1 Introduction

Carbon fibre reinforced polymer (CFRP) composite materials are becoming increasingly popular in high-performance part manufacturing in favour of their high stiffness and strength to weight ratios [1, 2]. Tubular structures have good flexural and torsional rigidities with respect to weight, making them ideal for structural applications such as spaceframe car and aircraft chassis [3]. CFRP composite tubes, therefore, combine the advantages of both the material and the geometry, opening up the possibilities for more versatile designs and applications, especially for high-performance industries.

Tubular composite structures can be manufactured using both thermoset and thermoplastic matrices. Thermoset tubes may be produced with discrete bends using techniques like compression moulding and hand layup [4]. However, these techniques are often multi-step and require design-specific tooling, resulting in high complexity and manufacturing cost [5, 6]. In an industrial setup where mass production is required, high costs and complexity are difficult to address as thermosetting matrices are non-thermo-formable after solidification, which excludes the possibility of post-forming thermoset tubes into various designs from a standardised geometry [7]. Unlike thermoset tubes, thermoplastic tubes can be reshaped into new geometries upon reheating their matrices to softening temperatures [7]. By utilising the thermo-formability of the thermoplastic matrix, thermoplastic tubes may be produced efficiently using techniques such as laser based thermoplastic automated tape placement (TP-ATP) [8]. In TP-ATP, the geometries of thermoplastic tubes produced are generally limited to straight or gentle curves, while production of tight or continuous bends is restricted by the clearance between the mandrel and the placement head, tape tensioning during winding, and the complex coordinated motion requirements [9].

Rotary draw bending (RDB) at elevated temperatures is a cost-effective technique to enable rapid post-forming of straight thermoplastic tubes into desired curvatures to produce customised parts for industry [8]. Using this technique, straight thermoplastic tubes made from processes such as TP-ATP are first heated to their matrix melting temperature to enable fibre movements within the softened matrix [10,11,12]. Using RDB, a straight tube with a softened matrix can be bent into various bending angles at a fixed bending ratio (i.e. the ratio between the bending radius and tube diameter) without changing the dies to cut the instrument cost for producing parts with varying curvatures [13]. RDB can also be performed either under isothermal conditions using oven heating or with localised heating of tube bending zones only [14].

Previous studies found that geometrical distortions occurred due to insufficient internal support when tubes were formed under isothermal conditions [8]. During forming under isothermal conditions, fibre movements are possible throughout the tube instead of being restricted within the tube bending zone. When bending stresses are applied to the straight tube, tube intrados (inner arc) experiences compression and tube extrados (outer arc) experience tension. An internal mandrel is needed to constrain the deformations so that the tube can be formed to the desired curvature following the designed geometry changes [8]. Currently, soft silicone mandrels have been commonly used in the RDB forming of thermoplastic tubes for their flexibility to be bent with the tubes [8, 9]. In a previous study, where forming was conducted using a soft silicone mandrel, additional extrados strains and reduced bending zone diameters were observed due to the displacement of tube extrados into the soft mandrel [8].

Conventional RDB of metallic tubing is typically performed at room temperature, where plastic deformation of the tubes occurs under yield [15]. In contrast, RDB of thermoplastic tubing is performed at elevated temperatures where the tubes are formed at relaxed state with softened matrices [8,9,10]. In this case, an internal mandrel is especially important to provide sufficient support to the thermoplastic tubes with relatively poor cross-sectional stability and little self-support to prevent defects and failures.

For conventional RDB of metallic tubing, solid bullet mandrels and ball mandrels made from metals are commonly used. When bending a metal tube, the bullet mandrel is fixed in place, and internal support is provided by the curvature of the bullet corresponding to the designed extrados curvature of the tube to be bent [16]. The bullet mandrel design is commonly used due to its simple design that is cheap and easy to manufacture [16]. However, as tube intrados is not supported by the bullet mandrel, precise control of bending parameters which may include additional bending steps is often needed to minimise wrinkling on tube intrados [16].

