Preparation and Mechanical Properties of UV?Assisted Filament Winding Glass Fiber Reinforced Polymer?Matrix Composite

，，，，

1.College of Materials Science and Technology，Nanjing University of Aeronautics and Astronautics，Nanjing 211106，P.R.China；

2.National Key Laboratory of Science and Technology on Helicopter Transmission，Nanjing University of Aeronautics and Astronautics，Nanjing 210016，P.R.China

（Received 17 October 2019；revised 22 November 2019；accepted 10 April 2020）

Abstract: This paper studied the preparation and mechanical properties of glass fiber reinforced polymer-matrix composite rings prepared by filament winding assisted by ultraviolet（UV）curing. A ray-tracing method was used to calculate the penetration ability of UV light in the resin casting，and then a typical composite ring with dual?curing characteristics was prepared by UV-assisted curing. The effects of winding speed and thermal initiator concentration on the distribution of fiber fraction and mechanical properties were studied. Microscopic morphology was used for the observation of the differences in fiber volume fraction. Mechanical properties tests and scanning electron micrographs were performed to investigate the failure and damage mechanisms of the composite ring samples. The ray tracing results indicate that the UV light can pass through a single yarn thickness and the energy transmitted is sufficient to cure the back side quickly. The experimental results show that the mechanical properties of the composite ring prepared in this paper are comparable to those of the heat-cured samples，which is sufficient to meet the requirements of the flywheel.

Key words：glass fiber reinforced polymer（GFRP）；ultraviolet（UV）curing；dual-curable resin；mechanical

0 Introduction

High-performance continuous fiber-reinforced resin matrix composites are widely used in various fields such as aircraft［1-3］， wind turbine rotor blades［4-6］and flywheel rotors［7-9］due to their advan?tages of high strength to weight and high stiffness to weight ratios. Among them，the composite flywheel rotor is generally produced by a filament winding process［10-12］，which is of low cost and high efficien?cy. The application of rapid curing technics could ef?fectively improve the energy storage performance of flywheel，such as ultraviolet（UV），microwave，and electron beam curing methods［13-15］.

The flywheel rotor made of high-strength con?tinuous fiber reinforced composite material can achieve a storage density of 17.6 times that of mar?tensitic steels［16］，as claimed by Conteh et al. The energy storage density is proportional to the mass and to the square of its rotational speed［17］. There?fore，the increase in rotational speed can achieve a significant increase in the energy storage density.Refs.［18-20］indicated that when the flywheel rotor rotated at high speed，the radial delamination may occur prior to the fiber breakage in the circumferen?tial direction. The introduction of fiber pre-stress can alleviate radial delamination damage，while rap?id curing resin can effectively maintain the stress on the fiber.

According to the reports of Gutowski et al.［21-25］，the fiber stress in filament winding is composed of three parts：the initial fiber tension caused by wind?ing tension，the change of fiber tension caused by fi?ber movement，and stress relaxation caused by ther?mal expansion or contraction and chemical changes.Applying of rapid curing resin as a matrix，the vis?cosity of the resin can be increased rapidly during the winding process，so that the radial displacement of the fiber can be reduced and the initial fiber ten?sion can be retained. The rapid curing resin system comprises thermal curing and radiation curing，wherein the time of thermal curing is as fast as a cou?ple of minutes，while the radiation curing may take only a few seconds. Among radiation curing meth?ods，the UV cationic curing resin has characteristics of rapid initiation，rapid growth（which will increase the viscosity in short time and maintain the winding tension），and difficulty in termination that provides an advantage for post-cure treatment to improve in?terlayer performance. Refs.［26-27］proposed a lay?er-wise curing method to overcome the poor pene?tration performance of UV curing，but the effect is not significant. The maximum tensile strength and shear strength of composite prepared only by UV exposure in Ref.［27］are 902 MPa and 25 MPa，re?spectively. Tena et al. reported a series research on UV-assisted out of die pultruded composites，the shear strength could reach 50 MPa and the flexural strength was about 833 MPa［28-30］.Endruweit et al.［31-33］addressed that UV light can transmit through voids between the fiber meshes and through the fiber meshes，and dual-curing resin formulations would post cure the resin shadowed by fibers. Compared with thermal curing，UV curing has characteristics of low energy consumption，fast curing speed and environmental friendliness. However，the poor pen?etrability of UV light restricts its application in com?posite preparation.

