Voids in Type IV Composite Pressure Vessels Fabricated Using a Dry Filament Winding Process

 

Type IV high-pressure hydrogen tanks for fuel cell electric vehicles are made of carbon fiber reinforced plastic. Their production involves winding carbon fiber reinforced thermosetting resin tows onto a plastic liner to form the tank shape, which is then heated in an oven to cure the thermosetting resin. The filament winding process produces many voids in the tank, which can lead to changes in quality and product life.

 

In this study, voids in the tanks were classified according to their sources. Type IV high-pressure hydrogen tanks with a burst pressure of 70 MPa were made and the voids produced in the tanks were observed. The same filament winding process was used to manufacture plates and tanks with simple winding trajectories to identify voids. The voids produced in the tanks by filament winding were divided into six groups based on their location, cross-sectional shape, and size, showing the sources of the voids.

 

Composite Pressure Vessel, Carbon Fiber, Filament Winding Voids, Hydrogen, Type IV

 

Hydrogen tanks for fuel cell electric vehicles need to have high-pressure storage capabilities to provide a sufficiently long driving range. Fuel cell electric vehicles use Type IV high-pressure hydrogen tanks, which are composed of carbon fiber reinforced plastic (CFRP) and a plastic liner, also known as composite pressure vessels. The tank stores more than 70 MPa of hydrogen and has a range of approximately 500 km. However, regulations require a high design safety factor, which increases weight and cost [3]. Further development of tank and structural health monitoring technologies is needed to achieve highly reliable, lightweight hydrogen tanks that reduce vehicle fuel consumption and are less expensive to manufacture.

 

Type IV high-pressure hydrogen tanks are typically manufactured using filament winding technology.

 

A tow prepreg consisting of carbon fiber tows impregnated with epoxy resin is prepared in advance and wound around a plastic liner to form the shape of the tank, which is known as a dry filament winding process. After the tow prepreg is wound onto the plastic liner, it is cured in an oven to reinforce the tank. These manufacturing processes can produce a large number of voids in the CFRP portion of the tank.

 

Voids typically reduce the mechanical properties of multi-directional laminated CFRP . Furthermore, voids often serve as crack initiation sites, thereby reducing fatigue strength. Therefore, voids are a measure of CFRP quality. Voids in filament-wound tanks have been studied previously. The volume fraction, location (e.g., interlaminar and intralaminar), and size distribution have been investigated using cross-sectional observations. X-ray computed tomography (CT) can reveal the three-dimensional shape and spatial localization of voids.

 

However, due to their complex winding trajectories, the fibers are arranged in multiple directions. Depending on the fiber orientation, voids develop in different locations and shapes. The filament winding process also causes errors in the winding trajectory. The complex laminated structure of filament-wound tanks makes it difficult to identify the source of voids.

 

In this study, voids generated in Type IV high-pressure hydrogen tanks manufactured using filament winding technology of carbon fiber reinforced thermosetting resin tow prepreg were observed and classified according to their sources. The tanks used a simple stacking sequence to identify the sources of voids. The voids were classified according to their location, cross-sectional shape, and size. The sources of voids in filament-wound tanks were identified and methods to reduce the number of voids were discussed. The experimental results can help in the design of tanks with reduced voids.

 

Materials and Methods

 

Materials and Winding Machine

 

Carbon fiber (T720SC-36K, Toray) and epoxy resin (XNR6880/XNH6880, Nagase Chemtex) were used (Table 1) [26]. Carbon fiber tows were impregnated with epoxy resin to form tow prepreg blanks. The resin weight fraction of the tow prepreg was about 25%.

 

 

The tow prepreg is wound on a bobbin and placed on a creel. The tow prepreg is then pulled from the creel and fed to the delivery head of a filament winder (PW30-236-5-1, Engineering Technology Company) (Figure 1(a)). During the filament winding process, tension is applied to the tow prepreg, which is controlled by the torque applied to the bobbin (Figure 1(b) and c). The elongation of the spring is measured using a linear encoder, which transmits the tension data to the torque control system of the creel stand.

 

Figure 1. Filament winding machine. (a) Winding section, (b) Creel, (c) Tension control system in the creel.

 

The tension was selected to be 20 N, which enabled the tow prepreg to be stably unrolled (Fig. 2 (a)). Under this condition, the width of the tow prepreg was about 9 mm (Fig. 2 (c)). Smaller tension resulted in intermittent release of the tow prepreg (Fig. 2 (b)).

 

A high tension of 30 N resulted in a relatively large change in the width of the tow prepreg. Fiber tension affects the quality and strength of the tank by affecting the compaction and void content of the tank [27, 28]. The dry fiber winding process is generally not equipped with a compaction roller [7]; therefore, a compaction roller was not used in this study.

