Charpy testing laser welds - the significance of fracture path deviation

TWI Bulletin, September/October 2001

A previous feature ^[1] described the first phase of work carried out with the aim of recommending a test method for qualifying the toughness of laser welds for shipbuilding use. It outlined the phenomenon of fracture path deviation (FPD) during Charpy testing, whereby a specimen, notched in the weld metal, does not fail in the plane of the notch; consequently the measurement of absorbed energy is not a true measure of weld properties. Isabel Hadley discusses the second phase of work aimed at qualifying laser welds for shipbuilding.

• a narrow weld zone
• increasing temperature (and toughness)
• increasing mismatch. ['mismatch' is defined here as the ratio of weld metal hardness to parent metal hardness, using hardness as a proxy for yield strength.]

Nevertheless, there is a view ^[2] that FPD will not occur in a genuinely brittle weld and that the occurrence of FPD therefore indicates that the weld has 'reasonable' toughness. The objective of the second phase of TWI's programme was therefore simply to test this hypothesis in the context of Charpy testing of laser welds for shipbuilding. Two possible outcomes, and hence two different approaches to Charpy testing, were envisaged:

1. A switch from fracture through the weld metal to failure by FPD is related to attainment of a particular value of Charpy energy, eg 27J, 47J. Under these circumstances, the occurrence of FPD at the test temperature could be regarded as a symptom of satisfactory inherent toughness.
2. FPD is related mainly to hardness overmatch, and occurrence of FPD can conceal poor inherent toughness. Under these circumstances it may be advisable to base the assumed toughness of production welds on their measured toughness atthe highest FPD-free temperature (or at the standard test temperature, whichever is the lower).

Method

• Aluminium foil was added to the bondline of the C-Mn steel weld tested in the previous phase of the work in a deliberate attempt to embrittle it. Results for embrittled material were compared with those from control welds ('narrow'autogenous welds).
• An HY80 weld known to be susceptible to temper embrittlement was tested in the as-welded and temper-embrittled conditions, and the Charpy transition curves compared.

In practice, it is difficult to alter the toughness of a weld without simultaneously altering its hardness and tensile properties, so two additional approaches were adopted:

• Carbon was added to the bondline of the C-Mn steel tested in the previous stage of the work with the aim of embrittling the weld. The carbon was added using shims of high-C steel; varying shim thicknesses were used, to producewelds containing 0.16, 0.21 and 0.24%C. The transition behaviour of the control weld was then compared with that of embrittled welds.
• Tests were carried out on two autogenous welds made in:

  - a carbon steel (BS EN 10025: 1993 grade S275JR, which has a specified minimum yield strength of 275N/mm ² and a parent metal Charpy requirement of 27J at room temperature),

  - a C-Mn steel (BS EN 10025: 1990 grade Fe510D, which has a specified minimum yield strength of 355N/mm ² and a parent metal Charpy requirement of 27J at -20°C).

  
  

Both steels had chemical compositions close to (or just beyond) the limit prescribed by current industry guidelines. ^[3] The welds were therefore expected to have lower toughness than the control weld.

None of the welds tested (except for the control weld) were expected to be representative of a ship laser weld; rather, they were made to test the hypotheses about FPD. As in the previous work, the hardness and mid-section width of the welds was measured, and a full Charpy transition curve generated. The temperature (or temperature range) associated with onset of FPD was then identified, as was the temperature at which 47J (or 27J) absorbed energy was observed. [These are typical Charpy criteria for high-strength and low-strength steels respectively.] Selected welds were also subjected to chemical analysis and metallographic examination; visual estimates were made of the different microstructural constituents.

Results

Table 1 Chemical composition of materials tested (weld metal analysed in the case of heterogeneous welds)

The properties of the welds are given in Table 2 in terms of weld metal hardness, weld width and mismatch ratio M (ratio of average weld metal hardness to average parent metal hardness). Details of Charpy test results are also shown, with an estimate of T _FPD, the temperature at which failure by FPD begins.

Results of hardness and Charpy tests

Three of the welds are very similar in terms of weld geometry and hardness mismatch: the control weld, the autogenous weld in Fe510D steel and the Al-treated weld. Metallographic examinations of these three welds showed the control weld to be almost fully martensitic, with very little variation in microstructure between the cap and root areas of the weld. By contrast, the Al-treated weld contained between 80% (cap area) and 85% (root area) martensite, and the Fe510D autogenous weld contained between 75% (cap) and 80% (root) martensite. The other constituent was ferrite with second phase.