In this paper，the UV penetrability in resin castings was quantitatively analyzed by ray tracing method，a dual-curable resin matrix with two-stage curing characteristics was proposed. A-stage was ex?cited by UV to achieve a rapid and high degree of crosslinking before the prepreg entering the winding mandrel. B-stage was initiated by heating to ensure the bonding strength between the layers. The objec?tive of this paper is to obtain the penetration efficien?cy of UV light in the sample，and prepare the UVassisted curing glass fiber reinforced composite ma?terial.

1 Modeling

In order to examine the UV exposure dose needed in the pretreatment process quantitatively，a model containing UV intensity distribution on the specimen surface and internal as a function of posi?tion was established. Firstly，the UV light source was divided into finite elements. Then a random sampling rule for light-emitting points and orienta?tions was set up. Finally every beam of light was al?located with a certain amount of intensity by

whereE0is the initial intensity of UV lamp andmthe sampling number of light source.

UV intensity distributions would be obtained by tracking the path of each light，and the loss of light passing through the medium was calculated without considering the fluctuations and polarization of light. The parabolic cylinder reflector was also di?vided into finite elements which were small enough to approach the real surface. Only one time reflec?tion was in consideration during the curing process and assuming that there was no intensity loss when reflected.

1.1 Coordinate of UV lamp

The UV lamp system used in this paper was consisted of a cylinder modulator tube and a parabol?ic cylinder reflector. Two coordinate systems were used in the simulation. A Cartesian coordinate sys?tem was used to describe the position of UV lamp（Fig.1）. Fig.1 contains a Cartesian coordinate sys?tem（X，Y，Z）and a spherical coordinate system（r，γ，ρ），wherePis a random point on the lamp，ρthe azimuth angle，γthe scattering angle ofP，andIthe normal of the plane tangent to the tube sur?face and passing throughP. The origin of coordinate is the center of the UV lamp，theYaxis is the same line as the central line of the cylinder UV lamp，and theZaxis is perpendicular to the specimen surface.The cylindrical lamp has a radius ofrand a length ofh. A moving spherical coordinate system，whose or?igin is determined by the position of emitting pointP(x1，y1，z1)，was used to trace the light trajectory.In this spherical coordinate system，the scattering angle isγ∈[0，π/2]，and the azimuth angle isρ∈[0，2π]. The direction of the emitting light at this point can be expressed asn=(n1，n2，n3)and de?termined by the included angle ofand the nor?mal of the tangent plane through the intersection point ofand cylindrical surfaceγ∈[0，π/2].

Fig.1 Coordinate system of UV lamp

1.2 Sampling of UV lights

The position and direction of UV lights were described based on a pseudo-random numberξi，which was uniformly distributed over the range of［0，1］. Combined with geometric constraint equa?tions and random sampling rules，the trajectory of each light can be determined.

The sampling of emitting pointPcan be de?rived from

whereris the radius of lamp，hthe length of the lamp，andξ1，ξ2are random numbers. (x1，y1，z1) is the coordinate ofP. The sampling of emitting direc?tionn= (n1，n2，n3)can be derived from

whereξ3andξ4are random numbers. A random pointF(x，y，z) on the parabolic cylinder reflector can be determined by

wherecis a constant（-a≤c≤a），pthe focal length of the parabolic cylinder，andathe half length of the parabolic cylinder. The sampling ofF(x0，y0，z0) on the parabolic cylinder reflector can be derived from

where (x0，y0，z0) is the coordinate ofF，lthe open?ing width of the parabolic reflector，andξ5，ξ6are random numbers. The vector of incident lightn0=(u0，v0，w0)can be described as

The normal of tangent plane through the inter?section point of incident light and cylindrical reflec?tor surface can be described asN. According to dif?ferential geometry［34］，the vector of reflected lightn= (n11，n22，n33)can be described as

