Bimodality revisited - split behaviour of weld metal

TWI Bulletin, September/October 1991

After several years in industrial research, Norman Bailey joined TWI in 1966. He is now a Principal Consultant Metallurgist in the Materials Department where he led a Research Section concerned mainly with welding of ferritic steels.

His researches into the problems of hydrogen cracking contributed to a method of predicting safe welding procedures now used worldwide. He has also led research into solidification cracking during submerged-arc welding and examined the problems of welding very high-strength steels. More recently he has been concerned with understanding how the composition of ferritic steel weld metals can be controlled to achieve consistently the required microstructure, strength and toughness properties in both normal and underwater welding. He has also led projects examining repair of pressure vessels without subsequent heat treatment, strain ageing behaviour of the weld metals used for such repairs, and development of electrodes for wet welding.

He has written numerous reports and papers including: 'The establishment of safe welding procedures for steels' Weld J 1972 51 (4) 169s-177s, which was awarded the Lincoln Gold Medal of the American Welding Society in 1972 and, with Dr S B Jones, 'The solidification cracking of ferritic steel during submerged-arc welding'. Weld J 1978 57 (8) 217s-231s, which gained the American Welding Society's Spraragen Award in 1979.

Bimodal behaviour - leading to occasional low Charpy values - and its implications were well understood 20 years ago. Now it's almost forgotten with the result that new specifications for weld metal lead to wasteful re-testing, and even disputes over anomalously low impact test results. Norman Bailey summarises what is known about bimodality in mild steel weld metal and offers guidance for specification writers.

Some scatter is inevitable when carrying out toughness tests on steel near its transition temperature. When adjacent specimens are tested at the same temperature and give values differing by a factor of seven or more, it is a cause for disbelief and usually taken as a sign that something has gone seriously wrong.

This is not necessarily the case, as the scatter is probably an example of bimodal behaviour, which was understood and accepted in the fifties and sixties, but currently seems to have been forgotten. Bimodal behaviour means that there is not a smooth transition in toughness between the upper and lower shelves of a toughness/temperature curve. Within the transition temperature range, specimens either break in a ductile manner with high toughness or in a brittle manner with low toughness.

An early investigation into scatter ^[1] , showed Charpy values in a 100mm thick weld ranging from 10-222J. Although the scatter was largely associated with varying proportions of coarse (as-deposited) and refined microstructures in the specimens tested ^[2] , bimodal behaviour was not apparent, largely because the specimens had originated from different parts of the weld, including the root, which tended to give lower values than elsewhere in the weld.

Some years later, Widgery examined the toughness of all-weld deposits from 12 different basic mild steel electrodes of AWS codings E6016, E6018, E7016, E7018 and E7028 types ^[3] . Clear evidence of bimodality was found in three of the welds (E6016, E6018 and E7016 types). These are illustrated in Fig.1, together with the results from a weld in the same investigation which showed normal behaviour ( Fig.1d). The three bimodal curves show an overlap of upper and lower shelf Charpy results with few values lying far from either the upper or the lower curve.

The greatest scatter was found in weld B11 at -10°C, where individual values of 19, 22 and 146J were measured. The curves overlapped over a temperature range of between 50 and 80°C. The compositions of Widgery's weld metals were lower in manganese than would be common today, the overall range being 0.5 - 1.3%Mn, the welds showing bimodality containing 0.7 - 1.1%Mn. However, bimodality is not confined to old-fashioned electrodes, and the author is aware of bimodality in a weld containing ~1.4%Mn made with a more recently developed electrode, where adjacent specimens gave Charpy values of 20 and 145J at the same temperature ^[4] .

Bimodality is not confined to weld metals, and was observed by Crussard and his colleagues ^[5] with open hearth killed and rimming steels, as well as with basic Bessemer steels and a low alloy, Ni-Cr steel. In this classic paper, bimodal behaviour is shown to be indicative of the existence of two modes of fracture co-existing at the same temperature. The authors observed this type of behaviour not only for Charpy specimens (both V and U notched), but also for slow notched bend specimens and for plain tensile test specimens, provided that the test temperature was sufficiently low.

A systematic, statistical investigation was carried out on some 4000 Charpy test results on welds on mild steel deposited with basic mild steel electrodes. ^[6] A typical set of results, reproduced in Fig.2, shows an almost complete separation between the low and high values from 0°C down to -80°C, although at both extremes of this range there was only one odd value. From another set of results in Fig.3a, it can be seen how the proportion of high values steadily falls with decreasing temperature. Increasing the alloy content of the weld metal ( Fig.3b) steepens the transition, so that bimodality is apparent at only two temperatures. Normalising the mild steel weld metal ( Fig.4) does not prevent bimodality, although it refines and homogenises the microstructure and apparently reduces the scatter. In this investigation, bimodality was obtained with a wide range of welding conditions and was not influenced by the welder nor by the position of the test specimen along the weld, nor was the result of an individual specimen related to that of the neighbouring specimen.