Charpy transition curves for the three welds (N, Al, Fe510D) are shown in Fig.1. Results for the control weld show failure through the weld metal only at -196°C; FPD intervenes at all higher temperatures, as reported in phase 1 of the work ^[1] . Addition of Al to the bondline increases the temperature of onset of FPD, from below -160°C to somewhere in the range -90 to -80°C.

Fig.1. Charpy transition curves for the control weld, an Al-embrittled weld made from the same steel and an autogenous weld made from 0.16%C steel (Fe510D). All welds have similar levels of mismatch

The transition curve for the Fe510D steel shows a still higher transition temperature, with T _47J ≈+60°C and no cases of FPD. Compositional and microstructural differences between the steels probably account for the very different behaviours shown. Nevertheless, the results from these welds show that a highly overmatched brittle weld fails in the weld metal, not by FPD. This supports the idea that FPD is associated with good toughness and not just with highly overmatched welds.

Figure 2 shows a comparison of Charpy transition curves for two autogenous welds made in 0.16%C steel. One result refers to the Fe510D steel, as discussed above. The other relates to the S275JR steel, which is much less hardenable (IIW CE = 0.293) than the Fe510D (IIW CE = 0.395), in view of its lower Mn content. This is reflected in a much softer weld metal and a lower mismatch ratio, around 1.4 in the S275JR weldment, compared with 1.8-2.0 in the Fe510D. The trend of failure through the weld metal at low temperature/low toughness and by FPD at high temperature/high toughness is nevertheless observed, in spite of the fact that the overmatch ratio in the S275JR weldment is similar to that observed in many conventional arc welds, which do not fail by FPD. Presumably the width of the weld zone (narrower in laser welds than in typical arc welds) also determines whether or not FPD will occur. 

Fig.2. Charpy transition curves for 0.16%C steel weldments, made by autogenous welding. Different symbols in the same colour denote separate welds made using noninally identical procedures

The two welds in HY-80 steel are reasonably similar in terms of hardness overmatch. Weld widths are identical, since only one weld was made, half of which was heat-treated. The Charpy transition curves in Fig.3 show a fairly clear transition from failure in the weld metal to failure by FPD at temperatures of around -37°C (in the as-welded condition) or 35°C (after an embrittling heat-treatment). Embrittlement is thus associated with an increase of approximately 70°C in the transition temperature, but the pattern of change from failure in the weld metal to failure by FPD is apparently unchanged.

Fig.3. Charpy transition curves for an HY-80 steel in the as-welded and temper-embrittled conditions. Both welds have similar levels of mismatch

Fig.4. Charpy transition curves for 0.16%C steel weldments, made by addition of high C steel at the bondline. Different symbols denote discrete welds made using nominally identical procedures

Fig.5. Charpy transition curves for 0.21 and 0.24%C steel weldments, made by addition of high C steel at the bondline

The final two columns of Table 2 show the best estimate of T _FPD and of T _27J and T _47J. There is a strong correlation between T _FPD and the corresponding transition temperature (see Fig.6), supporting the idea that FPD is associated with good toughness. Note, however, that T _FPD could not have been estimated accurately for the C-enriched welds ( Figs 4 and 5) without a knowledge of the full Charpy behaviour, including the bimodal Charpy curve. 

Fig.6. Correlation between temperature at which FPD intervenes and the 27 and 47J transition temperatures for laser welds

Discussion

The decision to use the narrow-profile weld ^[1,2] as the control weld in this programme was based on the fact that its width (between 1 and 2mm) is the lowest acceptable under current guidelines. The combination of high hardness overmatch and narrow weld width make it highly susceptible to FPD. Consequently, if the issue of FPD can be understood in a weld of this type, other practical shipbuilding welds will present less of a problem. More importantly, the high travel speed (1 m/min.) used for this weld makes it attractive to shipbuilders, provided that the problems associated with measurement of toughness can be resolved.

The decision to test several weldments which were borderline with respect to suitability for laser welding ( eg 0.16%C, 0.005%S, 0.01%P) was based on the LR guidelines. The principle was that if CO ₂ laser welds of appropriate strength, hardness and toughness could be manufactured even at the limits of the range currently considered acceptable in industry guidelines, this would reduce considerably the need for future qualification tests. In practice, obtaining materials with the desired borderline chemical composition was not possible and some of the steels tested would be considered unsuitable for laser welding, typically because of excessive S and P contents. Sound welds free of solidification cracking were nevertheless made.

