Analysis of all weld metal manual metal arc test pads and thick steel plate using optical emission spectrometry - an interlaboratory examination

TWI Bulletin, May 1987

Sheila Stevens, LRSC, is a Senior Research Chemist in the Materials Department.

Although information presented on a chemical analysis certificate is often regarded as absolute, the collaborative trials reported here demonstrate that results obtained using optical emission spectrometry can vary by as much as ± 60% for elements present at low levels (e.g. sulphur).

The successful industrial use of steel depends on knowledge of its chemical composition in relation to its specification and intended use. The steel composition, as supplied by the producers, is often a ladle analysis and is the chemical composition before casting occurs. This analysis may well describe an average composition. However, during casting and after solidification, changes in chemical composition occur which, in addition to elemental segregation, especially of elements such as C, S, P and Mn, may lead to heterogeneity in wrought steel products. Such heterogeneity can present difficulties in obtaining a representative chemical analysis, and this is particularly relevant to the fabricating industry, which as a consumer of steel products, is making increasing use of optical emission spectrometry (OES) for the analysis of wrought and other materials. Optical emission spectrometry is often the only economic alternative to conventional chemical analysis when confirmation of steel composition is required for quality assurance and other allied purposes.

However, sampling and calibration procedures can vary between OES users and at present there is no guarantee that different operators will produce consistent results for a given steel sample. In the UK, variations between operators may occur because of the lack of a national standard method for OES calibration. Although foreign standards are available, these are primarily aimed at type analysis encompassing only a narrow range of product composition and are not directly of general application to a wide range of element contents.^[1-3]

The validity of any analysis method is dependent in large part on the integrity of the sample. Where weld metal composition approval is required this entails making a sample which is, as far as possible, 100% weld metal. So the chemist who is responsible for defining sample taking and sample preparation procedures, also becomes involved in the actual manufacture of the sample. The following work relates specifically to MMA welding but could be extended to other processes.

A survey of national standards for classification of welding consumables^[4] according to the properties of the weld metal produced reveals a large number which specify preparation of a test pad for chemical analysis. Methods of preparation of such a pad vary with welding process and country, although in the UK there is no standard giving details of test pad preparation for mild steel electrodes. In fact the UK standard for these electrodes, BS 639,^[5] has no chemical analysis requirements at all, apart from hydrogen. Equivalent American, Australian, Canadian and French standards for mild steel all have chemical analysis requirements and give methods of preparation of a chemical analysis test pad. There can be changes in composition because of dilution from a preceding layer of weld metal, or perhaps because of changes in welding parameters if they are not adequately controlled. Thus the weld sample may be heterogeneous. In assessing the OES technique for analysing weld metal pads directly, it is recognised that the heterogeneity of the sample will diminish the reproducibility of the results and there may also be differences between weld pads. The reproducibility can be improved by remelting the weld sample, to form a homogeneous cast button.

In the steel industry, standard methods exist for sampling molten and wrought materials for OES^[6-8] and chemical analysis.^[6-9] Thick steel plate can present special problems for direct analysis by OES. It is well known that the surface C content may differ from that of the bulk C content and this difficulty can be overcome in some cases by analysing the material on a through-thickness plane. However, there may still be problems of inclusions such as manganese sulphide, and of heterogeneity, especially in the centre of concast steels. Modern OES spark source units reduce the effects of inclusions on analysis by means of powerful local remelting immediately prior to the analysis. The problem of heterogeneity can only be overcome by analysing a representative through-thickness sample, either by several direct analyses, or by analysing a button prepared by remelting a solid sample or millings taken across the thickness.

Two types of remelting unit are commonly used to produce buttons for OES analysis, incorporating either an argon arc to remelt the sample in a copper hearth, or radio frequency (RF) heating to remelt the sample in a ceramic crucible. To minimise losses of elements such as C and Mn, it is necessary to exclude air from the furnace prior to remelting by flushing with Ar or by evacuation followed by filling with Ar. The use of a deoxidant such as Zr, Al or Ti to combine with any oxygen present in the sample is also desirable, especially for rimming steels.