 

Figure 2. Tension of the drag prepreg. (a) Stable winding, (b) Unstable winding, (c) Winding the fiber tow onto a plate when the tension is 20 N.

 

Fabrication Method

 

Plate

 

The plates and cans were manufactured by filament winding. The tow prepreg was wound onto a steel plate for plate manufacturing (Fig. 3(a)). [0] 4 and [0/90] 2s laminations were selected. For [0/90] 2s lamination, the steel plate was rotated orthogonally at each lamination of a single layer. The prepreg tow must pass through the entire plate mandrel for continuous winding.

 

Therefore, the actual angles of the [0] 4 and [0/90] 2s laminations are [0 + θ ] 4 and [0 + θ /90 + θ ] 2s, and θ is 0.4° in this study. The tow prepreg is wound so as not to create gaps or overlaps, although gaps or overlaps do occur to some extent due to the variability of the winding process.

 

Figure 3. Fabrication of flat plates. (a) Winding process, (b) curing process, (c) consolidated plate.

 

The filament-wound steel plate was placed between steel plates. The tow prepreg was consolidated in a constant pressure (0.13 MPa) oven at 150°C for 3 h using a uniaxial loading machine (Figure 3(b)). The average thickness of the fabricated [0] 4 and [0/90] 2s plates was approximately 1.2 and 2.4 mm, respectively. The surface of the plate was relatively flat due to the pressure applied through the steel plate (Figure 3(c)).

 

Gas Cylinder

 

To fabricate the tank, a nylon-based polymer liner was prepared and the tow prepreg was wound around it. The outer diameter and length of the liner were 300 and 785 mm, respectively. During the winding process of the tow prepreg, internal pressure was applied to the liner to prevent deformation. The tank was consolidated in an oven at 150°C for 3 h, during which internal pressure was applied to the liner to prevent the tank from shrinking.

 

Two types of cylinders were manufactured by changing the winding trajectory. The first tank was designed with a burst pressure of 70 MPa for void observation and burst testing. This tank is called the burst test tank. The total number of laminations was 22, the same as the second tank.

 

The second tank was designed for void observation. A ±45° winding path was used to obtain an orthogonal lamination similar to the plate sample ([0/90] s ). However, it failed during the manufacturing process, resulting in large dimensional variations. Therefore, after many rounds of repeated trials, [±45 4 /89 2 /±45 4 /release film/89 2 /16 2 ] lamination was selected, and the ±45° lamination area was observed.

 

Although some gaps were observed due to the variability of the winding process, the tow prepreg was wound with an overlap of 30% in the width direction. The hoop layer of the 89° lamination was required to prevent the expansion of the cylindrical part during curing.

 

After curing, a release film was inserted to remove the outer layers 89 2 and 16 2, which also served as hoop and helical layers to prevent expansion in radial and axial directions. This tank is called an observation tank. No shrink tape was used on the tank during manufacturing. The surface of the tank was relatively uneven compared to the plate.

 

Observation Method

Cross-sectional observations were performed using an optical microscope (VHX-8000, Keyence and Eclipse L150, Nikon) and a scanning electron microscope (SEM) (JSM-6510, JEOL). Internal observations were performed using an X-ray CT system (SkyScan 1272S and SkyScan 2211, Bruker). The CFRP plate and tank body were cut into small pieces to observe the cross-section and internal structure. The cross-section was polished using an abrasive compound before optical microscopy. The cross-section and observation direction are provided in each section.

 

Burst Test

The burst test tank was placed in an underground pit. Internal pressure was applied to the can until failure by injecting water using a high-pressure pump (AJP-30020G2-E, Sugino Machine). The internal pressure was monitored during the test and the burst test was completed within 3 minutes.

 

Results

 

Observation Of Tow Prepreg

 

The tow prepreg used for plate and slot winding was observed before the winding process. Figure 4 (a) shows the observed tow prepreg, and Figure 4 (b) shows the SEM image of its surface. Fiber corrugations were observed on the surface. Some fibers were severely misoriented. A cross section perpendicular to the fiber direction was obtained by X-ray CT, as shown in Figure 4 (c). Many voids with a size of about 50 μm were observed inside the tow prepreg. There were severely misoriented fibers and voids in the tow prepreg before winding.

 

Figure 4. Observation of tow prepreg. (a) Towpreg, (b) SEM image of the “b” area in (a), (c) X-ray CT image of the cross section.

