesic winding from any starting point by calculating the virtual starting point for different
actual starting points.

We take a cycle as a unit and explore the starting point in a cycle. If the distance between

the point M and the mandrel edge is ax , then the distance between the virtual starting point

and the mandrel is L1 , which is expressed as L1=axctgφ. Similarly, when the yarns are wound

back, the contact points start from the virtual starting point on the oppositely elongated line of

the mandrel edge. The distance between the virtual starting point and the mandrel is L2 ,

which is expressed as L2=NL0-Lt-L1. N is a positive integer that meets the requirements of

0<L2<L0. L0 is the distance of the feed eye motion in a circuit, which can be obtained using Eq.

(6).

Considering the initial turn-around angle θi in the filament-wound pattern design, we

should calculate the turn-around angles of the mandrel according to length ax. Solving the

triangle OMM' in Fig. 9.

tan 
ax  a (7)
1  tan 


ii  N '  (8)
2

N' is a positive integer irrelevant to N and meets the requirement of 0≤β<π/2, and i is the

number of cycles for the feed eye to make a round trip and has a maximum value. From the

requirements of uniform coverage, we can derive the equation ibcosϕ=4a, where b is the

width of the yarn.

tan 
L1  ax ctg  a ctg (9)
1  tan 

Therefore, the mandrel turns a certain radian π/2 when the initial turn-around angle θi is

certain, iθi is increased, and N' is changed. Then, we calculate the virtual starting point on the

basis of the calculated value of β and realize the geodesic winding from any actual starting

point. The flow-process diagram is shown in Fig. 10.

4. Design of winding patterns

The winding patterns reflect the alignment law of the continuous fiber trajectories on the

mandrel surface. We set the mandrel turn-around angle with the round trip of the feed eye as

the symbol of the winding patterns. Continuous fibers start at a certain point on the mandrel

and wind for several cycles before finally returning to the starting point. We call this winding

process a complete cycle. A complete cycle has to satisfy the following conditions:

(1) The location of two sequential tangent points is adjacent. The polar location has one

tangent point before the appearance of the tangent point adjacent to the location of the starting

point. We call this winding pattern the single tangent point.

(2) The location of two sequential tangent points is segregative. The polar location has two or

more tangent points before the appearance of a tangent point adjacent to the location of the

starting point. This is called the multi-tangent point.

Each bundle of yarn corresponds to a tangent point on the circumference of the polar

location. Yarns are uniformly distributed piece by piece when every tangent point equally

divides the rotated angle of the mandrel and the staggered distance for the yarns in the

cylinder corresponding to the adjacent tangent points is equal to the width of the yarn. The

rotation angle of a round trip of the feed eye can be expressed as:

K 
n  (  N '')  360  (10)
n n

where n is the tangent point, N'' represents a non-negative integer, K is the integer inside the

scope of 1 to n-1, K/n is a simple proper fraction, and Δθ is the turn-around degree for the

yarn sheet distance in the cylinder corresponding to the adjacent tangent points. According to

Eq. (10), the values of n, K, N, and θn corresponding to various winding patterns are shown in

Table 1.
 Table 1. Winding patterns for S0, n, K, N, and θn.

n K
0 1 2 3 4 5 6 7 8
N
1 1 360° 720° 1080° 1440° 1800° 2160° 2520° 2880° 3240°

2 1 180° 540° 900° 1260° 1620° 1980° 2340° 2700° 3060°

1 120° 480° 840° 1200° 1560° 1920° 2280° 2640° 3000°

3
2 240° 600° 960° 1320° 1680° 2040° 2400° 2760° 3120°

1 90° 450° 810° 1170° 1530° 1890° 2250° 2610° 2970°

4 3 270° 630° 990° 1350° 1710° 2070° 2430° 2790° 3150°

1 72° 432° 792° 1152° 1512° 1872° 2232° 2592° 2952°

2 144° 504° 864° 1224° 1584° 1944° 2304° 2664° 3024°

5
3 216° 576° 936° 1296° 1656° 2016° 2376° 2736° 3096°

4 288° 648° 1008° 1368° 1728° 2088° 2448° 2808° 3168°

After determining the tangent points, we assign the mandrel an initial angle θi to meet the

requirements of uniform coverage in a complete cycle.

1 1 
n  2(i   L ) , i   n   L   n  N (11)
2 2 2
where θL is the one-way turn-around angle for the virtual length of the tube. When the

filament-wound patterns are certain, we select the proper value of θn and calculate the initial

turn-around angle θi. Finally, we obtain the virtual starting point and finish the winding.