A ball mandrel is made up of a base attached to a number of movable small balls corresponding to the inner diameter of the tubes [17]. During the RDB process, the balls move along the curvature of the bent tube to provide internal support and maintain its ovality [18,19,20]. Unlike the bullet mandrel, the ball mandrel design provides support to both tube extrados and intrados for more consistent bends [17,18,19]. However, the ball mandrel is unable to provide uninterrupted and continuous support due to the gaps between the balls, leading to the risk of wrinkling [21].

For applications in microsystems technology such as medical engineering, sensors, and optoelectronics, micro-tubes can be bent to tight radii using a micro-wire mandrel [22]. The micro-wire mandrel design consists of a bundle of tightly-packed thin micro-wires with a small clearance to the inner wall of the tube to provide continuous internal support to the entire tube during bending [22]. While winkling is prevented by the continuous support, the use of a miro-wire mandrel may lead to tube geometrical distortions due to the flattening of the wire bundle during bending [22].

By utilising features of the conventional mandrel designs used in the discussed RDB applications, four designs in the forms of silicone rod, bullet, wire bundle, and spring are tested in this study. An ideal design is decided upon comparing forming results of tubes using the four mandrel designs. As the most commonly used design [8,9,10,11], the silicone mandrel will be used as control. Four sets of carbon fibre reinforced polyamide 6 (CF/PA6) thermoplastic tubes will be post-formed under isothermal conditions using the four mandrels respectively. Tube axial and circumferential profiles will be characterised via optical measurements and micro-computed tomography (µCT). Qualitative assessments of the characterisation results will be performed to identify the mandrel design that enables tube forming with minimal distortions. In order to model tube mechanical properties for tubular component design in industrial applications, it is important to understand tube laminate thickness and void contents. Selected tubes will be analysed for tube wall thickness and void content and distribution after forming. The mandrels will also be discussed for their usability and potential when applied to other heating processes in industrial applications.

2 Materials and Testing Methods

2.1 Experimental Materials

This study used [± 60°]4 CF/PA6 tubes made a laser-assisted automeated tape placement (ATP) system (AFPT GmbH) [23, 24]. Straight tubes were manufactured using pre-impregnated (prepreg) CF/PA 6 unidirectional (UD) tapes (Celstran ® CFR-TP PA6 CF60-01, Celanese Corporation, Irving, Texas, USA). Each tape was 0.13 mm thick, 12 mm wide, and had a fibre volume fraction of 48%. The tapes were wound around a 20 mm diameter chrome steel mandrel at 12 m/min with a consolidation force of 200 N at a process temperature of 280 °C. The winding process adopted a conventional helical filament winding pattern with three circuits (P = 3) per ± 60° layer pair and a shift parameter of one (Ns = 1) following [25]. Tube fibre angles were defined where 0° is parallel to the longitudinal direction and 90° is parallel to the hoop direction.

The manufactured straight tubes had 8 plies and were measured with an inner diameter of 20 mm and an outer diameter of 22.6 mm, resulting in an average wall thickness of 1.3 mm. The length of tube bending zone is approximately 72.26 mm, corresponding to a 90° bend with a bending ratio of 2.

2.2 Experimental Setup

A bespoke rotary draw bender was used for the forming experiments, as shown in Fig. 1. The bender contains the same key components as a conventional RDB tool for metal tubing. The pressure and wiper dies were made from polytetrafluoroethylene (PTFE) to reduce heat loss of tube sections before forming. The bending angle selector on the main bending die enabled control of tube curvatures between 0 and 180° in 15° intervals using the same bending ratio.