 

Observation Of The Plate

 

The cross section of the CFRP plate is shown in Figure 5. The [0] 4 plate is observed in the fiber direction and transverse direction, marked as [0] 4 and [90] 4, respectively. The [0/90] 2s plate is observed in the fiber direction and oblique direction, marked as [0/90] 2s and [±45] 2s, respectively. Black, gray, and white represent voids, resin, and fiber, respectively.

There are many small voids of about 50 μm distributed in [0] 4; relatively large voids are marked with arrows in the figure, although many small voids are also distributed throughout the plate. Here, each void is identified using a magnified image.

 

[0/90] 2s shows interlayer voids. Interlayer voids may form when tow prepreg is wound at different angles.

[±45]2s shows larger voids than [0]4 because the voids are oriented in the fiber direction.

The voids in [±45]2s are larger in appearance due to being cut obliquely.

 

These voids are easier to identify when viewed in the fiber direction than when viewed in the transverse or oblique directions.

 

 

 

Figure 5 Cross-sectional observation of the plate. [0]4 and [90]4 represent the entire cross-sectional view, and [0/90]2s and [±45]2s represent the half cross-sectional view.

 

Due to the limited resolution (3.0 μm/pixel) of the X-ray CT system, the CFRP plate was cut into 7.5 mm × 7.5 mm sections. The X-ray CT image of the top surface of [0]4 is shown in Figure 6 (a). The internal voids were extracted from the X-ray CT image using image processing software (CTVox, Bruker), as shown in Figure 6 (b). The voids were extracted by setting a brightness value threshold that roughly matched the void ratio of the side surface.

The voids were accompanied by misoriented fibers.

Figure 6. X-ray CT images of the [0]4 plate. (a) Top surface image (gray indicates fibers and matrix, black indicates voids) and (b) void extraction 3D image (white indicates voids).

 

The X-ray CT image of the side surface of [0/90]2s is shown in Figure 7 (a). The internal voids are shown in Figure 7 (b). Voids were found in the tow prep, at the edges of the tow prep, between the tow prep due to gaps, and in the interlayer region. These voids have a finite length along the fiber direction.

 

Figure 7. X-ray CT images of [0/90]2S plates. (a) Side view image (gray indicates fibers and matrix, black indicates voids) and (b) 3D image of extracted voids (white indicates voids).

 

The void fraction of multiple cross-sectional X-ray CT images was measured using image processing software (Fiji) [29] and was found to be approximately 1.0% and 3.6% for [0]4 and [0/90]2S plates, respectively. Multiaxial lamination leads to an increase in void content due to an increase in the number of interlayer voids.

 

Explosion Test and Observation Of The Explosion Test Tank

The explosion test was conducted using an explosion test tank. The explosion pressure was 77.7 MPa, close to the design strength value (Figure 8). The cylindrical part of the cylinder failed.

 

Figure 8. Burst Test Results.

(a) Internal Pressure-Time Curve and

(b) Burst Test Can After Test

 

Another burst test can was used to observe the voids in the burst test can. The dome and cylinder parts of the burst test can were cut; its cross section is shown in Figure 9. The fiber volume fraction was measured using the combustion procedure (ISO 14127) to be approximately 65%. The fiber volume fraction inside the can was higher than that outside the can due to the negative thermal expansion coefficient of carbon fiber. The resin seeped out during the curing process because the carbon fiber shrinks at high temperature.

 

The porosity was approximately 5.9%, which was higher than that of the sheet. The voids inside the can were lower than those outside the can, which we attributed to the repeated compressive force caused by winding. A large number of voids were formed inside the can, and it was difficult to classify them because of the differences in the location, cross-sectional shape, and size of the voids due to different winding trajectories.

 

 

Figure 9. Observation of voids in the burst test tank. 3.4. Classification of voids

 

The burst test tank is wound in multiple directions, so it is difficult to determine the source of the voids. Therefore, the observation tank was made using a simple meandering trajectory. Figure 10 (a) shows the prepared observation tank, and Figure 10 (b) shows a small cutout tank to indicate the observation direction. The observed cross section is perpendicular to the fiber direction (Figure 10 (b)). Many voids were observed in the tank (Figure 10 (c)).

 

However, due to its simple meandering trajectory and appropriate observation direction, it can be classified into six types according to its location, cross-sectional shape and size. The enlarged image of each void is shown in Figure 11, and the corresponding schematic diagram is shown in Table 2. According to the cross-sectional observation of the 6 cut surfaces of the observation tank (Figure 10 (c) shows one of the cut surfaces), the following results were obtained.