The position of the reflected lights on specimen surface can be described asP1(x2，y2，z2)，that is

2 Experiment

2.1 Materials and equipment

The Diphenol A epoxy（E51）used as mono?mer was obtained from Nan Tong Xing Chen Syn?thetic Material Co.，Ltd，and mainly contributed to the mechanical properties of resin. The Triarylsulfo?nium hexafluoroantimonate（Chivacure1176） used as photoinitiator was provided by Chitec Technolo?gy Co.，Ltd. The boron trifluoride-benzylamine complex（BF3-BZA）used as thermal initiator was purchased from Shanghai Macklin Biochemical Co.，Ltd. This thermal initiator used for B-stage curing exists as solid，white powder at room temperature and sensitive to moisture. High-strength glass fiber（HS6）was supplied by Nanjing Fiberglass R&D In?stitute with a linear density of 800 g/km，virgin fi?ber tensile strength of 4 800 MPa，modulus of elas?ticity of 94 GPa and elongation to break of 5.7%.Acetone was used as solvent to dissolve BF3-BZA and regulate the viscosity of the resin to better im?pregnate the fibers，obtained from Shanghai Alad?din Biochemical Technology Co.，Ltd. The radia?tion intensity received by the prepreg was measured using an intensity meter（Sentry Optronics Co.，Ltd of Taiwan）which possessed a spectral region of 280—400 nm with a measurement range of 0—40 mW/cm2and a minimum resolution of 0.01 mW/cm2.

2.2 Preparation of reactive resin

In order to better dissolve the photoinitiator，the epoxy resin was heated to 60 ℃in a magnetical?ly stirred oil bath to reduce its viscosity while avoid?ing water absorption and then added 5% （in weight） photoinitiator dropwise. After that，the mixture was stirred for 30 min and cooled to room temperature. The thermal initiator dissolved in the acetone solution was then added to the mixture，fol?lowed by magnetic stirring for 15 min to uniformly mix. The resin matrix for studying the influence of thermal initiator was prepared in the following ra?tios：E51∶Chivacure 1176∶Acetone was 100∶5∶8，and the thermal initiator content was used in five dif?ferent ratios， 0%， 0.5%， 1.5%， 2.5%， and 3.5%，respectively，while the resin matrix for studying the influence of winding speed was pre?pared in the following ratios：E51∶Chivacure 1176∶BF3-BZA∶Acetone was 100∶5∶1.5∶8.

2.3 Fabrication of composite ring （A?stage curing）

The twistless glass fibers from the torque mo?tor was dipped in the resin tank，and the resin con?tent in the prepreg was controlled by adjusting the temperature of the solution and the wrap angle of the fiber on the squeezing rollers.

The preparation of naval ordnance laboratory（NOL）ring referred to American standard ASTM D2291，and the winding tension was maintained at 20 N. The dipped fibers first passed through the UV irradiation zone，and then through a tension detector before entering the winding mold. The two sets of rollers（Fig.2）before and after the UV irradiation area ensured that each fiber passed through the same area. And then，the fiber tension was adjusted to be stabilized at 20 N±2 N by the voltage value of the torque motor according to the feedback information on the tension detector. Finally，the prepreg yarn was guided by the guide mouth to prepare the NOL ring.

Fig.2 NOL ring fabrication scheme

2.4 Post?curing of composite ring （B?stage curing）

The prepreg was also seriously subjected to the thermal effects when exposed to the UV mercury lamp. In order to prevent the photoinitiator and the thermal initiator from being simultaneously trig?gered，BF3-BZA was chosen as the thermosetting agent. From the DSC curves of the dynamic heating process in Fig.2，it can be determined that the UV reactive resin matrix mixed with BF3-BZA has an initial curing temperature of 115 ℃and final curing temperature of 180 ℃. Therefore，after the wind?ing，the NOL ring can be post-treated with a final temperature（180 ℃）and holding time（30 min）to ensure complete curing.