Test specimens from the upper curve showed most deformation at the root of the notch, i.e. they broke in a ductile manner after considerable deformation of the notch root. From the lower curve, the specimens showed most deformation away from the root, i.e. typical of a mixed fracture, which initiates as cleavage from the root region and finishes by ductile failure of the remaining narrow ligament away from the root.

Statistical analysis of specimens tested at a single temperature showed that, for each type of fracture, the energy values each lay close to a normal Gaussian distribution ( Fig.5). Neumann et al concluded that bimodality is a fundamental phenomenon and cannot be removed by altering the notch form, notch orientation and weld metal structure in the fracture cross section. ^[6]

In recent years, however, the phenomenon of bimodality appears to have been forgotten, and a recent review of scatter when testing the toughness of weldments ^[7] appears to ignore bimodality completely.

Discussion

For bimodality to exist, it is likely that, in addition to two possible fracture modes, there must be some variability between specimens, either in regard to the presence of local brittle regions, or in their positioning in relation to the notch in the test specimen. Otherwise, the lower energy fracture initiation mode would always dominate, and bimodal behaviour would not be found.

Steels available over 30 years ago probably contained just sufficient, relatively large non-metallic inclusions or other imperfections to initiate cleavage fracture in some specimens and not in others. These imperfections occurred with sufficient frequency to give the bimodal behaviour found by Crussard and his colleagues. ^[5] It is likely that a gradual clean-up of high quality steels (i.e. the steels likely to be Charpy tested) over the years has reduced the incidence of such bimodality to zero. In fact, the author is aware of a Cr-Mo steel, recently tested at TWI, which has stubbornly refused to show transition behaviour; specimens were either fully ductile above a single transition temperature, or fully brittle below it. ^[8] The modern equivalent of bimodality in steel is found (admittedly rather infrequently) when Charpy tests are carried out on the centreline segregate region of continuously cast plate. Here, the author has encountered adjacent Charpy specimens in a thick microalloyed C-Mn steel plate which gave values of 140, 22, and 11J at the same temperature; specimens outside the segregate band gave over 200J.

In weld metals, bimodal behaviour now appears to be rare. A survey of recent relevant TWI research reports has revealed no examples of bimodality in Charpy transition curves of submerged-arc, MIG or MMA welds, although three possible examples were found in flux-cored wire welds, ^[9] reproduced in Fig.6a-c. Here, the available data are not sufficient to be as certain as in Fig.1 that bimodality was present, but it is reasonable to assume that it was. In his extensive papers on the mechanical properties of all weld MMA deposits, Evans mentions bimodality only once, ^[10] where he observed that scatter increased as carbon was decreased and that at 1.4%Mn, 'full bi-modal fracture occurred at -40°C'. The scatter was from approximately 30 to 220J at this temperature.

It has been said ^[11] that, with MMA deposits, bimodality is only found with certain brands of electrode. If this is so, its present rarity could be explained if electrode manufacturers knew how to avoid the problem, even if they do not know what causes it. One explanation for bimodality in mild steel multipass weld metals is that they contain a range of microstructures, namely soft, coarse, grain boundary (primary) ferrite; harder, coarse ferrite with aligned second phase; fine acicular ferrite; and fine, reheated weld metal refined by the succeeding weld run.

Of these, the last two are the toughest and the ferrite with aligned second phase the least tough. The amount and precise location of these different microstructures in the notch root area control the toughness of a particular specimen, and in certain circumstances can lead to a sharp difference between tough and brittle behaviour with no intermediate transition.

Even if certain electrodes may normally give Charpy curves free from bimodality, small changes in several factors are likely to bring about its re-appearance. The factors could include the change from all-weld tests to tests on butt welds, a change in welding position (as can be seen by comparing Fig.6c with 6d), dilution of elements from the parent plate influencing weld composition, a change in size of Charpy specimen (as suggested in Ref. ^[5] ), a change in the proportions of refined and as-deposited microstructure, straining or strain ageing of the weld metal, or even a small change in interpass temperature. Nevertheless, the cause of bimodality in MMA deposits is not yet known, even though some effort was expended in investigating the problem at TWI and elsewhere after the results in Ref. ^[3] and ^[6] had been reported.