 

Figure 10 Observation Tank (a) The Entire Tank, (b) The Cutaway Plane Showing The Observation Direction, (c) The Cross Section Observed From “c” in (b). R: Radial Direction, Z: Axial Direction

 

Figure 11. Magnified images of six types of voids. (a) Tow preg void, (b) Tow preg edge void, (c) Curl void, (d) Overlap void, (e) Tow preg void, and (f) Interlayer void

 

 

Figure 11(a) shows “tow voids” with a circular cross section and an average size of 46.1 μm (coefficient of variation (CV) = 63%, total number of measured voids (n) = 79). The voids in each layer are almost the same size as the voids in the tow prepreg shown in Figure 4(c). Even after water tank consolidation, the voids in the tow prepreg still exist. It is necessary to minimize the number of voids in the tow prepreg to reduce the number of small voids distributed throughout the tank.

 

Figure 11(b) shows “tow edge voids” with a triangular cross section and an average size of 527 μm (CV = 87%, n = 59). These voids are caused by the uneven edges of the tow prepreg when the tow prepregs with the same winding angle are adjacent to each other. To eliminate these voids, the tow prepreg needs to be pressed in the thickness direction with a compaction roller during winding [22, 30].

 

Figure 11 (c) shows “crimp voids” with an elongated triangular cross section and an average size of 512 μm (CV = 68%, n = 60). These voids are caused by the curling of the tow prepreg due to the different winding angles. These voids are considered to be crack initiation sites under load due to their sharp edges [15]. The tow prepreg needs to be compacted in the thickness direction with a compaction roller during winding to eliminate these voids. However, the compaction force increases the crimping angle, thereby creating local weak points and reducing the strength of the tank. It is recommended to use thinner tow prepreg to suppress curling.

 

Figure 11 (d) shows “overlap gaps” with an elongated triangular cross section and an average size of 1.77 mm (CV = 50%, n = 18). These voids are caused by the overlap of the tow prepreg. Similar to the curl voids, these voids are also considered to be crack initiation sites under load due to their sharp edges. Compacting the tow prepreg in the thickness direction with a compaction roller during the winding process is required to eliminate these voids.

 

Figure 11 (e) shows a rectangular cross-section of “tow gaps” with an average size of 3.02 mm (CV = 90%, n = 9). This is caused by deviation from the designed winding path due to slippage of the tow prepreg and misalignment of the filament winding machine. Changes in the tow prepreg tension during the winding process change the width of the tow prepreg, resulting in the formation of tow prepreg gaps. This type of void can be eliminated by limiting the placement of the tow prepreg so that the winding path does not move.

 

Figure 11 (f) shows a needle-shaped cross-section of “interlayer voids” with an average size of 3.41 mm (CV = 68%, n = 77). These voids form between layers when the tow prepreg is not in close contact in the thickness direction. The tension is not sufficient to generate compaction force to eliminate the voids. These voids are likely to be caused by thickness variations due to time-dependent tension changes and uneven surfaces due to severe misorientation of the tow impregnated fibers. To eliminate these voids, it is necessary to increase the tension to produce a higher compaction force. Compacting rollers are a more direct method that can increase the pressure on the fiber tow in the thickness direction and thus eliminate these voids.

 

The locations, cross-sectional shapes, and average sizes of the six voids generated in a Type IV high-pressure hydrogen tank are summarized in Table 2. The locations of the voids vary depending on the nature of their source. Each type of void has its own characteristic cross-sectional shape. The sizes of different types of voids range from a few microns to several millimeters. The average size varies for each type of void.

 

The compaction force during winding can reduce the number of voids [6,31,32]. It has been reported that recompacting can reduce the number of voids [33, 34]. It is recommended to use thin tow impregnated materials to suppress curling. The number of small voids during curing can be reduced by applying high pressure [9, 35]. The technology needs to be effectively applied to reduce the proportion of six characteristic voids in type IV high-pressure hydrogen tanks.

 

Voids are the main cause of CFRP strength variation.

In order to experimentally investigate these voids, a type IV high-pressure hydrogen tank was manufactured by a carbon fiber reinforced thermosetting resin tow winding method. During the fiber winding process, many voids were generated in the slots. The voids present in the tow prepreg remained in the slots after consolidation. The porosity and fiber volume fraction varied along the thickness direction. The origin of the voids was determined by manufacturing plates and tanks with simple meandering trajectories.

 

Based on their location, cross-sectional shape, and average size, the voids were classified into six categories: tow impregnation material voids, tow impregnation material edge voids, curling voids, overlapping voids, tow impregnation material voids, and interlayer voids.

 

The causes of voids in each category are different. The thin tow technology, compaction roller technology, high pressure during vulcanization, and other void reduction technologies need to be effectively applied to reduce the void ratio of type IV high-pressure hydrogen tanks. The effect of each void on the failure behavior also needs to be studied to reduce the effect of voids on the burst pressure.