When remelting takes place in a ceramic crucible, elemental changes can occur which may be additive because of pick-up from the crucible or losses may occur from reaction of reactive elements with the crucible. Remelting can be used to convert samples which are unsuitable for direct OES analysis, such as wires, drillings, and thin sheet, into a suitable sample form, at the same time overcoming problems of heterogeneity. The successful use of remelting to produce standards for OES calibrations has been demonstrated in other work.^[10]

Another problem facing users of OES for product analysis is that they may be required to analyse a wide range of compositions but are unable to carry out 'type analysis' using the narrow range 'in-house' reference materials available in the steel industry. Commercially available certified reference materials (CRMs) are not suitable for producing wide range calibrations.

Two collaborative investigations were undertaken to study the effects of sample preparation and OES calibration procedure on the analysis of carbon-manganese steel MMA test pads^[11] and the applicability of OES to the analysis of three compositions of thick steel plate,^[12,13] including low alloy material.

Materials

The weld metal trial was in two phases. In each phase six carbon-manganese steel MMA weld pads were prepared and circulated to the participating laboratories (four in Phase 1, and five in Phase 2) in turn, together with a remelt button prepared from each weld and a set of remelted CRMs for calibration in Phase 2.

Three plates were used and, after sampling, as shown in the Figure which gives details for Plate A, were distributed to each of seven laboratories. Samples 8A, 8B and 8C were used for conventional chemical analysis.

Experimental details

Weld pad preparation

Two sets of cylindrical weld pads were prepared on the ends of pieces of 38mm diameter mild steel bar using the same batch of AC manual metal arc electrodes, type E7016, for both sets. The first set, used for Phase 1, contained six layers of weld metal. The second set, used for Phase 2, contained twenty layers of weld metal to provide sufficient additional samples for remelting and for chemical analysis.

Preparation of remelt buttons from the weld pads

For Phase 2, approximately 70g were milled from the top of each weld pad, 45g were used to produce a remelt button and the remaining 25g were retained for chemical analysis. The millings were remelted in an argon-arc furnace, using approximately 0.04g of Zr deoxidant, and the top surface of the remelt buttons was ground to provide a flat surface for analysis.

Preparation of remelt buttons from plates

Six buttons weighing about 50g each were produced from the samples provided (samples A1 to A6 in the Fig.). Four of the seven laboratories possessed argon-arc remelters, and one of these prepared buttons for two of the other laboratories without remelting facilities, so that a total of six laboratories carried out analyses on remelted buttons. Three of the four remelters removed air from the furnace by flushing with argon and the other evacuated the air prior to argon filling. This last remelter was used to produce the buttons for chemical analysis. All four laboratories used a deoxidant, either 0.1%Zr remelted with the sample, or Ti melted in the furnace chamber prior to remelting the sample.

Preparation of remelted calibration CRMs

Six CRMs, covering the composition range of the weld pads, were prepared by one laboratory by pressing CRM millings into a cylindrical block and remelting in an argon-arc furnace using Zr deoxidant as for the weld remelt buttons. Duplicate buttons were prepared for each CRM and details of the compositions before remelting are given in Table 1.

Table 1 Calibration standards

OES analysis

In Phase 1 of the weld metal trial, each pad was analysed four times on a freshly prepared surface (ground or linished) with the instrument set up immediately prior to the analysis using the normal in house operating procedure. This was repeated not less than 24hr later, giving a total of eight results per sample for each element. For Phase 2 each laboratory prepared calibration graphs for C, S, P, Si and Mn, each point on the graph being the mean of six sparks (three on each of the duplicate buttons) obtained from six separate runs on the CRMs. The weld pads and remelt buttons were then analysed using these calibrations. Each sample was analysed six times, once on each of six runs, resurfacing after three analyses. The same surface preparation procedure was used for the remelted standards and the weld pads and buttons.