The laser power output, speed of welding and section thickness (hence the weld cooling rate) were broadly similar in all welds tested. Because of differences in chemical composition from weld to weld, there was considerable variation in microstructure, weld metal hardness, weld metal toughness and FPD behaviour. Most of the results supported the hypothesis that welds that fail by FPD have 'reasonable' Charpy energy. For example, when the control weld was embrittled by addition of Al at the bondline (without significant change of hardness properties), the effect was to shift the Charpy transition curve to the right. Failure through the weld metal continued to occur in 'brittle' welds, roughly defined as those absorbing less than 27J on fracture. Similar behaviour was observed in HY-80 steels subjected to a temper embrittlement treatment.

An exception to this was observed in highly overmatched welds in which carbon was added at the bondline to produce fully martensitic weld metal with carbon contents of 0.16-0.24%. In some cases, there was a considerable temperature range over which failure could occur either through the weld metal or by FPD. The conventional practice of testing only three Charpy specimens could be potentially misleading under these circumstances. It is statistically possible to observe failure by FPD (and very high Charpy energy) at some specified test temperature, but then to observe failure through the weld metal (and Charpy energy below 27J) at some higher temperature in such welds.

Numerous replicate tests were carried out on these C-enriched welds at temperatures close to the FPD transition temperature, in an attempt to define how many Charpy specimens should be tested for a weldment susceptible to FPD. Because of factors such as weld-to-weld variations in properties between nominally identical joints, it was not possible to make a simple recommendation on the appropriate number of specimens required to approve a CO ₂ laser weld for ship service. Furthermore, previous work ^[1] showed that none of the simple modifications suggested for prevention of FPD ( eg side-grooving, use of side welds) is fully satisfactory. It is therefore suggested that generation of both hardness data and a full Charpy transition curve should form part of the qualification of CO ₂ laser welds. There is a case for relying on FPD as an indicator of good toughness if some or all of the following properties are met:

• FPD occurs well below the normal specified Charpy test temperature
• 'bimodal' behaviour does not exist over a significant temperature range
• chemical composition is within industry guidelines
• fracture toughness and/or large-scale test behaviour are satisfactory
• satisfactory service experience exists

Weldments which do not meet the above should be subjected to more exhaustive testing, possibly to include:

• fracture toughness testing (FPD is less likely to occur in fracture toughness tests than in Charpy tests, because the former uses a sharp crack rather than a blunt notch)
• a larger number of Charpy tests than usual, to ensure behaviour of the weldment is well outside the brittle range

Summary and conclusions

• In most cases studied ( eg Fe510D, Al-embrittled weldments), brittle welds failed in the weld metal, where tough welds with a similar level of overmatch failed by FPD. This supports hypothesis one.
• In certain cases ( eg highly overmatched welds containing unusually high carbon concentrations at the bondline), the coexistence of the two failure modes over a wide temperature range (up to 70°C) suggests that the occurrence of FPD is nota sufficient indicator of good toughness in these cases, since the occurrence of FPD can conceal poor toughness in the weld metal.
• Some results (S275JR) show the occurrence of FPD in spite of relatively low overmatch ratios

There are a number of possible approaches to treating the occurrence of FPD in laser welds:

• Generate a complete Charpy transition curve so as to identify possible 'bimodal' behaviour. If there is little or no bimodal behaviour, then subsequent testing can be based on standard Charpy testing at the usual test temperature.The criterion would then be the attainment of a particular Charpy energy and/or observation of failure by FPD.
• Test at the lowest temperature at which FPD does not occur (or the standard test temperature, whichever is the lower). The usual energy requirement holds, with the additional requirement that failure should occur wholly through theweld metal.
• If bimodal behaviour occurs, it may be possible to derive a Charpy energy criterion based on testing more than the usual three specimens, to eliminate the possibility that the tests miss low-toughness behaviour.

Whichever approach is adopted, it is clear that at present Charpy testing of laser welds requires considerably more expertise in interpretation, and possibly a larger number of test specimens, than is the case with conventional fusion welds. If the structure is fracture-critical, then it may be prudent to qualify the weld for the first time by carrying out fracture mechanics tests. A Charpy criterion could be derived by correlation, and the Charpy test alone used thereafter. Work at TWI is continuing with a view to understanding FPD through numerical modelling and application of the local approach to fracture.

References