Fig. 1
figure 1

CF/PA6 tube post-forming operation: (a) loading and heating, (b) bending and cooling

Full size image

The mandrels shown in Fig. 2 were used to provide internal support to the inner ply of the tube during forming. The silicone mandrel was cast using Barnes M4470/T37 high temperature condensation cured RTV-2 in an acrylic tubing with an inner diameter of 20 mm. Its average diameter was measured to be 19.9 mm. The bullet mandrel was machined from mild steel with an average measured diameter of 19.5 mm. The curvature of the bullet corresponded to the designed curvature of tube extrados to constrain extrados strains. The bullet mandrel supported tube extrados only, where the most significant deformation was found in [8]. The wire mandrel was made by brazing together the base of a bundle of tightly packed copper wires stripped from a single core, double insulated (SDI) flex cable. The average measured diameter of the wire mandrel was 19.6 mm. The spring mandrel used a Clipsal 266MD25 medium duty steel conduit bending spring with an average measured diameter of 19.5 mm.

In the forming setup, the silicone mandrel and the spring mandrel covered the full length of the CF/PA6 tubes and were allowed to move with the tubes as they bent. The bullet mandrel and the wire mandrel were fixed in place by a steel rod screwed into their bases to support the tube bending zone only. When fixed in place, the start of the tube bending zone was aligned to the corresponding positions of the mandrels, as shown in Fig. 2(b) and (c).

Fig. 2
figure 2

The four mandrels tested in this study: (a) silicone, (b) bullet, (c) wire, (d) spring

Full size image

2.3 Experimental Procedure

The tubes were formed using the same technique and parameters as in [8], which consisted of two main steps, as shown in Fig. 1. In the first step, a straight CF/PA6 tube was loaded into the bender along with the mandrel, (Fig. 1(a)). The assembly was heated in a Steridium O75 oven to the matrix softening temperature of 220 °C. The temperature of the tube was monitored by an Omega 5SC-TT-K-40-36 fine wire K type thermocouple inserted between the mandrel and the tube prior to heating. The temperature data was recorded by a National Instrument cDAQ-9188 data logger throughout the forming operation.

After the tube temperature reading stabilised at 220 °C, the assembly was removed from the oven. The main bending die was rotated at a uniform rate of approximately 45 °/s to the bending angle of 90° immediately after removal to produce a curved tube, as shown in Fig. 1(b). The pressure die slid forward during this process due to friction between the die and the moving tube to provide external support to the tube section before bending. When a tube was bent to 90°, it was further clamped in position until it cooled down to 100 °C to minimise the effect of spring back [9].

2.4 Forming Result Characterisation

Defects such as wrinkles and delaminations were identified via inspections in the bending zones of formed tubes using a Micro-Vu automated precision measurement system. Three of the tubes without obvious defects were then selected to be imaged using a HeliScan™ µCT instrument hosted at the Australian National University (ANU) CTLab to obtain accurate measurements of their geometrical changes. The instrument is equipped with a 60 kV X-ray micro-focus source and a 3040 × 3040 pixels flat panel detector. The tubes formed using the wire and spring mandrels were scanned for a total 17 h. The voxel size (scanning resolution) of their reconstructed 3D volumes was 47.1 μm. The tube formed using the silicone mandrel was scanned for a about 21 h at the same scanning resolution of 47.1 μm.

From the µCT 3D images, tube axial profile was characterised by the measurements on post-formed bending zone extrados length, l1,m, and tube circumferential profile was characterised by bending zone diameters, d1,m following [8]. The measured values were compared with the original diameter d0 and theoretical extrados length after forming, l1,e to check for discrepancies. l1,e = 90.01 mm (Fig. 3) was calculated using tube bending radius, rb, and tube external radius, r:

$${l_{1,e}} = {\rm{ }}\left( {{r_b} + {\rm{ }}r} \right)\pi /2$$
(1)
Fig. 3
figure 3

Parameters of post-formed tubes

Full size image

Lastly, wall thickness and void content and distribution analysis were performed in Dragonfly (Object Research Systems, Canada) on the µCT 3D images of selected tubes without visible defects and significant geometrical distortions. Tube wall thickness was measured using Dragonfly’s sphere fitting algorithm to identify tube bending zone areas where significant thickness change occurred. Tube void content was mesasured throughout the interior of the tube walls. Voids were identified and separated from the 3D images of the tubes via an image segmentation process based on thresholding of air and tube voxel greyscales. Consistent upper and lower greyscale values were used for thresholding to appropriately account for partial volume voxels.