5. Simulations of winding patterns on a square tube

The results of winding deviate from the truth because of winding error. Thus, simulations

of theoretical winding patterns are necessary to provide a theoretical guidance for the actual

winding and a reference for the final result of winding. With the aid of MATLAB that can

calculate mathematical equations and graphic simulations, we build an entity model of the
square tube in the dimensional coordinate. Then, according to the requirements of the winding

parameters, we calculate the linear equations on the model of the square tube. Finally, we

formulate the linear equations of the trajectory on the square tube. On this account, we

explore the theoretical winding patterns of the square tube in geodesic winding. The

simulation of the winding patterns on a mandrel of the square tube is shown in Fig. 11.

6. Production for filament-wound composite square tubes

The numerical control system of the winding device SINUMERIK 810D is here used to

produce a composite square cross-sectional shape. We control the rotation of the spindle, the

linear motion of the winding trolley, and the arm movement of the feed eye by writing

procedure codes. From the system settings of the winding machine, X stands for the rotated

radian of the spindle, Y stands for the linear kinematic distance of the winding trolley, and Z

stands for the arm-moving distance of the feed eye. The winding trolley moves a distance ΔY

when the spindle rotates a radian Δx. A deliberately designed and produced fiber-hanging

device was installed at the ends of the square mandrel, in order to ensure a smooth transition

to the subsequent wound circuit when the fibers return back at the ends of the mandrel.

At the beginning of winding, the mandrel rotates at a degree of θi and the yarn is fixed on

the rod of the mandrel. The whole winding process has five tangent points, and the yarns

entirely cover the mandrel after the feed eye completes 19 round trips. The winding condition

is shown in Fig. 12.

7. Conclusions

In this paper, we study the winding patterns of composite square tubes with small winding

angles. Considering the relatively kinematic principle, studying the movement of the feed eye,

and analyzing the geometrical relationship of the feed eye and the mandrel, we meet the
requirements of non-slippage winding and implement the feed eye movement of geodesic

winding. Using the segmented cycle method and turning the center angle by 90° as a cycle,

we obtain the uniform equation of the feed eye motion for the small-angle winding-composite

square tubes. In addition, we achieve the winding uniformity of the composite square tubes by

controlling the NC winding machine. The results show almost no slippage during winding. At

the end of the winding process, the yarn uniformly spreads over the whole mandrel. However,

this paper does not cover the thickness uniformity of the square sections because of the

incomplete theory regarding the measure and control of section thickness during winding.

This topic will be the focus of our follow-up work.

The geodesic calculation of composite tubes in this paper shows the correct practice of

theory. The results enrich the geodesic calculation theory of the traditional surface for

filament-wound production. It also provides a reference for the geodesic calculation theory of

the profiled products with filament winding.

In the future, the structural performance (stiffness & strength) of the pultruded square

tubes and filament-wound square tubes will be evaluated and compared to each other, in order

to demonstrate that the filament winding process leads to preferable performance of

composite square tubes subjected to transversely distributed loads.

Acknowledgements

This research is supported by the National Natural Science Foundation of China (Grant No.

11302168).

References

[1] Wilson B. Filament winding: the jump from aerospace to commercial frame. SAMPE

Journal, 1997, 33(3): 25.

[2] Wang R, Jiao W, Liu W, Yang F. Dome thickness prediction of composite pressure vessels

by a cubic spline function and finite element analysis. Polym Polym Compos 2011, 19:

227-34.

[3] Wang R, Jiao W, Yang F. A new method for predicting dome thickness of composite

pressure vessels. J Reinf Plast Compos 2010; 29: 3345-52.

[4] Zu L, Koussios S, Beukers A. Integral design for filament-wound composite pressure

vessels. Polym Polym Compos 2011; 19: 413-420.

[5] Zu L, Koussios S, Beukers A. Optimal cross sections for filament-wound toroidal

hydrogen storage vessels based on continuum lamination theory. Int J Hydrogen Energy

2010; 35: 10419-29.

[6] Zhang XW, Su H, Yu TX. Energy absorption of an axially crushed square tube with a

buckling initiator. International Journal of Impact Engineering 2009; 36: 402-417.

[7] Oshkovr SA, Taher ST, Oshkour AA, et al. Finite element modelling of axially crushed

silk/epoxy composite square tubes. Compos Struct 2013 ; 95: 411-418.

[8] Mamalis AG, Manolakos DE, Ioannidis MB, et al. Crashworthy characteristics of axially

statically compressed thin-walled square CFRP composite tubes: experimental. Compos

Struct 2004; 63: 347-360.

[9] Kalhor R, Case SW. The effect of FRP thickness on energy absorption of metal-FRP

square tubes subjected to axial compressive loading. Compos Struct 2015; 130: 44-50.

