In-line reciprocating friction stir welding of plastics

A Scialpi¹, M Troughton², S Andrews², and L A C De Filippis¹

¹ Dipartimento di Ingegneria dell'Ambiente e per lo Sviluppo Sostenibile (DIASS), Politecnico di Bari, Taranto, Italy

² TWI Ltd, Cambridge, UK

Published in Joining Plastics/Fügen von Kunststoffen Magazine, Issue 1, May 2007.

In German

Abstract

Friction Stir Welding (FSW) can be applied to a multitude of materials of various thickness. However, the use of this technique on plastics has had very limited success due to the thermal and viscoelastic properties of thesematerials and, to date, no commercial applications have been reported. In this paper a preliminary investigation on a new variant of FSW, called Viblade welding, is presented. In Viblade ^TM welding, frictional heat is produced by the action of a blade which vibrates in a linear reciprocating motion parallel to the joint line. Welding trials were carried out on 9mm thick polypropylene sheet. Designof Experiments was used to evaluate the influence of the process parameters and blade geometry on the dimension of the heat affected zone and the mechanical performance of the joint. Results showed that conditions could be found wherethe Viblade welded joints achieved mechanical properties and weld speeds comparable with the conventional techniques of hot gas and extrusion welding.

1. Introduction

Friction Stir Welding (FSW) is a relatively new welding process developed in the early 1990s at TWI. ^[1] In conventional FSW (butt-joint) a rotating tool is fed into and along the joint between the two sheets to be welded. The tool rotation produces mechanical mixing of the materials on the advancing and retreating side of theweld. While FSW of metals has had a lot of success, its use on plastics has been limited due to the very different thermal and viscoelastic properties of these materials. In the published literature only Arici ^[2] has presented conventional FSW applied to plastics. By using a double pass Arici eliminated the root defect, obtaining satisfactory tensile and bending results, but with a poor surface finish and very low feed rate. Typically,FSW of plastics suffers from the following problems:

• Difficulty in retaining material in the joint line;
• Low speed of welding;
• Creating uneven mixing at the weld line;
• A crown with poor surface finish.

To solve some of these problems Nelson ^[3] developed and patented a shoe through which the rotating pin passes. In this solution, heat is not produced by frictional heating from the shoulder, but by the hot shoe that also constrains ejection of material.

A variant of FSW studied previously at TWI is the vertical reciprocating FSW, ^[4] which consists of a blade that reciprocates perpendicular to the joint line, the reciprocating action of the blade causing the plastic material to soften and reform. The main problem with this technique was that there was nomechanism for retaining molten material in the joint, which resulted in voids in the weld and an associated low weld strength.

Another variant of FSW that is covered under the TWI patent is the reciprocating motion of a tool in the direction of the weld. This technique, called Viblade ^TM welding, has the benefit that the blade remains fully within the joint at all times, making it easier to contain the melt in the weld. The welding process consists of a blade and a shoulder that run along thethermoplastic sheets with a downward force ( Figure 1); they generate frictional heat causing the sheet material at the interface to melt and form a weld behind the blade. Since thermoplastics have a very low thermal conductivity, the heat produced by the shoulder is notenough to melt the material near the root of the joint; almost all of the heat input on the faces of the butt-joint is generated by the blade.

Fig.1. A schematic illustration of Viblade welding

Since this is a friction welding process it is possible to distinguish two main actions of the pressure: one necessary for heat generation, due to the friction between the vibrating elements (blade and shoulder) and the plasticsheets, the other to generate the intimate contact between the parts. The vertical load of the shoulder produces the heat input to melt the material on the top of the joint but has no direct effect on the intimate contact of the parts.The horizontal load is characterized by three different components: one due to the imposed horizontal load ( Figure 1), one due to thermal expansion of heated plastic material, and the last due to the action of the blade, which as it moves forward, extrudes the plastic material round itself. These three components act directly toproduce frictional heating by the blade surface. The pressure that generates the intimate contact between the parts is produced only when the blade has passed and is due to the imposed load and the effect of thermal expansion. Themelted material flows around both sides of the blade into the cavity vacated by the blade as it moves forward.