3 Results and Discussion

3.1 Tube Surface Quality

No visible defects and defects were observed on the surfaces of tubes formed using the silicone, wire, and spring mandrels, as shown in the examples of Fig. 4. Amongst the tubes formed by these three mandrels, tubes formed using the silicone mandrel and the spring mandrel had the smoothest surface finish, based on optical inspections. While the tubes formed by the wire mandrel showed minor unevenness along tube extrados, no evident defect was found. Smooth surface finish of the bends produced using these three mandrels indicated relatively limited fibre movements during forming.

For tubes formed using the bullet mandrel, their extrados was smooth with no visible defects, reflecting sufficient support by the mandrel. Major failures were observed on the intrados in the form of wrinkles and fibre buckling, as observed in the Micro-Vu images. This was due to the lack of internal support to the tube intrados by the bullet mandrel. For the [± 60°]4 CF/PA6 tubes used, fibre movements were dominated by local transverse tension and compression as opposed to shearing dominated changes [8]. During forming, compressive stress applied at the tube intrados caused an increase in wall thickness. As tube intrados was only supported externally by the main bending die, it compressed into the tube cavity where no constraint was present. Without internal support, the compressive stress was not spread out evenly across the tube intrados to form a smooth surface. As a result, the local stress build-ups formed inward folds and wrinkles, as shown in Fig. 4(b). On the other hand, although tube extrados was not supported externally, it experienced tensile stress that led to inward displacement towards the tube neutral axis during forming. The internal support of tube extrados by the bullet mandrel corresponded to the direction of stress application, allowing the mandrel to constrain the inward displacement of tube extrados. As a result, tube extrados experienced smooth transitions and no defect was observed.

From the visual evaluation of tube surface quality, the bullet mandrel was concluded as unviable for causing tube intrados failures during forming. The other mandrel designs were then further evaluated for tube geometries.

Fig. 4
figure 4

Photos of tubes formed using (a) silicone, (b) bullet (not imaged by µCT), (c) wire, and (d) spring mandrel, respectively

Full size image

3.2 Tube Geometrical Changes

Analysis of the µCT images revealed discrepancies in all tubes when comparing their axial (Fig. 5) and circumferential measurements (Fig. 6) with the theoretical values. Tubes formed using the wire mandrel experienced the most significant distortions, followed by tubes formed using the silicone mandrel, while tubes formed using the spring mandrel experienced the least significant distortions. Firstly, tube axial profiles were evaluated in terms of tube extrados extension following [8]. During forming under isothermal conditions, matrix of the entire tube was softened. Fibre movements were thus possible throughout the tube rather than only within the bending zone. As tube intrados was fully supported and constrained by both the bender die and the mandrel, inward displacement was likely to occur at tube extrados due to insufficient mandrel support in response to the applied tensile stress during forming [8]. The displaced tube extrados would, therefore, extend beyond the designed tube bending zone.

From Table 1, the extended extrados length, Δl, was calculated from the difference between l1,m and 11,e using Eq. (2):

$$\Delta l = {\rm{ }}\left( {{l_{1,m}}-{\rm{ }}{l_{1,e}}} \right)/{l_{1,e}}$$
(2)
Table 1 Tube axial profile characterisation
Full size table

From the results of Eq. (2), tubes formed by the wire mandrel experienced the most significant extrados extension, which was 79.46% greater than the control. Tubes formed using the spring mandrel, on the other hand, experienced the least significant extrados extension, 14.64% smaller than the control.

Fig. 5
figure 5

µCT mid-plane cross-section images of tubes formed using (a) silicone, (b) wire, (c) spring mandrel

Full size image

For the tube circumferential profile, d1,m was measured at the bend centre corresponding to the angular position of 45° of a post-formed tube following [11] and compared to the original tube outer diameter, do of 22.6 mm, as summarised in Table 2.

