A C Woloszyn*, S Fisher** and R L Jones*
*Arcs, Laser and Sheet Processes Department, TWI, Granta Park, Great Abington, Cambridge, CB1 6AL, United Kingdom
**BNFL plc, Warrington, Cheshire, WA3 6AS, United Kingdom
Presented at JOM-9, 9th International Conference on Joining of Materials, Helsingør Denmark, 16-19 May 1999
Austenitic stainless steels tend to form large columnar grains in weld metal, and in extreme situations the grains can continue to grow through successive weld beads. This is detrimental to both weld mechanical properties and inspectability.
Work has been carried out at TWI on evaluating the effect of agitating the weld pool by applying ultrasound, with the aim of refining the microstructure in austenitic weld metals.
Metallurgical analysis and ultrasonic inspection were performed on the welds. Optical microscopy of cross-sections taken from the welds revealed a more refined structure in areas corresponding to maximum ultrasound energy within the plates. Ultrasonic NDE showed that the welds with the refined structure gave improved ultrasound transmission characteristics.
Austenitic stainless steel will form large columnar grains in the weld metal. These initiate from the pool edges growing towards the weld centre and, under certain grain growth conditions, very few equi-axed grains are formed. In extreme situations, the grains can continue to grow through the structure, i.e. through successive weld beads. This is detrimental as both weld mechanical properties and inspectability are adversely affected by increased grain size.
Ultrasonic NDE is often the most practical method to assess weld soundness. Large grains attenuate ultrasound reducing the effectiveness of ultrasonic NDE. This, in turn, increases repair rates and safety margins as small defects can not be readily detected.
Abramov evaluated the effect of ultrasound on the microstructure obtained with ferritic and austenitic steel castings  . He found that a refined structure was produced provided 800 W of ultrasound per kg of molten metal was applied. This work evaluated the effect of agitating a TIG weld pool, by applying ultrasound, on weld metal grain size.
The base materials were 100x100mm 304L stainless steel plates. 1.6mm dia. Grade 308 welding wire was used for all the welding, together with a pure argon shielding gas.
2.2 Welding procedure
For the trials, a number of 100x100mm square, 7mm thick 304L plates were used. These were prepared by machining a 60° groove of 6mm depth along the centre line of each plate. The ultrasound apparatus and plate specimens to be welded were fixed to a traverse that allowed mechanised TIG welds to be performed in the PA, flat position.
A series of TIG welds was made, which comprised bead in groove welds along the length of each plate. All the trials were made with a single pass, and produced at 175A and 10V. Various travel speeds and wire feed rates were used.
2.3 Ultrasound equipment
The method by which ultrasound was transferred to the weld pool is illustrated in Fig.1
Fig.1 Plate/Sonotrode fixture
The ultrasound was introduced into the plate being welded, underneath the joint using a sonotrode specifically designed for this system. The sonotrode was made from titanium because of its good ultrasound transmission characteristics. Also this metal would conduct heat from the joint area without being melted by the weld pool. The plate to be welded was fixed to the sonotrode as illustrated in Fig.1 using six M5 bolts.
To generate the ultrasound, a 2kW power source capable of generating ultrasonic vibrations at 20kHz was used. The power values available ranged between 0.2 and 1kW. The 1kW output level approached the maximum available from the system, which could be sustained for periods necessary to perform the weld.
2.4 Ultrasound parameters
Two modes of coupling were investigated; either the plate was tightly fixed to the sonotrode or it was held loosely and could vibrate separately from the sonotrode. With the bolts loose, the plate was free from the ultrasonic system with the sonotrode constantly hammering onto the plate. This mode was found to be the noisiest but was the least susceptible to changes in resonance as the weld metal was added. The second mode was to tighten the bolts so that the plate was part of the ultrasound system. In this way the plate was an extension of the sonotrode and the ultrasound could enter directly into the weld pool.