2. Experimental procedure

In order to investigate the potential of Viblade welding and the behaviour of the joints produced, a series of experimental trials were carried out on 9mm thick extruded polypropylene (PP) sheet. Two 100x220mm plaques werebutt-welded along their length using a Bielomatik linear vibration machine operating at resonance in order to maximize the energy given to the joint. The Viblade tool consisted of a PTFE shoulder and a titanium blade. Titanium was usedbecause of its relatively low thermal conductivity. In order to improve the frictional heating between the blade and the sheet it was necessary to increase the coefficient of friction. For this reason a series of grooves were markedonto the blade (see Figure 2).

Fig.2. Schematic of Viblade blade

The joints were sectioned perpendicular to the weld for macrographic analysis and local heating using hot gas was used to highlight the heat affected zone (HAZ). The width of the HAZ was measured at a position 4.5mm from the bottomof the sheets. The strength of the joints was evaluated using a 3-point bend test, measuring the ram displacement at crack initiation. ^[5]

Design of Experiments (DoE) analysis was used to study the effect of the main process parameters. The study was divided in two main parts. In the first, a preliminary investigation was carried out to analyse the effect of the mainprocess parameters on the dimensions of HAZ and the strength of the joint. In the second, a more detailed investigation was carried out to produce a response surface for the joint strength against the main process parameters.

2.1 Stage 1 - Preliminary investigation

The influence of the process parameters and blade geometry on the dimension of the HAZ and on the mechanical resistance of the joint was investigated during this work. In particular, the influence of five parameters were studiedusing the DoE technique:

1. Blade thickness (BT);
2. Blade length (BL);
3. Feed rate (FR);
4. Vertical load (VL);
5. Horizontal pressure (HP).

For each of these five parameters two levels were chosen, one low (identified by '-1') and one high (identified by '+1'). A fractional factorial experimental plan was used; the full 2 ⁵ factorial experimental plan was reduced to a 2 ^5-1 plan to decrease the number of welding trials. This 2 ^5-1 design would be expected to provide excellent information concerning the main effects and two-factor interactions. Table 1 shows the levels chosen for each of the above parameters and Table 2 shows the welds produced.

Table 1. Analysed process parameters and their levels

Table 2. Experimental trials

2.1.1 Mechanical performance

Figures 3 and 4 show the main and interaction plots for the tests on the weld root, while Figures 5 and 6 show the results for the tests on the weld face.

For the bend tests on the root ( Figure 3), the main effects were due to blade thickness and length. Blade length showed a positive effect, while the thickness showed a negative one. The vertical load, horizontal pressure, and feed rate only showed a smallnegative effect. Also from the interaction plots between blade length and thickness ( Figure 4), higher mechanical properties of the joint are observed with a long, thin blade. Furthermore, from Figure 4, a reduction in blade thickness and an increase in blade length produce an increase in mechanical properties independent of the level of the other parameters analysed. The first order interaction between verticalload, horizontal pressure, and feed rate shows that the maximum strength of the root is obtained when these parameters are at their lower level.

For the bend test on the weld face, the main effects were due to blade thickness, vertical load, and horizontal pressure ( Figure 5). As for the root, the effects of all the other parameters except for the blade length are negative. From the interaction plot between blade length and thickness ( Figure 6), higher mechanical properties of the joint are produced with a long, thin blade. Horizontal pressure has a high negative effect; this may be because a high horizontal pressure produces a flow of the material towardthe face of the joint, which generates a large bead, and an associated sharper notch on the face. Furthermore, from Figure 6 a reduction of the blade thickness produces an increase of mechanical properties independent of the level of the other parameters. An increase in blade length generates an increase in ram displacement only when thehorizontal pressure and vertical load are at their low levels. The interaction between blade length and horizontal pressure could be explained considering that a long blade will produce more heat and will consequently melt morematerial and produce a larger face bead when the horizontal pressure is high. Furthermore, a high horizontal pressure will also increase the frictional heating. A high vertical load will increase the frictional heating from theshoulder and therefore the amount of molten material on the face.

