Proposed standard procedures for the determination of diffusible hydrogen in submerged-arc weld metals

TWI Bulletin, May 1985

Norman Jenkins, AMet, is Head of the Chemical Laboratory in the Materials Department.

To optimise submerged-arc welding procedures to avoid hydrogen cracking it is necessary to develop a standard procedure for determining the contribution to weld hydrogen levels which is made by the welding consumables. This article describes the results of a co-operative analytical trial carried out by four laboratories to assess the suitability of a particular procedure for measuring diffusible hydrogen in submerged-arc weld metal.

The increasing use of the submerged-arc welding (SAW) process has led to the need for measurement of diffusible hydrogen in submerged-arc weld metal. Proposals within the International Institute of Welding (IIW) ^[1] suggest the possibility of using the same sample dimensions as those described in the manual metal arc (MMA) standard, ISO 3690, ^[2] ( Fig 1a) this having the advantage that the size of sample is compatible with existing analytical equipment. Although arc energies and welding speeds used with the SAW process are typically much greater than those of the MMA process, resulting in a larger weld bead width and weld pool crater, to some degree these increases may be accommodated by rotating the sample through 90° as shown in Fig 1b.

Production of a sample for diffusible hydrogen analysis requires the use of a welding jig to provide clamping and cooling of a testpiece assembly during welding. An improved welding jig was therefore designed at The Welding Institute, ^[3] a detailed sampling and analytical procedure specified, and the method was used by a number of co-operating laboratories to assess its reproducibility.

An essential part of the development of an analytical procedure is the statistical evaluation of the performance of the method. Details are given of a collaborative trial carried out by four laboratories using the jig with subsequent hydrogen measurement following ISO 3690. The work described in this article was carried out as a Group Sponsored Project with the objective of enabling recommendations to be made regarding standardisation of the procedure.

Design of new welding jig

The design of welding jig used was based upon the results of trials using three types of jig ^[3] which assessed their suitability for preparing submerged-arc test welds which were acceptable for measuring diffusible hydrogen by the standard ISO method. ^[2] These trials showed that the run on length of the testpiece assembly, 100mm, was insufficient to allow stable weld conditions to be established before the analysis sample was reached. Additionally, it was noted that the weld pool, which was typically about 75mm long, was barely clearing the analysis sample before the arc was extinguished. In order to overcome these faults the length of the run-on and run-off pieces was increased to 150mm so allowing a weld traverse of 200mm to be used.

Because this increased length was greater than could be accommodated by the three designs of welding jig which were tested, a new welding jig was designed and built by The Welding Institute. After welding trials to confirm its suitability for making test welds, plans were circulated to the collaborators who each built an identical welding jig.

The new design had four features which were intended to promote easy and reproducible use:

1. The increased overall length, which allowed adequate length for the run-on and run-off pieces;
2. Water cooling to promote quicker cooling of the jig between welds;
3. A quick clamping lever to allow rapid release of the sample on the finish of welding;
4. A continuous copper base plate which reduced the possibility of weld burnthrough.

This welding jig is illustrated in Fig.2.

Collaborative trial

Approach

The collaborative study of submerged-arc weld diffusible hydrogen measurement as described below drew upon the experience gained by the participants during a similar study on the manual metal arc welding process. [4] For this reason, factors which contribute to the variability of the method, such as sample cleaning, collection time and the influence of welding upon the results, were not individually studied.

With any analytical procedure, it is necessary to make a precise and reliable statement about the reproducibility of the results obtained. The assessment of this reproducibility must include both within and between laboratory variations, and involve replicate determinations by a number of laboratories.

The statistical examination of the procedure for the determination of diffusible hydrogen in submerged arc welds is based upon BS 4237: 1967. [5] The experimental design specified in this Standard states that where five or six laboratories are taking part, then it is sufficient for each laboratory to carry out five determinations. Regarding the two Welding Institute operators as separate laboratories, there were initially six laboratories, but one of these withdrew at an early stage of the work. It was therefore agreed that each of the remaining five participants should make six determinations of diffusible hydrogen. This was also a practical limit imposed by the availability of eudiometers in some laboratories. In accordance with the Standard each laboratory carried out the six analyses in a single batch.

Materials

Welding wire

The work being carried out was aimed at establishing the reproducibility of a specific sample production procedure. It was therefore necessary to avoid any possibility of variation in the analysis results obtained that might stem from non-uniformity of the welding consumables, such as differences in wire surface condition.