For the plate trial, the instruments were set up according to the normal operating procedures of the individual laboratories and the samples were analysed to give a total of six sets of data, each obtained from a separate run, setting the instrument up for each run. For each run six sparks were made on the side which was then linished to prepare for the next run. This was not applicable to the top and bottom analyses because linishing between each run would have removed the surface. Accordingly only one spark was made on the tops and bottoms of the blocks for each run. Each run therefore provided the following data:

Top - one figure (e.g. T1 (see Figure) for run 1, T2 for run 2 etc);
Bottom - one figure (as above);
Side - six figures (S1 to S6);
Buttons - six figures (A1 to A6).

Chemical analysis

Chemical analysis was performed on millings taken from the weld pads and from the through-thickness and remelted buttons for each plate. In the latter case the sample for analysis consisted of combined millings from all six buttons. Analysis was carried out by only one laboratory, apart from Al which was determined by two laboratories.

Results

The plate results from this study were obtained from analyses of only three steels and may not be applicable to all steels. In particular, steels cast by different procedures, e.g. large weight castings or concast steels, suffer segregation to differing degrees.

The data obtained were subjected to detailed statistical analysis (see Appendix for terminology[14,15]) to compare the performances of the various participating laboratories, and in particular to identify any major causes of variation in results. The OES results for the weld metal and plate trials are summarised in Tables 2a and 2b respectively, giving the means, ranges and reproducibility indices (see Appendix) for all laboratories, and illustrative data on one plate (A) derived by only one laboratory. Table 2b presents the data for the plate side and button analyses. Results for the top and bottom plate locations were similar, apart from carbon which showed more variability, Table 2c, presumably because of segregation and/or decarburisation, etc. Results of the chemical analyses are given in Tables 3a and 3b.

Table 2a OES results for weld metal

Table 2b OES results for plates

Table 2c Carbon from top and bottom samples in plate trial, OES results

Table 3a Chemical analysis results for weld metal

Table 3b Chemical analysis results for plates, wt%

Assessment of OES results

Repeatability

As described in Ref.[11] and [12], it was clear that all the laboratories were capable of working to the same instrumental precision at low concentration levels, while at higher levels there were slight variations possibly because of different instruments or operating procedures. So the instrument capability was generally considered satisfactory.

Results obtained in the weld metal trial showed that the repeatability of Phase 1 was better than that of the Phase 2 pads, suggesting that the former pads were more homogeneous.

A study of the within-laboratory standard deviations (sw) using OES was carried out by Bramhall et al[16] using mild steel and low alloy spectroscopic CRMs. Carbon, S, P, Si, Mn, Ni and Cr were determined and the results showed that, with the exception of S and P, sw increased with increasing content. Considering all elements other than S and P, Bramhall[16] found that the 95% confidence limits for within-laboratory variation were ± 0.008% at 0.1%, ± 0.014% at 0.5%, ± 0.018% at 1%, and ± 0.024% at 2%. An analogous appraisal of the plate data obtained in this study gave 95% confidence limits of ± 0.008% at 0.1%, ± 0.022% at 0.5%, ± 0.042% at 1%, and ± 0.081 % at 2%. Except for the lowest level, the results are all higher than those of Bramhall et al, but this discrepancy is to be expected. Apart from involving only one laboratory, Bramhall et al used spectroscopic CRMs for their tests and these materials are specially selected for their homogeneity. On the other hand, the present round robin involved seven laboratories which would be expected to increase sw as each laboratory is using different instruments and calibrations. The effect of different calibrations would be particularly important in the analysis of low alloy steels where corrections are required to allow for interelement interactions.

Reproducibility

In Phase 1 of the weld metal trial, the reproducibility values were higher than the repeatabilities, the 95% confidence limits for reproducibility being typically two or three times greater than those for repeatability, while in Phase 2 they were comparable. The combined effects of the within and between laboratory variations are given by the reproducibility indices in Table 2a from which it can be seen that considerable variation in analysis can arise for Si and Mn.