Fig.6. Interaction plots for the face test ram displacement

2.1.2 Macrographic analysis

The macrographs of the joints produced in these trials are shown in Figure 7. In each macrograph the characteristic T-shape, due to the different action of the shoulder (horizontal part) and of the blade (vertical part), is clearly evident. On the face of the joints there is a weld beadproduced by molten material flowing towards the upper surface. This is more evident in the welds made with high horizontal pressure. Near the weld root a reduction of width of the melted zone can be observed. This was possibly due toheat loss from the backing plate.

Welds Vb01 and Vb03 show a lack of penetration, (see for example Vb01 in Figure 7). This could be due to the low vertical pressure that reduced the blade plunge, resulting in low mechanical properties of the root in these joints. This lack of penetration was not evident in joints produced with along blade (see Vb05 in Figure 7); the longer blade length should generate a higher heat input, which will allow more melt to flow to the root even when the vertical load is low.

Figure 7 shows also Weld Vb09, which was made using a short, thick blade and Weld Vb13, produced with a long, thick blade. A number of the joints produced with thick blades exhibited high porosity, which, it is believed, isdue to a reduction in pressure of the molten material behind the blade. This effect is more evident with the long blade probably because it supplied more heat than the short one, and consequently the melted material was hotter.

The width of the HAZ was measured for each macrograph at a position of 4.5mm from the bottom of the sheet. Figure 8 shows the main effect plot. Horizontal pressure and blade length cause a positive effect, due to the higher heat input produced when they are at their high level, while the positive effect of the blade thickness isdue to the greater amount of material that the thick blade displaces. The negative effect of the feed rate can be related to the lower specific heat energy (energy per unit weld length) that is generated when the sheets are welded athigher speed. No significant interactions were observed between the process parameters regarding the width of the HAZ.

Fig.8. Main effect plots for width of HAZ for 2 ^5-1 experimental plan 

2.2. Stage 2 - Evaluation of joint performance with a modified blade

Since it was shown in the previous section that a thinner blade produced a better quality joint, it was decided to reduce the blade thickness towards the root, i.e. taper the end of the blade, in order to improve the mechanicalproperties of the weld root (see Figure 9).

Fig.9. Vertically tapered blade

In this initial study three parameters were studied:

• Vertical load;
• Horizontal pressure;
• Feed rate.

Since a strong interaction of the process parameters was expected, a full 2 ³ experimental plan was carried out. Table 3 shows the chosen levels for the above parameters.

Table 3. Analysed process parameters and their levels

The analysis of variance (ANOVA) for the root bend tests is shown in Table 4 and indicates that only feed rate and vertical load are significant (low P-value). Figure 10 shows the main effect plot for the root test and shows that vertical load produces a high positive effect, while the feed rate produces a negative effect. This could be due to the high blade penetration and high heatinput produced when the feed rate is low and the vertical load is high. The effect of horizontal pressure is not significant. Figure 11 shows the interaction plots and indicates that vertical load has a positive effect independent of the levels of the other parameters. Feed rate always has a negative effect, but this is low when horizontal pressureis at a low level. A strong interaction can be observed between horizontal pressure and feed rate. This suggests that to get a high feed rate, which is important for production, it is necessary to weld with a low horizontalpressure.

Table 4. ANOVA for root and face bend test

Fig.11. Interaction plots for the root test ram displacement

Table 4 also shows the ANOVA for the face bend test and indicates a high influence of feed rate and of the interaction between vertical load and horizontal pressure. Figure 12 shows the main effect plots, in which the high negative effect of the feed rate is clearly highlighted. The interaction plot ( Figure 13) shows a strong interaction between horizontal pressure and vertical load; at a low level of horizontal pressure, the vertical load produces a high positive effect, however, when the horizontal pressure is at a highlevel the vertical load produces a negative effect. Also, when the horizontal pressure is at a low level the negative effect of the feed rate is smaller than when it is at a high level. The interaction between vertical load and feedrate is opposite to the interaction between horizontal pressure and feed rate and is also stronger.

Fig.13. Interaction plots for the face test ram displacement

The joints produced show some porosity in the top half of the joint, which resulted in a reduction of the ram displacement of the face. The two most likely reasons why porosity is occurring in this location are: 1) the greaterthickness of the top of the blade (with a consequently higher expansion of the plastic material behind the blade), and 2) a higher temperature caused by the shoulder. The shape of the vertically tapered blade therefore seems to havegreatly influenced the flow of the melted material in the joint, producing a better quality weld at the root but reducing the quality of the face.