It was thought that a degassed wire would overcome this difficulty and a 20kg reel of 4mm diameter SD3 wire was obtained which was divided into six portions. The potential hydrogen content of the wire was determined by an encapsulation technique [3] and found to be 0.8ml H 2 at STP/100g. The composition of the wire is given in Table 1.

Table 1 Composition of welding wire (S/83/578) and testpiece plate (S/83/488), wt%

Submerged-arc flux

For the purpose of this collaborative trial, a fully fused basic submerged-arc, welding flux, for DC welding was obtained. To reduce exposure to atmospheric humidity, portions of the flux were sealed into tins. Each participant in the trial received six tins, one for each weld, with sufficient spare flux to establish welding conditions. The flux was used as-received and surplus flux remaining after each weld was not recycled. Flux composition is given in Table 2.

Table 2 Composition of fused flux (S/83/246), wt%

Testpiece assembly

Steel bar stock to BS970 Pt 1 1972 070 M20 of 30 x 10mm cross section was used, the composition being given in Table 1. The surface was ground to remove any surface irregularities, but maintaining the dimensional tolerance stated in the draft IIW standard. [1] Lengths of bar were circulated to each participant in the trial.

Production and analysis of the weld specimens

From earlier work [4] it was known that variations in results stem from a combination of effects arising from choice of welding parameters and specimen cleaning method. In order to obtain uniform operation of the proposed procedure it was agreed to use a common laboratory work sheet giving detailed instructions for carrying out the test. The welding parameters which were to be measured and the diffusible hydrogen results obtained were recorded on three result forms (ref. [3] ), and the nominal welding parameters are given in Table 3. Six test welds were made in each of the co-operating laboratories. The welds were cleaned by vigorous wire brushing, the run-on and run-off pieces were detached from the specimens which were then analysed for diffusible hydrogen by collection over mercury at room temperature.

Table 3 Nominal welding parameters for submerged arc test welds

Atmospheric humidity was recorded

Results

The results obtained for the diffusible hydrogen measurements on deposited and fused weld metals are given in Table 4. The results from Laboratory B are much lower than those obtained by the other four laboratories. Examination of the weight of weld on the run-on and run-off pieces showed that the instruction regarding a 200mm weld traverse had not been followed and an excessive weight of deposit was made on the run-off piece. The delay in quenching the samples which was caused by the extra welding time could account for hydrogen loss and the low results obtained. It was decided that the Laboratory B results should be omitted from the reproducibility calculations.

The frequency distribution curve for the results obtained by Laboratories A, C, D and E is shown in Fig.3 and hydrogen evolution curves in Fig.4.

Reproducibility index

The reproducibility of an analytical method is defined in BS 4237 as '. . . the consistency between replicate tests, indicating the measure of agreement between values obtained when using a particular method . . .' The term 'reproducibility' should not be taken to imply accuracy, which can be assessed only when a true value is known. Thus the reproducibility index of an analytical method represents the range about an overall mean within which 95% of the results by any one operator could be expected to fall. For this purpose, the range is defined by plus or minus twice the method standard deviation.

The reproducibility achievable by one operator was studied by considering the between-laboratory and within-laboratory variances for both deposited and fused weld metal diffusible hydrogen measurements.

Table 4 Diffusible hydrogen measurements on deposited and fused weld metal. (Final figures after 21 days)

Table 5 Diffusible hydrogen results, ml/100g

Table 5 shows the between-laboratory and within laboratory mean squares, variances and standard deviation analyses.

Discussion

Statistical evaluation of the results

The results in Table 4 show that both the deposited and fused weld metal diffusible hydrogen results may be separated into two different levels, those obtained by Laboratories A and C, and those obtained by Laboratories D and E. Examination of the welding and sample weight records did not indicate any reason for this difference and so did not justify rejection of any results on a technical basis. It was noted that there was an overlap of the range in the two groups of results for the deposited weld hydrogen values, but not for the fused metal results. If the collaborative trial had been conducted on the same basis as most trials of an analytical method, i.e. with a homogeneous sample supplied to each laboratory, then the results would probably have been rejected and a repeat trial carried out. However, as with an earlier manual metal arc trial ^[4] the programme of work also included the variations in the sample production stage; since this is inherent in the practical usage of the whole procedure, it was considered valid to carry out a statistical examination to indicate the reliance which may be placed upon the results obtained.

The method standard deviation is obtained from the formula

where Sb is the 'between-operator' standard deviation and Sw is the 'within-operator' standard deviation. 95% of results obtained by any one operator should be reproducible to within two standard deviations of the overall mean value derived from all laboratories ( i.e.