As would be expected, the data from the plate trial showed that the reproducibility of the results derived for individual elements was substantially worse than the repeatability, the 95% confidence limits being about 1.5-3 times greater in the former case. The reproducibility indices in Table 2b show that, for example, the carbon content of plate A as determined on a through-thickness section by a number of laboratories employing IDES would be expected to be 0.143 ± 0.018% at the 95% confidence level. Similarly the sulphur content of plate B would lie in the range 0.0060 ± 0.0029%; A comparison of the reproducibility indices with those published for standard chemical techniques showed the latter to be much lower. However, these standard techniques are referee methods, while OES was tested as a routine method in common use. It is also true that since OES is a secondary technique it cannot improve upon the accuracy of the CRMs used for calibration.

Effect of sampling procedure

The weld metal pad and button results in Phase 2 were similar and unaffected by the sampling procedure. The reproducibilities for the pads and buttons were comparable for elements present at low levels, i.e. C, S, P, as were the repeatabilities, but the analysis of buttons gave considerably better reproducibilities and repeatabilities than pads for Si and Mn. In all cases the repeatability was less than, or similar to, the reproducibility. This is as expected since the former is a measure of within laboratory performance, whilst the latter measures between laboratory performance. The advantages in analysing buttons are reflected in the improved reproducibility indices for Si and Mn shown in Table 2a.

As noted above, the plate results obtained for four different sampling procedures, namely direct analysis on the top, bottom, or through-thickness surfaces, or analysis of remelted buttons, were the same for all elements except C, where the results on the outer surfaces differed from those on the buttons. Variations in C content at the surface are not uncommon and have been reported elsewhere.[17] Some caution is necessary, in that the results for S and Mn may not be typical as these elements can cause problems for direct analysis because of heterogeneous distribution of sulphide inclusions, and in such instances the analyses can be greatly improved by analysing remelted buttons.[17,18]

The reproducibilities obtained were in general similar for sides and buttons at low concentration levels, but at high levels (of e.g. Si, Mn, Ni, Cr, Mo) the buttons exhibited slightly better reproducibility. The repeatabilities for the sides and buttons were comparable, and better than the reproducibilities. The reproducibility indices showed that method performance may be improved by analysing buttons as opposed to plate sides or surfaces, primarily for elements present at higher concentration levels such as Si, Mn, Ni, Cr and Mo.

Discussion

Accuracy of the OES analyses

The accuracy of OES data can be assessed only by comparison with results from conventional chemical methods. The chemical results were obtained by one laboratory (with the exception of Al), and it is assumed that they are correct and without bias.

Weld metal trial

The effect of calibration procedure and remelting on the accuracy of the results for pads and buttons

As the same batch of electrodes was used for each phase it is possible to include Phase 1 pads in an assessment of accuracy, since although these pads were not analysed by chemical methods the composition should be the same as for the Phase 2 pads.

For Phase 1 pads three of the four laboratories are in reasonable agreement with the true value for carbon while one is high. In Phase 2 all the results for both pads and buttons are high. Sulphur results agree quite well with the true value for both phases for the pads and buttons, with the exception of results from one laboratory which were consistently low.

A true value for phosphorus was only determined on one sample. However, assuming that the compositions of the samples are the same, all the phosphorus results on pads and buttons are in agreement with the true value, with the exception of results from one laboratory for each of the Phase 1 and 2 pads.

For Si also, a true value was only determined for one sample, but assuming this to be representative then in Phase 1 two of the four laboratories agreed with the true value and two were low, while in Phase 2 the pad and button results were in satisfactory agreement apart from one laboratory whose results were low for the pads and buttons.

An assessment of the accuracy of Mn results for Phase 1 pads is more difficult because of the variation in true content of individual samples. However, it would appear that two laboratories agree reasonably well with the true value while one is high and one is low. In Phase 2 the spread between the overall means for each laboratory has been considerably reduced compared to Phase 1 and the pads and buttons are now fairly near to the true value. However, an examination of individual sample means for each laboratory shows that overall the results for the buttons are more accurate than those for the pads.