Table 2 Tube circumferential profile characterisation
Full size table
$$\Delta d{\rm{ }} = {\rm{ }}\left( {{d_{1,m}} - {\rm{ }}{d_o}} \right)/{\rm{ }}{d_o}$$
(3)

Similar to extrados extension, tubes formed using the wire mandrel experienced the most significant diameter change, while tubes formed using the spring mandrel experienced the smallest change. When the bending load was applied during forming, tube extrados experienced tension and displacement towards tube neutral axis. As a result, tube extrados pressed onto the wire mandrel, forcing the wires to pack more tightly towards tube intrados. During forming, the outer surface of tube extrados was constrained by the main bending die while that of tube intrados was unconstrained (Fig. 1(b)). Tubes formed using the wire mandrel, therefore, experienced significant inward displacement of tube extrados, resulting in flattened extrados cross-section as shown in Fig. 6 (b).

Fig. 6
figure 6

µCT bend centre circumferential cross-section images of tubes formed using (a) silicone, (b) wire, (c) spring mandrel

Full size image

3.3 Tube Wall Thickness and Void Distribution

As tubes formed using the silicone, wire, and spring mandrels showed varying degrees of geometrical distortion without defects, one specimen of each series was analysed for wall thickness and void distribution.

Average bending zone wall thicknesses were taken from tube intrados and extrados, where the most significant changes were expected, as summarised in Table 3. From analysis of the µCT images, the tube formed using the silicone mandrel experienced the greatest average bending zone wall thickness change, followed by the tube formed using the spring mandrel, and the tube formed using the wire mandrel.

Voids were mainly distributed along tube extrados in all the tubes analysed, as shown in Fig. 7 (a) – (c). During forming, tube intrados was supported by the main bending die on the outside. Meanwhile, the bending stress forced the mandrel to bend inwards toward the bend centre, pressing onto the inner surface of the tube intrados. The main bending die and the mandrel, hence, applied pressure to the tube intrados, resulting in less voids.

Fig. 7
figure 7

3D visualisations of µCT tomogram images showing void distribution (in blue) of tubes formed using (a) silicone, (b) wire, (c) spring mandrel; wall thickness of tubes formed using (d) silicone, (e) wire, (f) spring mandrel

Full size image

All tubes were found to have a relatively low void content in terms of void volume fraction (i.e. < 1.5%). The tube formed with the spring mandrel was measured to have the highest void content (0.72 ± 0.43%), followed by the tube formed using the wire mandrel (0.28 ± 0.19%), while the tube formed using the silicone mandrel had the lowest void content (0.19 ± 0.13%). The higher void content of the tube formed using the spring mandrel was likely due to the gaps between the coils when the spring was bent with the tube during forming. These gaps allowed the molten matrix to flow in without sufficient pressure applied causing the tube to have a wavier bending zone inner surface, as shown in Fig. 8, which may lead to a higher void content particularly along its bending zone extrados (Fig. 7 (c)) in addition to the unconstrained outer surface of the tube.

Table 3 Tube bending zone wall thickness
Full size table
Fig. 8
figure 8

3D visualisations of subsets of the µCT tomogram images showing extrados surface of tubes formed using (a) silicone, (b) wire, and (c) spring mandrel

Full size image

The silicone and the wire mandrels, on the other hand, provided continuous internal support along the longitudinal axis of the tubes during forming. Hence, less voids were found on the extrados of the tubes formed by these two mandrels (Fig. 7 (a) and (b)). An area of relatively high void concrentration was found at the end of the tube formed using the wire mandrel, as shown in Fig. 7 (b). This area was within the hanging length of the tube, where it was only partially supported by the end of the wire mandrel. When the wire mandrel bent with the tube during forming, the end of the mandrel deformed from flat to wedge-shape due to the strain between tube extrados and intrados. As the opposite end of the individual thin wires were brazed together, wires closer to tube intrados had smaller bending lengths and greater hanging lengths, and wires closer to tube extrados had the opposite. As a result, the wedge-shaped end of the wire mandrel provided less internal support to the hanging extrados during forming, allowing for voids to form along the hanging length of the tube.