2S). (For further information see BS 4237).

The findings of this collaborative trial show that for the determination of diffusible hydrogen in submerged-arc deposited weld metal, the method standard deviation is 0.81 ml at 7.9ml STP/100g, and 95% of results by one operator will fall within the range 6.30 to 9.54ml STP/100g. Similarly the fused weld metal method standard deviation is 0.41ml at 2.5ml STP/100g and the range for 95% of results is 1.69 to 3.31ml STP/100g. When these findings are expressed as percentage of content then we have a reproducibility of ±20% of content for the deposited metal and ±32% of content for the fused metal. This compares well with the figures of ±18% and ±30% respectively which were obtained during the MMA study. [4]

These statements apply only to the levels of hydrogen studied. The similarity in the reproducibilities obtained for the two different welding processes indicate that with both procedures for diffusible hydrogen measurement equal reliance may be placed upon the results. Although these reproducibilities may not appear to be good in absolute terms, it is pointed out that they are equivalent to the reproducibilities of measurements of other impurities in steels occurring at the low parts per million concentration level, [6] Table 6. Because of the wide range of results which is represented by this high value for the reproducibility index, it is customary to make triplicate welds when preparing samples for hydrogen analysis. When the mean of triplicate results is reported, then the standard deviation, and hence the reproducibility index is reduced by a factor of 1/ √3.

Table 6 Reproducibility of the method for diffusible hydrogen compared with that for other analytical methods

Hydrogen evolution

Figure 4 shows that hydrogen evolution is incomplete after three days. A clean specimen, free of surface oxide, takes about ten to twelve days, but Laboratory A specimens took 21 days, perhaps because of insufficient cleaning prior to analysis.

Sources of hydrogen in the weld

The hydrogen potential of the degassed wire was 0.8ml at STP/100g which would contribute a reasonably constant but low amount of hydrogen to the weld deposit which was analysed. It may be surmised therefore that the hydrogen content of the weld was derived mainly from either the flux or ambient humidity. The latter was recorded by each participant, and it was noted that the welds made with the highest ambient humidity (from Laboratory E at 91%) recorded the lowest diffusible hydrogen content in the weld. Therefore it is unlikely that ambient moisture would make any significant contribution to the weld hydrogen, but this aspect should be further investigated.

If the contribution of the wire and ambient moisture to the weld pool hydrogen is minimal for the reasons given above, then it is apparent that the flux itself plays an important role in influencing the level of hydrogen in the weld. If a flux contains only 0. 1% by weight of moisture, then this is equivalent to 124ml of hydrogen at STP/100g of flux. It is thought that the procedure which is described in this article may be used to examine the behaviour of a submerged-arc flux with respect to weld hydrogen in the same way that ISO 3690 [2] is used to categorise the weld hydrogen characteristics imparted by the coating of a MMA electrode.

Use of the new design of welding jig

The use of a 10mm thick testpiece imposes limitations upon the arc energy which can be used. This study showed that an energy of 3kJ/mm can be accommodated with the new design of welding jig without the weld breakthrough described in ref. [3] . This agrees with the limit suggested by the IIW draft standard. [1]

Conclusions

1. The work described is considered to provide a sound basis for the standardisation of a method for the determination of diffusible hydrogen in submerged arc test welds. Further studies of different hydrogen concentrations are desirable.
2. At the hydrogen level examined of about 8ml/100g the reproducibility of the procedure described is ±20% of the content for deposited metal. This improves to ±10% of content as determined by the mean of triplicate results.
3. The reproducibility can be equated with that of the standard method for diffusible hydrogen in MMA test welds.
4. Under the welding conditions used, the new design of welding jig can accommodate arc energies up to 3kJ/mm on the 10mm thick testpiece assembly.
5. As with MMA weld hydrogen measurements, evolution is not complete after three days and up to 14 days collection time may be required. If the weld specimen is not cleaned sufficiently well, then up to 21 days may be required.

Acknowledgements

The author acknowledges the assistance and advice of his colleagues, J A Barlow, M Taylor and Mrs D F Pargeter.

The following individuals took part in the collaborative work described in this report: Dr B Chew (CEGB Marchwood Engineering Laboratories); Dr D N Shackleton and J I Duce (Head Wrightson Teesdale); K Calder and C Herrod (NEI International Combustion); C R Dye and D H Parker (The Welding Institute).

The United Kingdom Atomic Energy Authority and the Department of Industry provided financial assistance towards the costs of this project.

References