Therefore the use of a uniform calibration procedure in Phase 2 improves the accuracy for Si and Mn, while sulphur and phosphorus are unaffected, but slightly reduces the accuracy for carbon. But this can still be considered acceptable, taking into account that there is often a wide range of individual values on certificates supplied with CRMs (Table 4).

Table 4 Typical variability of analyses quoted for commercial certified reference materials

*British CRMs supplied by Bureau of Analysed Samples

The production of remelt buttons from the weld pads further improves the accuracy for Mn and overcomes the problems of through-thickness variation in Si and Mn composition for weld pads.

The spectrographic results obtained using the standardised calibration procedure, and the remelt buttons are in satisfactory agreement with those produced by standard chemical methods.

Plate trial

Comparisons of the OES results with chemical results showed that in general the OES results were satisfactory for C, P (depending on plate composition), Si, Ni at low levels, Mo and V. However, within these results, some laboratories were consistently high or low. The results for S, Mn, Ni at high levels, Cr, Cu and Al were disappointing. Only about half of those for Mn and Cu were in agreement with the chemical value and the level of agreement for Ni and Cr decreased as the content increased.

Sulphur results were particularly unsatisfactory, with only one laboratory consistently in agreement with the true value, and the same was true for Al (although there are doubts associated with the chemical analysis method for this element in that fusion may not be complete). It was noticeable that laboratories tended to produce results which were either consistently high or low, rather than results scattered above or below the true value, indicating a calibration bias. This was especially evident for the sulphur analyses, where all except two laboratories were high. All the instruments used the same sulphur line, 180.7nm, which is known to be subject to interelement interferences caused by overlapping spectral lines.[19-21] If the calibration techniques do not fully compensate for these effects, high results will be obtained. Plates A, B and C contain relatively large amounts of alloying elements such as Mn, Ni, Cr and Mo which may interfere. Of these elements, Mn has the largest effect on the sulphur line which could explain why the highest differences between the true value and the OES figures were obtained on Plate C which contained nearly 2% Mn. It is perhaps relevant that only the laboratory which was consistently in agreement with the true value routinely uses OES to determine sulphur in the range 0.003% to 0.005%, and the others would normally carry out a combustion sulphur determination for sulphur levels below 0.01%. Difficulties in the determination of low levels of C, S, P, Cr and Cu by some laboratories occurred because of a lack of low level CRMs, and accuracy may have been improved if narrow range calibration curves were prepared and more account taken of interelement interferences. However, existing OES standards[1,2] as exemplified by ASTM E 403 and E 415 do not take full account of interference corrections. With regard to the elements covered in this round robin, they are further limited in that ASTM E 403 does not include C, S or P, and the detection limits for Mn, Ni, Cr, Cu and Al are too high for current day analytical demands. The comment regarding high detection limits also applies to ASTM E 415 with the exception of V, although this standard does cater for all the elements measured by the collaborative trial. Modern instruments are capable of producing extremely low detection limits providing suitable CRMs and calibration techniques are employed.

The study by Bramhall et al[16] compared OES results on their solid spectroscopic CRMs with those on the certificates and obtained good agreement between the two, but they pointed out that OES results are really determined by comparative measurement with the CRMs used to calibrate the instrument and that the accuracy of results is therefore dependent on the validity of the standard samples. An examination of the certificates supplied with standard samples shows that there is often a considerable range of individual figures reported (Table 4).

The effect of remelting on accuracy

Chemical analyses of the buttons showed that Ar arc remelting can be used successfully to convert thick plate into a suitable sample for OES analysis, although there may be variations over the area of the plate. The small C and Mn losses obtained were not evident from the OES analysis. The laboratory which prepared the remelt buttons for wet analysis employed a system of calibration using remelted standards in order to correct automatically for any element loss. All other elements were unaffected by the remelting procedure apart from Cu where there was a pick-up of about 0.002% from the Cu hearth. However, low results from one laboratory for C, Mn and Al did show that the successful use of this technique depends upon the correct operation of the remelting equipment.