3 Characterizations

3.1 UV absorption coefficient measurement

In order to eliminate the measurement error caused by the height of UV-meter itself，a UV lamp system was mounted on a mechanical arm to mea?sure the UV light intensity after a relative displace?ment in theZdirection. During the curing process，the absorption coefficient of the resin with photoiniti?ator would increase due to the shrink of the volume as presented by Lam et al［35］. Thus，the UV intensi?ty meter would fix on the under glass plate to record the UV intensity at intervals（Fig.3）.

Fig.3 Configuration for measurement of the transmitting ra?diation and curing mould

3.2 Test of fiber volume fraction

The distribution of fiber volume fraction in typi?cal samples varies significantly with the winding speed and the curing agent content. It can be seen from the metallographic picture in Fig.4 that there is a significant interfacial resin-rich zone between the layers，and the fiber accumulation in the layer is rel?atively tight.

In this article，the volume content of fiber and volume content of resin-rich area were calculated by means of the metallographic photographs using Im?ageJ software. From the metallographic photo?graphs of the cross-section in Fig.4，it can be seen that under the 20 N winding tension，fibers are closely packed with rarely voids，and each fiber is uniformly surrounded with resin. Therefore，it is possible to obtain a more accurate fiber volume frac?tion using microscopic techniques. In order to re?duce errors caused by fiber incompleteness in the metallographic picture，ImageJ software was used in this paper to obtain the average diameter of the fi?bers，and then the fiber volume content was calcu?lated by

whereis the volume content of fiber，nthe fiber numbers of interested region，d0the average diame?ter of fibers，andSthe total area of interested re?gion.

3.3 Mechanical properties of composite ring

The tensile strength and shear strength test were performed on the MTS test machine with ref?erence to the American standard ASTM D 2290 and ASTM D 2344， respectively. The tensile strength can be determined with the help of Eq.（10），while the shear strength was obtained ac?cording to Eq.（11），as shown in Fig.5.

whereσis the apparent yield or ultimate tensile stress of the specimen（MPa），Pmthe maximum or breaking load（N），bthe average width of the speci?men（mm），anddthe average thickness of the spec?imen（mm）.

whereτis interlaminar shear strength of the speci? men，MPa.

Fig.5 Test of mechanical properties

4 Results and Discussion

4.1 Results of UV absorption coefficient mea?surement

The UV light would pass through three differ?ent media：air，glass plate and resin before reaching the UV-meter. UV light distribution after transmit?ting through these media would maintain a slight dif?ference（less than 5% in the range of 10 cm cen?tered on thex-axis origin），if the surrounding envi?ronment remains stable. However，after transmit?ting through resin with photoinitiator the UV light distribution would change a lot as presented by Lam et al［35］.According to the Beer-Lambert law

whereE0is the initial intensity，μthe absorption co?efficient andHthe distance. After passing through the medium，the UV light intensity received by UVmeter can be obtained by

whereμiandHiare the absorption coefficients and thicknesses of the air，the upper glass plate，the res?in and the lower glass plate，respectively. The mea?sured values for different media with a relative dis?placement and corresponding absorptions forμi，de?termined according to Eq.（13），are listed in Table 1. Distribution of absorption coefficient is shown in Fig.6.

Table 1 UV light intensity with relative displacements and the absorption coefficient μi as derived from Eq.(13)

Fig.6 Schematic diagram of distribution of absorption coefficient

In addition，the UV ht distribution after pass?ing through resin with photoinitiator was measured as shown in Fig.7.