 

Source of this article (Voids in type-IV composite pressure vessels manufactured by a dry filament-winding process)

 

Figure 11 (a) shows “tow voids” with a circular cross section and an average size of 46.1 μm (coefficient of variation (CV) = 63%, total number of voids measured (n) = 79). The voids in each layer are almost the same size as the voids in the tow prepreg shown in Figure 4 (c). Even after the tank is consolidated, the voids in the tow prepreg still exist. The number of voids in the tow prepreg needs to be minimized to reduce the number of small voids distributed throughout the tank.

 

Figure 11 (b) shows “tow edge voids” with a triangular cross section and an average size of 527 μm (CV. = 87%, n = 59). These voids are caused by the uneven edges of the tow prepreg when the tow prepregs with the same winding angle are adjacent to each other. To eliminate these voids, the tow preform needs to be pressed in the thickness direction with a compaction roller during winding [22, 30].

 

Figure 11 (c) shows a “curl void” with an elongated triangular cross section, with an average size of 512 μm (CV = 68%, n = 60). These voids are caused by the curling of the tow preform due to the different winding angles. These voids are considered to be crack initiation points under load due to their sharp edges [15]. The tow preform needs to be compacted in the thickness direction with a compaction roller during winding to eliminate these voids. However, the compaction force increases the crimping angle, thereby creating local weak points and reducing the strength of the tank. It is recommended to use a thinner tow preform to suppress curling.

 

Figure 11 (d) shows an “overlap gap” with an elongated triangular cross section, with an average size of 1.77 mm (CV = 50%, n = 18). The overlap of the tow prepreg leads to these voids. Similar to the crimp voids, these voids are also considered to be crack initiation sites under load due to their sharp edges. Compacting the tow prepreg in the thickness direction with a compaction roller during the winding process is required to eliminate these voids.

 

Figure 11 (e) shows a “tow gap” with a rectangular cross section, with an average size of 3.02 mm (CV = 90%, n = 9). This is caused by deviation from the designed winding path due to slippage of the tow prepreg and misalignment of the filament winding machine. Changes in the tow prepreg tension during the winding process change the width of the tow prepreg, resulting in the formation of tow prepreg gaps. This type of void can be eliminated by limiting the placement of the tow prepreg so that the winding path does not move.

 

Figure 11 (f) shows a “ply gap” with a needle-shaped cross section, with an average size of 3.41 mm (CV = 68%, n = 77). These voids are formed between layers when the tow preform is not in close contact in the thickness direction. The tension is insufficient to produce a compaction force to eliminate the voids. These voids are likely to be caused by thickness variations due to tension changes over time and uneven surfaces due to severe misorientation of the tow impregnated fibers. To eliminate these voids, it is necessary to increase the tension to produce a higher compaction force. Compacting rollers are a more direct method to increase the pressure of the fiber bundle in the thickness direction and thus eliminate these voids.

 

The locations, cross-sectional shapes, and average sizes of the six voids generated in a Type IV high-pressure hydrogen tank are summarized in Table 2. The locations of the voids vary depending on the nature of their source. Each type of void has its own characteristic cross-sectional shape. The sizes of different types of voids range from a few microns to several millimeters. The average size varies for each type of void.

 

The compaction force during winding can reduce the number of voids [6,31,32]. It has been reported that recompacting can reduce the number of voids [33, 34]. It is recommended to use thin tow impregnated materials to suppress curling. The number of small voids during the curing process can be reduced by applying high pressure [9, 35]. This technology needs to be effectively applied to reduce the proportion of six characteristic voids in type IV high-pressure hydrogen tanks.

 

Voids are the Main cCause of CFRP Strength Variation.

 

In order to experimentally investigate these voids, a type IV high-pressure hydrogen tank was manufactured by a carbon fiber reinforced thermosetting resin tow winding method. During the fiber winding process, many voids were generated in the slots. The voids present in the tow prepreg remained in the slots after consolidation. The porosity and fiber volume fraction varied along the thickness direction. The origin of the voids was determined by manufacturing plates and tanks with simple meandering trajectories.

 

Based on their location, cross-sectional shape and average size, the voids were divided into six categories: tow impregnation voids, tow impregnation edge voids, curling voids, overlap voids, tow impregnation voids and interlayer voids.

 

The causes of voids in each category are different. It is necessary to effectively apply thin tow technology, compaction roller technology, high pressure during vulcanization and other void reduction technologies to reduce the void content of Type IV high-pressure hydrogen tanks. It is also necessary to study the effect of each void on the failure behavior to reduce the effect of voids on burst pressure.