Several studies have been carried out on the suitability of Ar arc remelting for the preparation of buttons for OES analysis from drillings, millings and from solid materials.[10,17,22] Two of these studies used CRM millings for the tests.[10,22] The excellent reproducibility of the technique was demonstrated by results obtained on ten replicate buttons and was shown to be as good as that of listed chemical methods of analysis.[15,22] Investigations have shown, either by comparison of OES button results with solid results or by chemical analysis of the buttons, that Ni, Cr, Mo, V, Cu and Al are unaffected by the remelting procedure as was the case in this round robin. Small reproducible losses of C[10,17,22] and Mn have been reported,[10,22] which were confirmed by chemical analysis,[10] and an apparent loss of S[10,22] which was not shown by chemical analysis.[10] Apparent gains have also been reported for Si and P[10,22] but these were not substantiated by chemical analysis.[10]

All of these investigations indicated that accurate analyses could be performed on remelted buttons for all of the above-mentioned elements, provided that appropriate calibrations were used to compensate for changes in elemental composition or for differences in instrumental response. A study[23] on the use of centrifugal casting for preparing buttons for OES analysis gave variable reproducibility on ten buttons prepared from a setting-up sample, which was attributed to the effect of a typical composition on the remelting process. The same study also showed that small reproducible losses were obtained for C and Mn, and irregular losses for Al. The other elements analysed in the current round robin were unaffected by the centrifugal remelting procedure.

Implications

The present work has shown that, depending on the element concerned, OES analysis can give results broadly comparable with those obtained from standardised methods. However, in a number of cases the former data show significantly more scatter, especially in terms of between laboratory variations. To a large extent, industry tends to regard information on a chemical analysis certificate as absolute, and the possibility of variation is not recognised. These collaborative trials have clearly demonstrated that appreciable differences in reported composition can arise between laboratories operating OES equipment. This is also true for chemical analysis, as evidenced by the range of individual element figures cited for CRMs, but the discrepancy may well be greater for OES analysis. In most situations, the reliability of OES results will be perfectly acceptable, but the present results show that the range of a given element may well be no better than '± 10% of the amount present for major elements and as much as ± 60% for elements present at low levels, such as sulphur.

In large part, the scatter of data in Table 2a for Phase 1, and in Table 2b, stems from between laboratory variations. This almost certainly originates from different calibration and operation procedures in the participating laboratories, since more consistent analyses were obtained when a uniform calibration procedure was used, or when results from only one laboratory were considered. It must therefore be concluded that a need exists for an internationally standardised calibration procedure, with full allowance for interelement spectral interferences to minimise the variability in reported analyses as identified in the present work.

Summary

1. A uniform calibration procedure using remelted CRM millings can reduce the variations between laboratories for the analysis of weld pads as compared with in house procedures, giving accurate results and improving the accuracy for certain elements. Accuracy and the reproducibility index can be improved further by analysing remelted buttons in some instances.
   
   
2. Using modern spectrometers incorporating a high repetition rate source or a high energy pre-spark, it is possible to analyse thick low alloy steel plate, either directly on the through-thickness plane or as remelted buttons. However, the plate surfaces should be avoided. It is considered that remelted buttons are preferred to obtain a representative through-thickness analysis and to minimise the reproducibility index.
   
   
3. While chemical analyses of steel carried out by OES will be sufficiently reliable for many practical purposes, reproducibility indices of individual elements may be about twice those for conventional chemical analysis methods.
   
   
4. Considering OES results obtained by different laboratories, the range reported for major elements may be ± 10% of the amount present, and possibly ± 60% for elements present at low levels.
   
   
5. The main source of variability in OES data is interlaboratory inconsistency, mainly because of the lack of a standard calibration procedure.
   
   
6. A need exists for standardising the sample preparation and instrument calibration technique. The standardised calibration technique should include the application of interference corrections and a means of regular assessment of instrument performance in terms of accuracy and reproducibility.