Fig.7 Light intensity and light absorption value change over time for resin with photoinitiator

With the increasing of exposure time，the re?ceived UV light intensity decreased to nearly a plat?form value after a quasi-linear drop in the first 20 s as shown in Line 1.As mentioned previously，the re?action resin consists of monomer and photoinitiator.The photoinitiator has a strong selective absorption of UV light，while the monomer is almost transpar?ent to UV light. The transmittance of the resin with the UV exposure process seems to increase gradual?ly with the gradual depletion of the photoinitiator.However the resulting reactive groups will rapidly initiate polymerization of the monomers with contin?uous consumption of the photoinitiator，which will lead to a rise in resin viscosity and a significant change in the absorption coefficient. The absorption of resin with photoinitiator was obtained as shown in Line 2 according to Eq.（13）. It was found that with?in a point the refractive index increased linearly with conversion to gel point［35］. Based on the optical theo?ry of light，if the refractive index of a medium is de?fined as a complex number in Eq.（14），it can be seen from Eq.（15）that the imaginary part of the re?fractive index of the medium reflects the attenuation due to the absorption of light［36］. Combined with Fresnel’s law，the linear variation of the intensity distribution of the first 20 s was obtained by linear fitting in Eq.（16）.

wherenandkare both real numbers.

whereλ0is the wavelength in vacuum.

wheretis the curing time andμthe coefficient of ab?sorptance.

4.2 Modeling of light distribution in resin

For the light transmission in resin with photo?initiator，the absorption coefficient can simply be taken as 10.425 cm-1for the initial absorption and 11.585 cm-1for the cured absorption according to Eq（.16）in the first 20 s. During the initial curing process，assuming that the polymerization occurred instantaneously［37］，the absorption coefficient would be updated with the curing process.

Since the location and orientation of UV light received on the sample was obtained，the distribu?tion of UV intensity on the upper surface can be quantified（Fig.8）.

Fig.8 Three-dimensional morphology picture of light inten?sity distribution in the resin thickness direction

It can be seen from Fig.8 that there was a quasiuniform area centered on zero point of position axis with a higher light intensity. The intensity distribu?tion curve of light intensity at different depths was shown in Fig. 9. The quasi-uniform area was of 20 cm long in theydirection and of 15 cm wide in thexdirection. It can be seen that the middle area of the light intensity distribution inx-axis plane（Fig.9（a））gradually stabilized with the increased UV pen?etration depth.

Fig.9 Light intensity distribution curve of resin in the thick?ness direction

When the thickness increased to 0.08 cm，the penetrated UV intensity was almost below the threshold value that was determined by Ma’s experi?ment（28.79 mw/cm2）［38］（Fig.9（b））. It was report?ed that when ultraviolet light penetrated through one layer of glass fiber fabric，the remaining energy of UV light remains more than 62%［31］. The penetra?tion efficiency of ultraviolet light in unidirectional glass fiber is larger than that of fabric. Therefore，the intensity of UV light after it penetrates 0.02 cm thick prepreg （47 mW/cm2×62%>28.79 mW/cm2）when the fiber is present can still reach the cur?ing threshold. Once the UV energy exceeds the cure threshold，the photosensitive resin polymerizes and crosslinks rapidly，as shown in Fig.7.

4.3 Effect on fiber volume fraction

The fiber volume fraction in the component is largely determined by the resin content of the pre?preg. In this paper，the resin content in the prepreg was controlled in the range of 30% ± 5%，the fi?ber volume fraction depends mainly on the percola?tion of the resin in the yarn. If the viscosity of the resin in the prepreg is low，the interlaminar resin will penetrate the surface of the part or accumulate in the gap of the yarn under the action of the fiber pressure，resulting in a high fiber volume fraction. If the viscosity of the prepreg is high，the resin flow ef?fect is not significant，resulting in a low fiber vol?ume fraction.

It can be seen from Fig.10 that the fiber vol?ume fraction first increases to 60.3% and then de?creases as the winding speed increases from 50 mm/s to 150 mm/s. The volume fraction of the resinrich zone first drops to 0.6% and then rises，the minimum fiber volume fraction and the largest vol?ume content of resin-rich zone all appear at a wind?ing speed of 50 mm/s，which are 51.9% and 8.2%，respectively. At the same time，with the increase of the percentage content of BF3-BZA，the fiber vol?ume fraction also increases first and then decreases，and the volume fraction of the resin-rich zone de?creases first and then increases. When the content of the thermal curing agent is 1.5%，the fiber volume fraction reaches the maximum percentage at 60.3%，on the other hand the volume fraction of resin-rich zone is the minimum percentage at 4.0%.