Prior to depositing welds subjected to ultrasound, the plate and sonotrode were 'tuned'. That is, the frequency of the power source was adjusted so that the system was in resonance, thus maximising the ultrasound input. By observing sand sprinkled into the groove of the prepared plates, the resonant frequency was found. The sand was seen to bounce above and out of the groove when resonance was achieved. The ultrasound generator power consumption was found to dip around resonance. If the frequency was adjusted away from resonance, the current drawn by the resonator increased.
The power values were arbitrarily chosen, limited primarily by the power source. A range of values was chosen. The lowest value of 0.2 kW is still in excess of the value identified by Abramov  (0.8 kW per kg of metal) to produce grain refinement. The higher value of 1 kW was close to the maximum available from the system. The power source will deliver 2 kW, but at 1 kW, the maximum current capacity of the resonator windings had been reached.
2.5 Metallographic examination of welds
Each weld was sectioned transverse to the welding direction, and approximately ¼ or ¾ along the length of the weld. In addition, two welds were sectioned at 5-7mm intervals along the length of each weld, in order to observe any variations in the microstructure along each weld run. Both of the latter specimens had been welded with the same level of ultrasound input (0.8kW).
All sections were ground, polished and etched with a 20% sulphuric acid solution; then examined by optical microscopy.
2.6 Ultrasonic inspection of welds
Five of the mechanised TIG welds were skimmed to remove any excess metal. One mm and 0.8mm diameter holes were drilled into each plate, along the weld and from either end of the plate. These samples were then inspected using ultrasonic NDE. The following techniques were applied:
- 0° probe measurements at 2.25 and 10 MHz.
- 45° shear wave measurements at 5 and 10 MHz.
- Pitch and catch system operating at 10 MHz.
The ultrasonic test procedures are detailed in Table 1.
3.1 Observations made during welding
Throughout the welding trials, several notable observations were made. Firstly, in tuning the sonotrode/plate prior to welding, during resonance the sand on the plate was seen to be more active in certain areas. This is illustrated in Fig.2
, the sand being seen to 'jump' in the areas between the bolt holes. Next to the holes, and in certain areas on top of the plate, the sand would be relatively inactive; tending to pool. This suggested an uneven distribution of ultrasonic energy across a given plate.
Fig.2 Observation of sand during resonance
Also, during a weld run with ultrasound, there was a significant variation in input current from the ultrasound power source throughout the respective weld. This followed tuning to the resonant frequency, and pre-setting the current and voltage to give the correct power; after which the weld runs would be performed. During the time it took to complete the weld, the current would fluctuate around the original level by 0.25-0.5A. This also represents a fluctuation in the power input into the plate/weld during each respective weld run.
3.2 Metallographic examination of welds
The results from metallographic examination for two specimens (tightly coupled to the sonotrode) are shown in Figs 3
. There is a marked level of refinement for the welds where ultrasound was applied. Given that the grain boundaries are not clearly visible, it was difficult to discern the true level of grain refinement. However, refinement in the structures which were welded with ultrasound is clearly visible in the dendrite microstructure. The greatest effect was noticed at the weld pool edge where the columnar grains predominate in weld metal where ultrasound was not applied. As the level of ultrasound is increased so the microstructural refinement is also increased, this was most marked with the tight coupling.
However, this result was not consistently seen for all the matching specimens welded with/without ultrasound that were compared. For example, one pair of welded specimens showed very little difference in microstructure, despite the fact that one specimen had received 0.8kW ultrasound, and the other received no ultrasound. This was the case with several cross-sections taken from specimens welded with/without ultrasound.
Optical microscopy of the transverse weld sections taken along the length of two specimens welded with ultrasound (again using a tight coupling) revealed no notable differences in structure between both specimens, and those receiving no ultrasound, up until the sections corresponding to the dashed line on Fig.2. A micrograph of one of these sections is shown in Fig.5. Apart from the solidification crack, the microstructure is much finer than that of Fig.3, and shows a mottled appearance towards the lower/central regions of the weld bead. Fully austenitic regions, with no ferrite were identified. The solidification crack was observed to pass almost entirely through these regions.