Since the effect of the horizontal pressure is not significant, and because the feed rate effect is not as pronounced for both root and face when it is at a low level, it was deduced to carry out the response surface study usingonly the vertical load and feed rate, and keeping the horizontal load at its low level. A matrix of three vertical loads and five feed rates was tested ( Table 5). In this table the shaded cells are characterized by a high specific energy (energy per length of weld), due to either a low feed rate (Vb55, Vb56 and Vb62) or a combination of low feed rate and high vertical load(Vb61). These welds produced excessive flash and gas bubbles ( Figure 14). No mechanical tests were carried out on these welds. Figure 15 shows the top surface of Weld Vb49, which exhibits no bubbles or excessive flash.

Table 5. Matrix of welds for response surface analysis

In Figures 16 and 17 the contour plots for the face and root are shown, respectively.

To evaluate the acceptability of the joints a comparison with the bend test requirements for the hot gas and extrusion welding techniques was carried out. According to DVS 2203-1, ^[6] for the thickness and material studied a 13mm minimum ram displacement is required both for root and face: the two responses must be considered simultaneously.

Figure 18 shows the overlaid contour plots where the white zone is the region that satisfies the criteria (ram displacement > 13mm) for both root and face. As can be seen, with a high feed rate and low vertical load, bothroot and face exhibited poor mechanical properties. This is probably due to the specific energy being too low. This figure suggests that to increase the welding speed the vertical load should be in the range 2500-3000N.

Furthermore, these trials show that Viblade welding can produce good quality welds at a feed rate of about 150-180mm/min. In comparison, hot gas welding a 9mm thick sheet of PP requires an average weld speed of 15mm/min. Extrusionwelding the same material and sheet thickness can be performed at an average speed of 170 - 225mm/min.

 Fig.18. Overlaid contour plot for face and root test 

3. Conclusions

A preliminary investigation on a new variant of FSW, called Viblade ^TM welding, has been presented. A Design of Experiments approach was used to investigate the effect of vertical load, horizontal pressure, feed rate, blade thickness, and blade length on the HAZ dimensions and onthe mechanical performance of the joint. These trials were carried out with a titanium blade, which reduces heat loss from the weld due to its relatively low thermal conductivity. The blade length and thickness were shown to have asignificant effect on the weld quality, and the use of a long, thin blade produced joints with higher mechanical properties.

A vertically tapered blade was used to try and increase the mechanical performance of the joint root, and resulted in joints with bend test properties comparable with hot gas and extrusion welding. The main effects plot showed ahigh negative effect for the feed rate and a high positive effect for the vertical load. The effect of the horizontal pressure was not significant. Nevertheless, the interaction plot showed a high interaction between horizontalpressure and feed rate; when the horizontal pressure is at low level, the negative effect of the feed rate is small. This suggests that when horizontal pressure is at low level it should be possible to weld at high feed rates withoutan excessive loss of mechanical properties of the joint.

The overlaid contour plot for the face and root tests showed a wide zone in which Viblade welds could be produced with bend test ram displacement values greater than the minimum required values for hot gas and extrusion welding,according to DVS 2203-1, and at comparable weld speeds.

4. References

1. W.M. Thomas, E.D. Nicholas, Needham J.C., M.G. Murch, P. Temple-Smith, C.J. Dawes, International Patent Application No. PCT/GB92/02203.
2. A. Arici, T. Sinmaz, Effects of double passes of the tool on friction stir welding of polyethylene, Journal of Materials Science 40 (2005) 3313 - 3316 - Letters.
3. T.W. Nelson, C.D. Sorensen, C. Johns, S. Strand, J. Christensen, Joining of Thermoplastics with Friction Stir Welding, Proceedings, 2nd International Friction Stir Welding Symposium, Gothenburg, Sweden, 26-28 June 2000.
4. Leading edge - friction stir welding?, TWI Connect, March 1993.
5. BS EN 12814-1: 2000 Testing of welded joints of thermoplastics semi finished products - Part. 1: Bend Tests.
6. Directive DVS 2203-1 Testing of welded joints of thermoplastics semi-finished products - Test method - Requirements.