[10] Palanivelu S, Van Paepegem W, Degrieck J, et al. Experimental study on the axial

crushing behaviour of pultruded composite tubes. Polymer testing 2010; 29: 224-234.

[11] Shin KC, Lee JJ, Kim KH, et al. Axial crush and bending collapse of an aluminum/GFRP

hybrid square tube and its energy absorption capability. Compos Struct 2002; 57:

279-287.

[12] Mamalis A G, Manolakos D E, Demosthenous G A, et al. Energy absorption capability of

 fibreglass composite square frusta subjected to static and dynamic axial collapse.

Thin-Walled Structures, 1996, 25(4): 269-295.

[13] Mamalis A G, Manolakos D E, Demosthenous G A, et al. The static and dynamic axial

crumbling of thin-walled fibreglass composite square tubes. Composites Part B:

Engineering, 1997, 28(4): 439-451.

[14] Zu L, Koussios S, Beukers A. Design of filament-wound toroidal hydrogen storage

vessels based on non-geodesic fiber trajectories. Int J Hydrogen Energy 2010; 35:

660-70.

[15] Zu L, Koussios S, Beukers A. Design of filament-wound domes based on the continuum

theory and non-geodesic roving trajectories. Compos Part A: Appl Sci Manuf 2010; 41:

1312-20.

[16] Zu L, Koussios S, Beukers A. Shape optimization of filament wound articulated pressure

vessels based on non-geodesic trajectories. Compos Struct 2010; 92: 339-46.

[17] Zu L, Koussios S, Beukers A. Design of filament-wound isotensoid pressure vessels with

unequal polar openings. Compos Struct 2010; 92: 2307-13.

[18] Vasiliev VV, Krikanov AA. New generation of filament-wound composite pressure

vessels for commercial applications. Compos Struct 2003; 62(3): 449-59.

[19] Liang CC, Chen HW, Wang CH. Optimum design of dome contour for filament-wound

composite pressure vessels based on a shape factor. Compos Struct 2002; 58(4): 469-82.

[20] Teng TL, Yu CM, Wu YY. Optimal Design of Filament-Wound Composite Pressure

Vessels. Mech Compos Mater 2005; 41: 333-40.

[21] Vafaeesefat A, Khani A. Head Shape and Winding Angle Optimization of Composite

Pressure Vessels Based on a Multi-level Strategy. Appl Compos Mater 2007, 14 (5-6):

379-91.

[22] Hojjati M, Safavi AV, Hoa SV. Design of domes for polymeric composite pressure
 vessels. Compos Eng 1995; 5(1): 51-59.

[23] Fukunaga H, Uemura M. Optimum design of helically wound composite pressure vessels.

Compos Struct 1983; 1(1): 31-49.

[24] Braun CA, Haddock RC. Manufacturing process controls for high reliability carbon

filament wound seamless aluminum lined composite pressure vessels. AIAA, SAE,

ASME, and ASEE, Joint Propulsion Conference and Exhibit, 28th, Nashville, TN, July

6-8, 1992. 9

[25] Seereerem S, Wen JTY. An all-Geodesic algorithm for filament winding of a T-shaped

form. IEEE Transactions on Industrial Electronics 1991; 38: 484-490.

[26] Mazumdar SK, Hoa SV. Analytical models for low cost manufacturing of composite

components by filament winding, part I: direct kinematics. Journal of Composite

Materials 1995, 29(11): 1515-1541.

 List of Figures

Fig.1. Spatial model of the feed eye and the mandrel.

Fig.2. Approximate circle in the X-Y plane.

Fig.3. Deviation of the approximate circular method.

Fig.4. Geometrical relations of various fiber positions and the mandrel.

Fig.5. Triangular envelope for the projections of various fibers in the X-Y plane.

Fig.6. Geometrical relations of the X-Y plane and the mandrel.

Fig.7. Y-coordinates of the feed eye with turn-around angles of the mandrel.

Fig.8. Virtual starting point corresponding to the actual starting point.

Fig.9. Relationship between the turn-around angle β and the length ax.

Fig. 10. Flow diagram of finding the virtual starting point.

Fig. 11. Simulation of winding patterns for a square cross-sectional tube: (a) a square tube; (b)

fiber trajectories at the junction of the adjacent mandrel edges; (c) Uniform coverage patterns

on the mandrel.

Fig. 12. Winding process of a square tube with small-angle geodesics: (a) feed eye and the

mandrel at the beginning of winding; (b) fiber trajectories for two winding circuits; (c)

uniform and full coverage.