Fig.3 Transverse section of specimen welded without ultrasound
Fig.4 Transverse section of specimen welded with 0.8kW ultrasound
Fig.5 Transverse section taken from position indicated by the dashed line on Fig.2, welded with 0.8kW ultrasound
3.3 Ultrasonic inspection of welds
The NDE results are summarised in Table 1. It was possible to detect both the flaws (i.e. pre-drilled holes) in all the weld samples, irrespective of any ultrasound used during manufacture. However, differences were noticed in the attenuation of the ultrasonic NDE signal, particularly at the higher frequency. This is consistent with attenuation due to larger weld metal grain size. Using the results from the pitch and catch technique the welds were ranked according to ease of ultrasound transmission, these are listed in the Table.
|Sample No.||Ultrasound power, kW|
L - loosely attached
T - tightly attached
|Ranking for ease of NDE|
1 - easiest
5 - hardest
|Measurements at Normal (0°) incidence||Shearwave 45° 1mm diameter Hole Target||Pitch-|
|Shearwave 2.2MHz 1|
No. of Echoes >80% FSH
||Amplifier GAIN for Amplitude of 80%FSH
The microstructures observed indicated that the weld became more refined as ultrasound is introduced to the molten weld metal. This refinement is most prominent at the weld pool edges where the columnar structure is replaced by a more equiaxed dendritic structure (see Figs.3
). The effect of increasing the ultrasound power was to increase the degree of refinement, but the coupling mode also had a significant effect on the level of refinement i.e. tighter coupling resulted in greater refinement.
As mentioned, the ultrasound generator power consumption was found to dip around resonance, the current drawn by the resonator increasing if the frequency is adjusted away from resonance  . However, during a weld run this frequency is altered by the addition of the filler metal, resulting in a given plate alternating in and out of resonance. Thus, it appears that some sections taken from ultrasonic welds showed a change in microstructure, whilst others did not; depending on whether the section fell through an area that received a sufficient level of ultrasonic energy.
The transverse sections taken along two welds both showed the greatest level refinement to lie in the area indicated by the dashed line in Fig.2. The fact that this area corresponds to a high resonance area (i.e. active region of sand) further suggests that the refinement observed is due to the application of ultrasound energy. The solidification crack observed travelling through the austenite rich areas of the weld, can be attributed to the high restraint imposed on the welded plates, necessary to transfer the ultrasound to the weld pool, coupled with the tight v-joint design.
The lack of consistency of the grain structure within the welds made using ultrasound would appear to be due to a combination of the following:
- The individual resonance characteristics of the plate to be welded.
- The addition of filler metal altering the resonance of the plate being welded.
- The increased level of ultrasound energy at the midpoint between bolt holes in a given plate, and lower energy 'dead zones' near the bolt holes.
The refined grains did improve ultrasound transmission through the weld for NDE purposes. At only 7mm thick, there was insufficient difference between the microstructures for any significant difference in defect determination to be possible. The pitch-catch technique at 10 MHz showed the greatest difference between the welds because this was at the higher frequency (therefore more readily attenuated) and was transmitted through the greatest length of weld metal (so having the most chance of being attenuated).
If the weld had been bigger the difference between the 'natural' weld and the 'ultrasound' weld grain sizes would most likely have been larger. This would have given a more definitive conclusion from the NDE inspection.
- The application of ultrasound during welding has been shown to refine the weld metal microstructure. The effects of the ultrasound power and clamping technique are critical for the degree of refinement achieved.
- Ultrasound input into the weld-pool can lead to a finer dendritic structure.
- The finer microstructure was shown to enhance NDE ultrasonic transmission.
- For the apparatus used in this investigation, there is an uneven distribution of ultrasound energy across a clamped resonating plate.
- The addition of filler metal alters the resonant frequency of a plate being welded. As such, a plate receiving ultrasound during welding fluctuates in and out of resonance during the weld run. The result was significant variability in the degree of refinement observed along the weld joints.
||'The Action of Ultrasound on Solidifying Metals', Advances in Sonochemistry Vol. 2. JAI Press 1991, pp 135-186.