Appendix - Terminology[14,15]

where S_b is the between-laboratory standard deviation and S_w is the within-laboratory standard deviation.

Acknowledgements

The work in this article was funded jointly by Research Members of The Welding Institute and the Minerals and Metals Division of the UK Department of Trade and Industry. Members of participating laboratories are thanked for their co-operation.

References

1. ASTM E 403-82: 'Standard method for optical emission spectrometric analysis or carbon and low alloy steel by the point-to-plane technique'. Publ American Society for Testing and Materials.
2. ASTM E 415-85: 'Standard method for optical emission vacuum spectrometric analysis of carbon and low-alloy steel'. Publ American Society for Testing and Materials.
3. JIS 1253-1976 'Method for photo-electric emission spectrochemical analysis of iron and steel'. Publ Japanese Standards Association.
4. Stevens S M: 'Sample preparation for ferrous weld metal composition analysis'. Welding Institute Members Report 207/1983.
5. BS 639: 1986 'Covered carbon and carbon manganese steel electrodes for manual metal arc welding'. British Standards Institution.
6. ASTM E 59-78 (Reapproved 1982), 'Standard method of sampling steel and iron for the determination of chemical composition'.
7. JIS G 0303-1972: 'General rules for inspection of steel'. Publ Japanese Standards Association.
8. JIS G 0321-1966: 'Product analysis and its tolerance for wrought steel'. Publ Japanese Standards Association.
9. BS Handbook No 19: 'Methods of sampling and analysis of iron, steel and other ferrous materials'. Publ British Standards Institution.
10. Stevens S M: 'The effect of remelting on OES calibrations for ferritic steels'. Welding Institute Members Report 246/1984.
11. Welding Industry Chemists Group Report. 'The preparation and analysis of all weld metal manual metal arc test pads'. Project 9409, January 1984, publ The Welding Institute.
12. Stevens S M: 'An interlaboratory examination of the analysis of thick steel plate using optical emission spectrometry'. Welding Institute Members Report 312/1986.
13. Stevens S M: 'How good are chemical analyses of steels?' Advanced Materials and Processes inc Metal Progress 1986 130 (4) 48-51.
14. BS 4237: 1967: 'Report on the reproducibility of methods of chemical analysis used in the iron and steel industry'. British Standards Institution.
15. BS 5497: Part 1: 1979: 'Precision of test methods - Part 1 - Guide for the determination of repeatability and reproducibility for a standard test method'. British Standards Institution.
16. Bramhall P S, Greenfield A A and Scholes P H: 'The precision of direct-reading optical emission spectrometric analysis'. BISRA Report MG/CC/560/71.
17. Stamp C, McKenzie P and Harrison T S: 'Spectrographic analysis of mild and low alloy rolled steel using the argon arc vacuum melting technique'. Metallurgia and Metal Forming 197340 (7) 222-5.
18. Jenkins N and Carter G J: 'Thick plate analysis using a direct reading spectrograph'. Welding Institute Research Bulletin 1972 13 (4) 107-111.
19. Lude M and Vialette B: 'The use of binary alloys for systematic study of interferences in optical emission spectrometric analysis of steel'. Analysis 1980 8 (9) 403-9.
20. Endo Y et al: 'Multiplicative interferences of coexistent elements on emission spectrochemical analysis'. Trans Iron Steel Inst Japan 1980 20 223-7.
21. Stevens S M: 'Calibration of a direct reading optical emission spectrograph'. Welding Institute Research Bulletin 1982 23 (6) 190-3.
22. Tunney A A and Hughes H: 'Development of argon arc melting techniques for the preparation of solid specimens for instrumental analysis from drillings'. BSC Report, GS/TECH/251//1/73/C.
23. Brookes L and Tunney A A: 'An assessment of the performance of the MBLE PV8910 high frequency melting unit for the preparation of solid samples from drillings'. BSC Report, GT/TECH/505/1/75/C.