3.4 Mandrel Usability

The silicone mandrel was found to snap easily when being removed from the formed tubes. This resulted in broken silicone mandrel sections being left inside the formed tubes, hence making the mandrel non-reusable. In an industrial setup, where rapid post-forming of tubes is required, the single-use silicone mandrel may increase tooling costs as well as cause difficulties in mandrel removal. The broken silicone sections also prevent applications such as piping which require clearance within the tube cavity.

The wire mandrel and the spring mandrel made from metals were more durable and could be reused without breaking throughout the forming experiments. The removal processes of these two mandrels were different. While the spring mandrel could be removed relatively easily from a bent tube by twisting the spring anticlockwise without affecting the RDB setup, the wire mandrel had to be pulled out from the tube. It was more difficult to remove the wire mandrel by pulling as (1) the bent thin wires could be entangled, and (2) it required disassembly of the RDB setup.

4 Conclusion

In this study, four (silicone, bullet, wire, spring) mandrel designs for post-forming of thermoplastic tubes via RDB at elevated temperatures were investigated and compared to determine which design caused the least geometrical distortion and generated fewer defects during forming. In the forming experiments, four sets of [± 60°]4 CF/PA6 tubes were formed using the four mandrels respectively. The tubes were first heated to 220 °C, then formed to the bending angle of 90° with a bending ratio of 2 via RDB under isothermal conditions.

Through the forming experiments, the silicone mandrel, the spring mandrel, and the wire mandrel were found to be capable of providing continuous internal support to both tube intrados and extrados. During bending, adequate support to tube intrados prevented uncontrolled deformation in the form of inward compression. Similarly, adequate support to tube extrados minimised the inward displacement of tube extrados due to tension during bending. For the bullet mandrel, while it provided sufficient support to tube extrados, the tube intrados was left unsupported. As a result, tubes formed using the bullet mandrel were found to experience failures in the forms of wrinkling and fibre buckling at their intrados.

Tube geometrical distortions were found when forming a tube using the silicone mandrel in a previous study [8]. Inspection of µCT images showed that similar distortions were also found when using the wire mandrel and the spring mandrel. Significant geometrical distortions were found in tubes formed using the wire mandrel due to flattening of the wire bundle when bent, which led to flattened tube extrados.

When removing the mandrels from the formed tubes, the silicone mandrel tended to get trapped as it was relatively soft. Consequentially, pulling of the silicone mandrel out from the formed tubes may result in defects of the mandrel. On the other hand, the wire mandrel required disassembly of the bending setup and its thin wires were prone to entanglement, making removal difficult.

Analysis of tube wall thickness revealed that the tube formed using the silicone mandrel experienced the greatest average bending zone wall thickness change, followed by the tube formed using the spring mandrel, and the tube formed using the wire mandrel. Void distribution analysis showed that the average void content was low (< 1.5%) for all tubes analysed. Within the tube formed using the spring mandrel, voids were found to accumulate in areas with insufficient support, suggesting the importance of support for consistency and quality in post-forming of thermoplastic tubes.

While small distortions, including 14.61% extension in bending zone extrados and 4.51% reduction in tube diameter, were found in tubes formed using the spring mandrel, it outperformed the other three mandrel designs for:

  • forming tubes with smooth surface finish without defects;

  • forming tubes with the least significant geometrical distortions;

  • its reusability; and.

  • ease of removal from formed tubes by twisting the spring anticlockwise.

Future work will include design optimisation of the spring mandrel to reduce the effects of gaps between coils as well as the clearance between mandrel and tube to further minimise voids and geometrical distortions. Post-forming of thermoplastic tubes with other types of matrices, laminate configurations, and bending ratios may also be conducted using the optimised mandrel designs for design validation and to determine their effects on laminate quality. Continuous forming of tighter and / or multiple bends using the spring mandrel may also be investigated to determine its usability limits.