Fig.10 Influence of winding speed and percentage content of cure agent on fiber volume fraction

Prior to thermal curing，the viscosity of the res?in matrix increased due to the passage of UV lamp.The degree of crosslinking depended on the intensi?ty of ultraviolet radiation on the surface of the pre?preg，which was determined by the winding speed because the power of the UV lamp and the distance between the lamp and the fiber yarn were fixed. As?suming that the ultraviolet irradiation intensity did not change with time，the ultraviolet light intensity received per unit distance of the prepreg can be cal?culated from Eq.（17）.

wherePis the light intensity of the surface of the prepreg yarn（mW），L0the effective irradiation length of the ultraviolet mercury lamp（mm），vthe winding speed of the prepreg yarn（mm/s），bthe width of prepreg，anddthe unit length of prepreg.

4.4 Effect of winding speed on mechanical properties

The role of the matrix material in the compos?ite material is mainly to bond the dispersed fibers and transfer the load to fibers through the interface between fiber and resin during loading so as to achieve the purpose that the reinforcement fibers can evenly bear the external load. If the fiber volume content of the composite material is distributed in the thickness direction unevenly，especially in the large tension winding samples，there will be a grad?ual failure in the loading process.

As the winding speed decreased，the UV irradi?ation intensity increased correspondingly，and the tensile strength and interlaminar shear strength（ILSS）of composites shared a consistent trend，which increased at first and then decreased（Fig.11）. When the winding speed reached 75 mm/s，the highest average tensile strength was 1 582.32 MPa and the maximum shear strength was 35.37 MPa，which was 50.3% and 60.0% higher than the sample without UV light，respectively. However，when the winding speed changed from infinity（no UV exposure）to 150 mm/s，which means a slight increase of UV intensity，the tensile strength was in?creased by 30.7%，and the shear strength was in?creased by 22.3%. It can be seen that proper in?crease of illumination can significantly improve the mechanical properties of the composite. The reason may be that the heat curing dose of 1.5% （in weight）is too small，and in the absence of the Astage curing，the degree of crosslinking of the resin during the B-stage is insufficient，so the interlami?nar shear strength is low，and the interlayer resin cannot sufficiently transfer the load of the fiber dur?ing the tensile test，therefore a small amount of UV exposure can significantly improve the mechanical properties of the sample. When the speed was less than 75 mm/s，the interlaminar shear strength and tensile strength began to decrease，which may be caused by the high degree of crosslinking in A-stage curing. The viscosity of prepreg became high due to the crosslinking in A-stage curing，and resin-rich re?gions were then formed between the layers，which cannot be eliminated during the B-stage curing，re?sulting in the presence of resin-rich zone in the sam?ple as shown in Fig.4.

Fig.11 Mechanical properties of different winding speed

Fig.12 Force-displacement curve of NOL ring tensile test

From the force-displacement curve of the NOL ring tensile test（Fig.12），it is obvious that the sam?ple prepared by winding speed of 75 mm/s shows a slight load reduction before the failure，and the load change value is relatively small，while the remain?ing samples show more than a few different levels of load reduction before failure. There are two typical failure modes as shown in Fig.13，namely fiber cracking and delamination. A lower fiber volume fraction tends to exhibit a uniform fracture pattern，and a higher fiber volume fraction is prone to delami?nation and lateral tear fracture.

Fig.13 Different failure mode of tensile samples

It can be seen from the scanning electron mi?croscopy（SEM）fracture photograph（Fig.14）of the typical interlaminar shear specimen that the frac?ture surface has obvious cusps and fiber imprints.When the winding speed increases，the cusps steps become shallower following deeper cusps on the sur?face at winding speed of 75 mm/s（Fig.14（a））.Here the cusp tilt direction on the fiber dominated fracture surfaces referred to the opposite direction to the crack growth direction［39］. At winding speed of 150 mm/s，a ribbon was observed in Fig.14（b）.The cusps of the fracture surface of shear sample without UV exposure almost converges to a plane（Fig.14（c）），and the crack propagation path is sig?nificantly shortened，resulting in low shear strength.