Bimodality and standards

Severe scatter caused by bimodality is implicitly recognised in specifications for manual electrodes. The international standard ISO 2560, ^[12] and standards derived from it, take account of scatter and ignore isolated low results. The requirement at the appropriate test temperature is for an average minimum value of 28J from 18 specimens, although if the first six specimens tested exceed an average of 35J (or fall below 16J), no further testing is needed and the test is deemed to have passed (or failed). A different approach is used in the American standard AWS A5.1, ^[13] where five Charpy specimens are tested. The highest and lowest values from these are discarded. The average of the remaining three must be at least 27J, two of the three must exceed 27J and the third must not be below 20J.

Unfortunately, a less realistic view is apparent in most applications standards. These usually ignore the possibility of an isolated low result and impose the same requirements as for the parent steel, e.g. the average energy from three specimens shall exceed the stated value and no single value shall fall below about 70% of that value. If the value is set at 27J (as it is in BS 5500, ^[14] for steels of less than 450 N/mm ² tensile strength) there may not be much of a problem, as most of the lower shelf values in bimodal regions (e.g. Fig.1 and 3) appear to exceed 70% of that value, i.e. 19J. However, the modern trend, particularly with higher strength steels, is to require higher Charpy values, and 70% of such higher values would lie above the lower shelf in the bimodal region. For example, BS 5500 requires 40J minimum for steels (and weld metals) stronger than 450 N/mm ² tensile strength; 70% of this value is 28J, well above the lower shelf within the bimodal region of several curves in Fig.1 and 3.

One consolation is that, for these higher strength steels, the weld metal needs to contain a high proportion of fine acicular ferrite in the as-deposited microstructure to achieve the required strength and toughness. This type of structure is likely to give as-deposited and refined microstructures of comparable ferrite grain size and hence comparable toughness, so that bimodality would be less likely or, if present, less marked, as can be seen in Fig.3. Nevertheless, designers and others should be aware of bimodality, and consider reducing the test temperature, rather than increasing the Charpy energy requirement, when a tougher weld metal is required.

Fracture toughness

Apart from the significant difference from Charpy testing of giving practically useful values for assessing fracture behaviour, fracture mechanics tests are much more discriminating, in that they are used to assess fracture initiation behaviour in a small zone just beyond the tip of a sharp crack. Thus, they are more likely to show up bimodal, or even multi-modal behaviour. Indeed, test specimens with significant pop-ins could be considered to be exhibiting bimodality in a single specimen.

However, because weld metals usually contain several different microstructural regions (e.g. as-deposited, refined, as-deposited sub-critically reheated, and as-deposited intercritically reheated), a wide scatter of values is seen in the transition region, rather than the simple bimodality illustrated in Fig.1-3. Unlike bimodal behaviour in Charpy testing, scatter in CTOD testing is recognised in assessing the toughness of the different microstructural regions of weld metal. In many cases, the lowest toughness value of those measured is commonly used for calculation to determine critical defect sizes, etc, when making fitness-for-purpose assessments.

Practical implications

Bimodality appears to be a problem found mainly in unalloyed mild steel weld metal and is nowadays probably rare because the more common, tough basic electrodes are sufficiently highly alloyed with manganese to give a sharper transition and less likelihood of isolated low Charpy energy values.

However, it is important to be aware of the phenomenon when testing such materials.

For example, low strength weld metals may nowadays be selected to reduce residual stress levels to minimise risks of lamellar tearing and other types of cracking. In such low strength weld metals, an occasional low Charpy value is to be expected and is not the sign of a deterioration in weld quality. If one is faced with such an isolated low value, the first action is to check that no defects can be seen on the fracture surface which might have led to reduced energy absorption.

The second step is to check whether the weld metal microstructure is sufficiently varied, i.e. with fairly large coarse and fine grained regions, which could be expected to give rise to bimodality. The third step depends on individual circumstances. One option would be to test one or more additional sets of Charpy specimens (as in ISO 2560) to assess whether the low value is a truly isolated phenomenon (such as the single low value at 0°C in Fig.2). Alternatively, particularly if the tests are being made to assess the safety of a structure, fracture mechanics testing and a fitness for purpose analysis are likely to be more appropriate.

However, the writers of standards and specifications also need to be aware of bimodality and not set the minimum values too high; it is more sensible to set a low value at a lower temperature. Weld metals are not wrought steels and do not behave as such.

Summary

Bimodal behaviour, occasionally seen during Charpy testing of mild steel weld metal and other forms of steel, occurs within the ductile-brittle transition range so that specimens fracture either in a ductile manner with a high absorbed energy or in a brittle manner with low energy. The two populations are quite distinct and the energies may not overlap.

Nowadays the problem is rare, but may still occur and give rise to unnecessary concern. It is particularly important that writers of standards and specifications for mild steel weld metals are aware of the phenomenon and do not set unrealistically high values for single Charpy energy values; 19J is a realistic lower value. If tougher weld metal is needed, then a lower test temperature should be selected.

References