Fig.14 SEM images of ILSS samples at different winding speed

4.5 Effect of thermal curing agent on mechani?cal properties

The tensile strength and interlaminar shear strength of the composite material also increase in the first and then decrease with the percentage con?tent of thermal initiator growing（Fig.15）. The in?terlaminar shear strength reaches a maximum value of 37.04 MPa when the percentage of thermal initia?tor is 2.5%，which is 81.9% higher than the sample without thermal curing，meanwhile when the per?centage of the thermal initiator is 1.5%，the NOL ring tensile properties of the composite reaches a maximum value of 1371.59 MPa，which is 23.8%higher than the sample of no thermal initiator.Therefore，the amount of thermal curing had a sig?nificant effect on the mechanical properties of the NOL ring of the composite material. The best per?formance of mechanical properties is achieved be?tween 1.5% and 2.5% of the thermal initiator.When the percentage of thermal initiator exceeds 1.5%，the crosslinking density of B-stage curing of the resin increases，which might lead to a decrease of resin toughness and an increase of resin brittle?ness. This may be the reason why the optimum ten?sile strength and the best shear strength do not ap?pear in the same curing agent concentration. At the same time， when the thermal initiator content changed from 0 to 0.5%，the tensile strength was in?creased by 6.4%，while the shear strength remained almost the same. Thus，the excessive or insufficient heat curing agent content cannot obtain good me?chanical properties.

Fig.15 Mechanical properties of different thermal initiator content

There is no obvious cusp can be observed in the scanning photomicrograph without thermal initia?tor in Fig.16（a）. It may be caused by the insuffi?cient cure in B-stage. As the thermal initiator per?centage content increased from 2.5% to 3.5%，the size of cusps become smaller while the number has increased. This may be due to the increased brittle?ness of the resin. The tensile strength and shear strength obtained by the optimum winding speed and the optimum heat curing agent content reached 70% and 71%，respectively，of the properties of the heat-cured sample.

Fig.16 SEM images of ILSS samples at different thermal initiator concentration

5 Conclusions

In the preparation of the composite flywheel ro?tor，the pre-stress level is difficult to accurately con?trol due to the radial displacement of the fiber. The UV-assisted rapid curing resin system with twostage curing characteristics can alleviate the fiber dis?placement along radial direction while ensuring the mechanical properties of the component. In this pa?per，an UV light initiated dual-curable resin matrix was used to study the mechanical properties of com?posite samples with constant winding tension. The effect of different winding speeds and different ther?mal initiator concentrations on performance was in?vestigated. Moreover， microscopic observations were used to understand the failure mechanisms.The following conclusions are obtained：

（1）A model of UV light was established and the distribution of light intensity on the surface of the specimen and inside the specimen was quanti?fied. When the thickness of the unidirectional pre?preg is less than 0.2 mm，the ultraviolet light trans?mitted through the prepreg is sufficient to cure the resin quickly.

（2）A resin-rich zone was observed in most of the samples herein，when the winding speed was the 75 mm/s，the thermal initiator content was ze?ro，and the volume fraction of the resin-rich zone was the largest，respectively，but as the winding speed increased and the thermal initiator increased，the volume fraction of the resin-rich zone eventually began to increase.

（3）When the winding speed changed from in?finity（no UV exposure）to 75 mm/s，the tensile strength was increased by 50.3%，and the shear strength was increased by 60.0%，meanwhile when the thermal initiator content changed from 0 to 1.5%， the tensile strength was increased by 23.8%，and when the thermal initiator content changed from 0 to 2.5%，the shear strength was in?creased by 81.9%.

（4）A small amount of light intensity can signif?icantly improve the mechanical properties of the component，while the effect of a small amount of thermal initiator was not significantly increased，but excessive UV or thermal initiators would lead to performance degradation.

（5）The composite material prepared by the UV-assisted dual-curable resin in this article had ex?cellent mechanical properties，and reliable perfor?mance data. It can be used as a resin matrix for large tension winding experiment，and further research should be conducted in the tension retention rate.